Abstract
The aim of this work is to discuss the proportionality of the multifractal measures. We will prove that the ratio of the multifractal measures is bounded. In addition, for a class of homogeneous Cantor sets, we find an explicit formula for their multifractal Hausdorff and packing function dimensions and discuss some interesting examples.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\mu \) be a probability measure on a metric space X. The Hausdorff multifractal spectrum function, \(f_\mu \), and the packing multifractal spectrum function, \(F_\mu \), of the measure \(\mu \) are defined respectively by
where
During the past 25 years there has been an enormous interest in computing the multifractal spectra of measures in the mathematical literature. Particularly, the multifractal spectra of various classes of measures in Euclidean space \(\mathbb {R}^n\) exhibiting some degree of self-similarity have been computed rigorously. The reader can be referred to the paper [13], the textbooks [8, 15] and the references therein. Some heuristic arguments using techniques of statistical mechanics (see [11]) show that the singularity spectrum should be finite on a compact interval, noted by \(\text {Dom}(\mu )\), and is expected to be the Legendre transform conjugate of the \(\tau _\mu \)-function, given by
where the supremum is taken over all centered packing \(\big (B(x_i,r)\big )_i\) of \(\text {supp}(\mu )\). That is, for all \(\alpha \in \text {Dom}(\mu )\),
The multifractal formalism (1.1) has been proved rigorously for random and non-random self-similar measures, for self-conformal measures, for self-affine measures and for Moran measures. We notice that the proofs of the multifractal formalism (1.1) in the above-mentioned references (see for example [2, 3, 13, 14, 21, 23,24,25] and references therein) are all based on the same key idea. The upper bound for \(f_\mu (\alpha )\) is obtained by a standard covering argument, involving Besicovitch’s covering theorem or Vitali’s covering theorem. However, its lower bound is usually much harder to prove and is related to the existence of an auxiliary measure (Gibbs measures, Frostman measures) which is supported by the set to be analyzed. In an attempt to develop a general theoretical framework for studying the multifractal structure of arbitrary measures, Olsen [13], Pesin [15] and Peyrière [16] suggested various ways of defining measures analogous to those of Gibbs measures in very general settings. For an arbitrary Borel probability measure \(\mu \) on \(\mathbb {R}^n\), they introduced two parameter families of measures,
based on certain generalizations of the Hausdorff measure and of the packing measure. One of the main importance of the multifractal measures \(\mathscr {H}_{\mu }^{q, t} \text { and } \mathscr {P}_{\mu }^{q, t}\), and the corresponding dimension functions b and B is due to the fact that the multifractal spectra functions \(f_\mu \) and \(F_\mu \) are bounded above by the Legendre transforms of b and B, respectively, i.e.,
These inequalities may be viewed as rigorous versions of the multifractal formalism. Furthermore, for many natural families of measures we have
see for example [2, 3, 5, 12,13,14, 21, 23,24,25,26]. It is clear by comparing the definitions of the measures \(\mathscr {H}_{\mu }^{q, t} \text { and } \mathscr {P}_{\mu }^{q, t}\), and definition of the \(\tau _\mu \)-function which appears in the multifractal formalism that b(q) and B(q) are mathematically rigorous versions of \(\tau _\mu (q)\), and that the one-parameter families
play the role of the auxiliary measures \(\left\{ \mu _{q};\; q \in \mathbb {R}\right\} \). In particular, we would expect that the measures \(\big \{\mathscr {H}_{\mu }^{q, b(q)};\; q\in \mathbb {R}\big \}\) and \(\big \{\mathscr {P}_{\mu }^{q, B(q)};\; q \in \mathbb {R}\big \}\) have similar properties to those of the auxiliary measures \(\left\{ \mu _{q};\; q \in \mathbb {R}\right\} \). This has been proved rigorously for self-similar, quasi self-similar, homogeneous Moran measures, self-conformal measures and for arbitrary measures.
Even though it seems rather unlikely that the multifractal Hausdorff and packing measures are proportional in general. In this paper, we will prove that the ratio of the measures \({\mathscr {H}^{q,b(q)}_{\mu }}\) and \({\mathscr {P}^{q,B(q)}_{\mu }}\) might still be bounded, i.e., there exists a number \(0< c_q < +\infty \) such that
which provide a positive answer to Olsen’s questions [13, Question 7.6] and [14, Question 4.1.12] in a more general framework. We give also a reasonable lower and upper bound for the multifractal Hausdorff and packing measures of homogeneous Moran sets. In particular, these results find an explicit formula for their multifractal Hausdorff and packing function dimensions. We note that our results, due to the use of the multifractal Hausdorff and packing measures introduced in [13], appear as natural multifractal generalizations of some of the main results in [1, 9, 10, 17, 18] and completely different from those found in [19].
We will now give a brief description of the organization of the paper. In the next section, we recall the definitions of the various multifractal dimensions and measures investigated in the paper. Section 2 recalls the multifractal formalism introduced in [13]. Section 3 contain our main results. The proofs are given in Sect. 4. The paper is concluded with Sect. 4 that, lists some interesting examples.
2 Preliminaries
We start by recalling the fine multifractal formalism introduced by Olsen in [13]. The key ideas behind the fine multifractal formalism in [13] are certain measures of Hausdorff-packing type which are tailored to see only the multifractal decomposition sets \(E(\alpha )\). These measures are natural multifractal generalizations of the centered Hausdorff measure and the packing measure and are motivated by the \(\tau _\mu \)-function which appears in the multifractal formalism. We first recall the definition of the multifractal Hausdorff measure and the the multifractal packing measure. Let \(\mu \) be a compactly supported probability measure on \(\mathbb {R}^n\). For \(q, t \in \mathbb {R}\), \(E \subseteq {\mathbb R}^n\) and \(\delta >0\), we define
where the supremum is taken over all centered \(\delta \)-packing of E. Moreover we can set \(\overline{\mathscr {P}}^{q,t}_{\mu ,\delta }(\emptyset )=0\). The packing pre-measure is then given by
In a similar way, we define
where the infinimum is taken over all centered \(\delta \)-covering of E. Moreover we can set \(\overline{\mathscr {H}}^{q,t}_{\mu ,\delta }(\emptyset )=0\). The Hausdorff pre-measure is defined by
Especially, we have the conventions \(0^q=\infty \) for \(q\le 0\) and \(0^q=0\) for \(q>0\).
\(\overline{{\mathscr {H}}}^{q,t}_{\mu }\) is \(\sigma \)-subadditive but not increasing and \(\overline{{\mathscr {P}}}^{q,t}_{\mu }\) is increasing but not \(\sigma \)-subadditive. That’s why Olsen introduced the following modifications on the multifractal Hausdorff and packing measures \({\mathscr {H}}^{q,t}_{\mu }\) and \({\mathscr {P}}^{q,t}_{\mu }\),
In follows that \(\mathscr {H}^{q,t}_{\mu }\) and \({\mathscr {P}}^{q,t}_{\mu }\) are metric outer measures and thus measures on the Borel family of subsets of \(\mathbb {R}^n\). An important feature of the Hausdorff and packing measures is that \({\mathscr {P}}^{q,t}_{\mu }\le {\overline{\mathscr {P}}}^{q,t}_{\mu }\). Moreover, there exists an integer \(\xi \in \mathbb {N}\), such that \({\mathscr {H}}^{q,t}_{\mu }\le \xi {\mathscr {P}}^{q,t}_{\mu }.\) The measure \(\mathscr {H}^{q,t}_{\mu }\) is a multifractal generalization of the centered Hausdorff measure, whereas \(\mathscr {P}^{q,t}_{\mu }\) is a multifractal generalization of the packing measure. In fact, it is easily seen that if \(t\ge 0\), then \(\mathscr {H}^{0,t}_{\mu }=\mathscr {H}^t\) and \(\mathscr {P}^{0,t}_{\mu }=\mathscr {P}^t\), where \(\mathscr {H}^t\) denotes the t-dimensional centered Hausdorff measure and \(\mathscr {P}^t\) denotes the t-dimensional packing measure.
The measures \(\mathscr {H}^{q,t}_{\mu }\) and \(\mathscr {P}^{q,t}_{\mu }\) and the pre-measure \({\overline{\mathscr {P}}}^{q,t}_{\mu }\) assign in the usual way a multifractal dimension to each subset E of \(\mathbb {R}^n\). They are respectively denoted by \(b_{\mu }^q(E)\), \(B_{\mu }^q(E)\) and \(\Lambda _{\mu }^q(E)\) and satisfy
and
The number \(b_{\mu }^q(E)\) is an obvious multifractal analogue of the Hausdorff dimension \(\dim _H(E)\) of E whereas \(B_{\mu }^q(E)\) and \(\Lambda _{\mu }^q(E)\) are obvious multifractal analogues of the packing dimension \(\dim _P(E)\) and the pre-packing dimension \(\Delta (E)\) of E respectively. In fact, it follows immediately from the definitions that
Next, for \(q\in \mathbb {R},\) we define the separator functions \(b_{\mu },\) \(B_{\mu }\) and \(\Lambda _{\mu }\) by
It is well known that the functions \(b_{\mu }\), \(B_{\mu }\) and \(\Lambda _{\mu }\) are decreasing and \(B_{\mu }\), \(\Lambda _{\mu }\) are convex and satisfying \( b_{\mu }\le B_{\mu }\le \Lambda _{\mu }.\)
The multifractal formalism based on the measures \({\mathscr {H}}^{q,s}_{\mu }\) and \({\mathscr {P}}^{q,s}_{\mu }\) and the dimension functions \(b_{\mu }\), \(B_{\mu }\) and \(\Lambda _{\mu }\) provides a natural, unifying and very general multifractal theory which includes all the hitherto introduced multifractal parameters, i.e., the multifractal spectra functions \(f_\mu \) and \(F_\mu \), the multifractal box dimensions. The dimension functions \(b_{\mu }\) and \(B_{\mu }\) are intimately related to the spectra functions \(f_\mu \) and \(F_\mu \), whereas the dimension function \(\Lambda _{\mu }\) is closely related to the upper box spectrum (more precisely, to the upper multifractal box dimension function \(\overline{\tau }_{\mu }\), see [13]).
The reader is referred to as Olsen’s classical text [13] for an excellent and systematic discussion of the multifractal Hausdorff and packing measures and dimensions.
3 Main results
In the next, we suppose the existence of a Gibbs’ measure at a state \((q, { B}_\mu (q))\) for the measure \(\mu \), i.e., the existence of a measure \(\nu _q\) on \(\text {supp}(\mu )\) and constants \(\underline{K}, \overline{K}> 0\) and \(\delta > 0\) such that for every \(x \in \text {supp}(\mu )\) and every \(0< r < \delta \),
to obtain the following result which provides a positive answer to Olsen’s questions [13, Question 7.6] and [14, Question 4.1.12] in a more general framework.
Theorem 1
Let \(q\in \mathbb {R}\) and we assume that there exists a Gibbs measure \(\nu _q\) for \(\mu \) at \((q, B_\mu (q))\), then there exists \(C>0\) such that
where \(\xi \) is the constant that appears in Besicovitch’s covering theorem. In addition, if \(\mu \) satisfies the doubling condition, then there exists \(C_1>0\) such that
Example A
Let \(\mu \) be the Bernoulli measure with parameters \(P_1\) and \(P_2\) which is defined by the repeated subdivision of a unit mass between the basic intervals of the pre-fractals of the middle-third Cantor set C. Then,
Example B
To define the Bedford–McMullen carpets, we introduce a digit set
where \(1<m\le n\). For each \((i, j)\in A\), we define \(T_{i,j}:\mathbb {R}^2\rightarrow \mathbb {R}^2\) by
We divide the unit square into nm congruent rectangles
It follows immediately from the definitions that
We let E be the unique non-empty compact set which satisfies
Sets of the form E are usually known as Bedford–McMullen carpets.
We introduce a positive probability vector \({\textbf {p}}\) with element \(p_{i, j}\) for each \((i, j)\in A\). We also define the related probability vector \({\textbf {q}}\) where
Thus, we can define a self-affine measure \(\mu \) which is the unique probability measure satisfying
Now, suppose that E satisfies these following disjointness conditions: for all distinct ordered pairs \((i, j)\in A\), \((i', j')\in A\), we have
Put \(\Omega \) as the set of all infinite sequences of ordered pairs belonging to A,
By the disjointness conditions, we known that there is a bijection \(\pi : \Omega \rightarrow E\) defined by
The value of \(\pi (\omega )\) is independent of the initial value \(v\in [0,1]\times [0,1]\). For \(x\in E\), let us write \(\omega _x=\pi ^{-1}(x)\) and let us denote \(\omega (n)=(\underline{x}_1,\underline{x}_2,\ldots ,\underline{x}_n)\) for all \(\underline{x}_i\in A,\;i=1,2,\ldots ,n\). Let \(\{a_{i,j};\;\; (i, j)\in A\}\) be a set of real numbers indexed by A, then write \(a_{\omega (n)} = a_{\underline{x}_1} a_{\underline{x}_2}\ldots a_{\underline{x}_n}\). Similarly \(T_{\omega (n)}\) means that the map \( T_{\underline{x}_1}\circ T_{\underline{x}_2}\circ \ldots \circ T_{\underline{x}_n}.\) Now, for \((x,y)\in \mathbb {R}^2\), let
For \(k\in \mathbb {N}\), we let \(l=l(k) = \lfloor \sigma k\rfloor \), where \(\lfloor .\rfloor \) denotes the integer part and \(\sigma =\frac{\log m}{\log n}\). Then the k-th level approximate square is defined as
It follows that
Therefore, we define \(\beta (q)\) (\(q\in \mathbb {R}\)) as the unique solution to
Let \(\gamma _i=\sum _{j,(i, j)\in A}p_{i, j}^q\). We then define the function \(\varphi _k^q(\omega )\) by
Now, we define the set \(\widetilde{F}\) as a subset of \(\Omega \) satisfying the following condition
Now, we write
It follows from the definition of \(\beta \) that
Denote by \(\mu _q\) the self-affine measure generated by \(P_{i, j}\) and \(T_{i, j}\), then
If we assume that \(\text {supp}(\mu )=\pi (\widetilde{F}),\) \(x\in \text {supp}(\mu )\) and \(r>0\), then for all \(\omega =\pi ^{-1}(x)\), we can choose \(h, k\in \mathbb {N}\) such that
It follows from (3.2) and the disjointness conditions that
Now, (3.1) implies that \(\mu _q\) is a Gibbs measure for \(\mu \) at \((q, \beta (q))\) and
Finally, by using Theorem 1, there exists \(C_1>0\) such that
Note that this example is already discussed much more comprehensively and complexly in [14, Section 6.7]. Indeed, in [14], Olsen considers Bedford–McMullen sponges in \(\mathbb {R}^d\) rather than Bedford–McMullen carpets in \(\mathbb {R}^2.\)
In the following, we give an example that had not already been investigated and studied for which the conditions of our main theorem are satisfied.
Example C
Let p be an integer with \(p\ge 2.\) Theorem 1 applies to a family of measures supported by the full p-adic grid of [0, 1], namely the quasi-Bernoulli measures.
We denote \({\mathscr {A}}\) the set of words constructed with \(\{0, 1,..., p - 1\}\) as an alphabet. Provided with concatenation, \({\mathscr {A}}\) is a monoid: if a and b are two elements of \({\mathscr {A}}\), we denote by ab the word obtained by concatenation of a and b. The empty word, which is the unit, is denoted by \(\varepsilon \). we denote the set of words of length n by \({\mathscr {A}}_n\). Now, we consider a sequence \(\big \{\{I_a\}_{a\in {\mathscr {A}}_n}\big \}_{n\ge 1}\) of nested finite partitions of the interval [0, 1[ in right half-open intervals: the intervals \(I_{al}\), \(l = 0, 1,..., p - 1\) constitute a partition of the interval \(I_a\). If \(x\in [0, 1[\), we denote by \(I_n(x)\) the element of the n-th generation \(\{I_a\}_{a\in {\mathscr {A}}}\) which contains it. The length of an interval I is denoted |I|. We assume (\(|\cdot |\) is almost multiplicative) that there is a positive constant L such that
Let us now consider the particular case where the sequence of partitions is given by the p-adic intervals \(\left\{ \left\{ I_{a}\right\} _{a \in \mathscr {A}_ {n}}\right\} _{n \ge 1}\):
A probability measure on [0, 1[ is said to be quasi-Bernoulli if there exists \(M > 0\) such that, for any a and \(b \in {\mathscr {A}},\)
Let \(\mu \) be a quasi-Bernoulli measure, for any \(q,t\in \mathbb {R}\), we define
where the star \(*\) means that the terms for which \(\mu (I_a) = 0\) are removed (a convention valid throughout this example), and let
It follows from Bhouri [4] and Peyrière [16] that, if \(\mu \) is a quasi-Bernoulli measure then there exist \(K>0\) and a measure \(\nu _q\) such that for all \(a \in {\mathscr {A}}\),
In the next, we will compare the function \(\tau _{\mu }(q)\) to \(\Lambda _{\mu }(q)\). For this we need the following extra condition:
Lemma 1
One has \(\tau _{\mu }(q)=\Lambda _{\mu }(q).\)
Proof
Let \(x \in {\text {supp}}(\mu ), 0<r<\frac{1}{p}\) and n such that \(p^{-n-1} \le r<p^{-n}\). To prove Lemma 1, we will prove that, there exist \(a \in {\mathscr {A}}\) and \(j \in \{0, \ldots , p-1\}\) such that
Proof of (3.5). Without loss of generality, we can assume that \(x\in [0, 1[.\) Two cases can then arise.
Case 1. \(\mu (I_n(x))\ne 0\).
Note that in this case, \(\mu (B(x, r)) \le \mu \left( I_{n}(x)\right) \). Indeed, the condition (3.4) implies that \(I_{n}(x)\) is the only interval of the n-th generation, of non-zero measure, meeting the ball B(x, r). Moreover, we have \(I_{n+1}(x) \subset B(x, r)\) since \(p^{-n-1} \le r\). The property (3.5) is then verified if \(\mu \left( I_{n+1}(x)\right) \ne 0\). It therefore remains to study the case where \(\mu \left( I_{n+1}(x)\right) =0\). Let \(I_{a}\) be the interval \(I_{n}(x)\), given (3.4) and the fact that \(x \in {\text {supp}}(\mu ), I_{n+1}(x)\) is different from \(I_{a 0}\) and therefore the interval \(I_{a j}\) which is just to the left of \(I_{n+1}(x)\) is contained in B(x, r) and has a non-zero measure. The property (3.5) is therefore satisfied in the case where \(\mu \left( I_{n+1}(x)\right) =0\).
Case 2. \(\mu (I_n(x))=0\).
Since \(x \in {\text {supp}}(\mu ),\) x is necessarily the left end of the interval \(I_{n}(x)\). Consider now the interval \(I_{a}\) of the n-th generation which is to the left of \(I_{n}(x)\) and which is contiguous to it. We then have \(\mu \left( I_{a}\right) \ne 0\), since \(x \in {\text {supp}}(\mu )\). The real r being strictly less than \(p^{-n}\), \(I_{a}\) is therefore the only interval of order n, of non-zero measure, meeting B(x, r). As consequence \(\mu (B(x, r)) \le \mu \left( I_{a}\right) \). Moreover, taking into account the inequality \(p^{-n-1} \le r\), we have \(I_{a j} \subset B(x, r)\) for \(j=p-1\). The assumption \(x \in {\text {supp}}(\mu )\) then implies that \(\mu \left( I_{a j}\right) \ne 0\), which establishes the property (3.5).
The proof of Lemma 1 follows immediately from the property (3.5) (for more details see [20]). \(\square \)
Let us now show that (3.3) is valid for intervals centered in the support of \(\mu \), in other words \(\nu _{q}\) is a Gibbs measure for \(\mu \) at \((q, \Lambda _{\mu }(q))\).
We consider here only the case \(q \ge 0\), the other case is treated in the same way. Let \(x \in {\text {supp}}(\mu )={\text {supp}}(\nu _{q}), 0<r<p^{-1}\) and n be the integer such that \( p^{-n-1} \le r<p^{-n}.\) It follows immediately from (3.3) that
By using (3.5) and (3.6), there is a word \(a \in \mathscr {A}_{n}\) and a letter \(j \in \{0,1, \ldots , p -1\}\) such that
and
It follows from (3.3) and (3.8) that
Since \(\mu \) is a quasi-Bernoulli measure and \(\mu \left( I_{a j}\right) \ne 0\), it results that
where \(\rho =\inf \big \{\mu \left( I_{b}\right) ; b \in \{0,1, \ldots , p-1\}\;\text {and}\;\mu \left( I_{b}\right) \ne 0\big \}.\) Since \(r<\left| I_{a}\right| \le p r\), we, therefore, have by using (3.7) and (3.9),
Now, Lemma 1 implies that, there exists a constant \(K_1>0\) such that
where \(K_1=\left( \frac{M}{\rho }\right) ^{q}\left( {2p}\right) ^{\Lambda _{\mu }(q)}\). Which implies that
and \(\nu _{q}\) is a Gibbs measure for \(\mu \) at \((q, B_{\mu }(q))\). Finally, it follows from Theorem 1 that, there exists \(C_1>0\) such that
where \(\xi \) is the constant that appears in Besicovitch’s covering theorem.
3.1 Moran sets
Let us recall the class of Moran sets. We denote by \(\{n_k\}_{k\ge 1}\) a sequence of positive integers with \(n_k\ge 2\) and \(\Phi =\{\Phi _k\}_{k\ge 1}\) be a sequence of vectors satisfying
Define \(D=\bigcup _{k\ge 1}D_{k}.\)
Definition 1
We say that the collection \(\mathscr {F}=\left\{ J_{\sigma }, \sigma \in D\right\} \) fulfills the Moran structure if it satisfies the following conditions:
-
(1)
For all \(\sigma \in D\), \(J_{\sigma }\) is similar to J, that is there exists a similarity mapping \(S_{\sigma }:\mathbb {R}^d\rightarrow \mathbb {R}^d\) such that \(S_{\sigma }(J)=J_{\sigma }\). Here we set \(J_{\emptyset }=J\).
-
(2)
For all \(k\ge 0\) and \(\sigma \in D_{k}, J_{\sigma * 1}, J_{\sigma * 2}, \ldots , J_{\sigma {*} n_{k+1}}\) are subsets of \(J_\sigma \), and satisfy that \(J_{\sigma * i}^{\circ } \cap J_{\sigma *j}^{\circ }=\emptyset \;(i \ne j)\) [we call such assumption open set condition (OSC)], where \(A^{\circ }\) denotes the interior of A.
-
(3)
For any \(k\ge 1,\sigma \in D_{k-1}\) and \(1\le j\le n_k\), \( c_{k,j}=\frac{\left| J_{\sigma {*} j}\right| }{\left| J_{\sigma }\right| }, \; 1 \le j \le n_{k}\), where |A| denotes the diameter of A.
Let \(\mathscr {F}=\mathscr {F}\left( J,\left\{ n_{k}\right\} ,\left\{ \Phi _k\right\} \right) \) be a collection having Moran structure. The set \(E(\mathscr {F})=\bigcap _{k \ge 1}\bigcup _{\sigma \in D_{k}} J_{\sigma }\) is called a Moran set determined by \(\mathscr {F}\). It is convenient to denote \(M\left( J,\left\{ n_{k}\right\} ,\left\{ \Phi _k\right\} \right) \) the collection of Moran sets determined by J, \(\left\{ n_{k}\right\} \) and \(\left\{ \Phi _{k}\right\} \).
Remark 1
If \(\lim _{k \rightarrow +\infty } \sup _{\sigma \in D_{k}}\left| J_{\sigma }\right| >0\), then E contains interior points. Thus the measure and dimension properties will be trivial. We assume therefore \(\lim _{k\rightarrow +\infty } \sup _{\sigma \in D_{k}}\left| J_{\sigma }\right| =0.\)
Now, we consider a class of Moran sets E which satisfy a special property called the strong separation condition (SSC, which is stronger than OSC), i.e., take \(J_\sigma \in {\mathscr {F}}\). Let \(J_{\sigma *1}, J_{\sigma *2}, \ldots , J_{\sigma *n_{k+1}}\) be the \(n_{k+1}\) basic sets of order \(k + 1\) contained in \(J_\sigma \), then we assume that for all \(1\le i\ne j\le n_{k+1}-1\), \( \text {dist} (J_{\sigma *i}, J_{\sigma *j}) \ge \Delta _k |J_\sigma |,\) where \((\Delta _k)_{k\in \mathbb {N}}\) is a sequence of positive real numbers, such that
Then the assumption \(\lim _{k\rightarrow +\infty } \sup _{\sigma \in D_{k}}\left| J_{\sigma }\right| =0\) follows.
If we ask \(c_{k,j}=c_k\) for all \(1\le j\le n_k\), where \(\{c_k\}_{k\ge 1}\) is a sequence of positive numbers, we can get the Moran structure and Moran sets. In this situation, we call them by homogeneous Moran structure and the collection of Moran sets, and denote by \(\mathscr {F}=\mathscr {F}\left( J,\left\{ n_{k}\right\} ,\left\{ c_k\right\} \right) \) and \(\mathscr {M}=\mathscr {M}\left( J,\left\{ n_{k}\right\} ,\left\{ c_k\right\} \right) .\)
3.2 Moran measure
Let \(\left\{ P_{i, j}\right\} _{j=1}^{n_{i}}\) be probability vectors, i.e., \(P_{i, j}>0\) and \(\sum _{j=1}^{n_{i}} P_{i, j}=1(i=1,2, \ldots ),\) suppose that \(P_{0}=\inf \left\{ P_{i, j}\right\} >0.\) Let \(\mu \) be a mass distribution on E such that for any \(J_{\sigma }\left( \sigma \in D_{k}\right) \) \(\mu \left( J_{\sigma }\right) =P_{1,\sigma _{1}}P_{2,\sigma _{2}} \cdots P_{k,\sigma _{k}}\) and \(\mu \left( \sum _{\sigma \in D_{k}} J_{\sigma }\right) =1,\) we call \(\mu \) a Moran measure on E.
For \(q\in \mathbb {R}\), we define the following functions
In the following theorem, we find an explicit formula for the multifractal Hausdorff and packing function dimensions of a homogeneous Moran set satisfying the strong separation condition.
Theorem 2
Let E be a homogeneous Moran set satisfying (SSC) and \(\mu \) be the Moran measure on E. Then for all \(q\in \mathbb {R}\), we have
Remark 2
The results developed by Beak in [1] and Feng et al. in [10] are obtained as a special case of the multifractal theorems when q equals 0.
Remark 3
Let E be a homogeneous Moran set satisfying (SSC) and \(\mu \) be the Moran measure on E. If the limit \(\lim _{k\rightarrow +\infty }\beta _k(q) = \beta (q)\) exists, and for all \(k\ge 1\), \(k(\beta (q)-\beta _k(q))<+\infty \), then by using [24, Proposition 3.1] there exists a probability measure \({\nu }_q\) supported by E such that for any \(k\ge 1\) and \(\sigma _0\in D_k\),
It follows from \(k(\beta (q)-\beta _k(q))<+\infty \) for all \(k\ge 1\) that
Now, by using the strong separation condition, we have
and \(\nu _q\) is a Gibbs measure for \(\mu \) at \((q, B_\mu (q))\) (it is the case of [6, 7]) which implies that the conditions in Theorem 1 are satisfied.
4 Proof of main results
4.1 Proof of Theorem 1
This theorem follows immediately from the following lemma.
Lemma 2
For any \(q\in \mathbb {R}\), there exist two constants \({K}_1>0\) and \({K}_2>0\) such that
Proof
Fix \(\delta >0\) and let \(\Big (B\left( x_{i}, r_{i}\right) \Big )_{i\in \mathbb {N}}\) be a centered \(\delta \)-covering of \(\text {supp}(\mu )\). Then
Consequently
Let F be a closed subset of \(\text {supp}(\mu )\). For \(\delta >0\) write
Since F is closed, \(B(F,\delta )\searrow F\) for \(\delta \searrow 0\). Then for all \(\varepsilon >0\), there exists \(\delta _0\) satisfying
Now, fix \(\delta >0\) and let \(\Big (B(x_i,r_i)\Big )_{i\in \mathbb {N}}\) be a centered \(\delta \)-packing of F. Observing that
It results that
Letting \(\varepsilon \downarrow 0\), now yields
which proves the desired result with \(K_2=\frac{1}{\;\underline{K}\;}\) and \(K_1=\frac{1}{\;\overline{K}\;}\). \(\square \)
4.2 Proof of Theorem 2
We present the tools, as well as the intermediate results, which will be used in the proof of our main result. First, we express the multifractal Hausdorff and packing measures of a homogeneous Moran set as the explicit form with \(n_k,\) \(c_k\) and \(P_{i,j}\).
Proposition 1
Let E be a homogeneous Moran set satisfying (SSC) and \(\mu \) be the Moran measure on E. Then for all \(q,t\in \mathbb {R}\),
-
(1)
there exists \(A>0\) such that
$$\begin{aligned} A~\displaystyle \liminf _{ k\rightarrow \infty } \prod _{m=1}^k \sum _{j=1}^{n_m}P_{m,j}^q c_m^t \le \mathscr {H}_{\mu }^{q,t}(E)\le \displaystyle \liminf _{ k\rightarrow \infty } \prod _{m=1}^k \sum _{j=1}^{n_m}P_{m,j}^q c_m^t. \end{aligned}$$ -
(2)
If \(\limsup _{k\rightarrow \infty } \prod _{m=1}^k \sum _{j=1}^{n_m}P_{m,j}^q c_m^t=0\) or \(+\infty \) then \(\mathscr {P}_{\mu }^{q, t}(E)=0\) or \(+\infty \) which implies that \(\overline{\mathscr {P}}_{\mu }^{q, t}(E)=0\) or \(+\infty \).
-
(3)
There exist C and \(c_{p}>1\) such that
$$\begin{aligned}{} & {} \max \Big (1,C^{2q}\Big )c_{p}^{-1} \limsup _{k \rightarrow \infty } \prod _{m=1}^k \sum _{j=1}^{n_m}P_{m,j}^q c_m^t\le \mathscr {P}_{\mu }^{q, t}(E)\\{} & {} \quad \le \overline{\mathscr {P}}_{\mu }^{q, t}(E)\le c_{p}\limsup _{k\rightarrow \infty }\prod _{m=1}^k \sum _{j=1}^{n_m}P_{m,j}^q c_m^t. \end{aligned}$$
Proof
The verification of this proposition now follows routinely from the theory described by Wu et al. [24, Propositions 3.3 and 3.4] and [22, Theorem 1]. \(\square \)
Remark 4
For any \(k\ge 1\) and \(\sigma \in D_{k-1}\), \(J_{\sigma }, J_{\sigma {*} 1},...,J_{\sigma {*} n_k}\) are arranged from the left to the right, \(J_{\sigma {*} 1}\) and \(J_{\sigma }\) have the same left endpoint, \(J_{\sigma {*} n_k}\) and \(J_{\sigma }\) have the same right endpoint, and the lengths of the gaps between any two consecutive sub-intervals are equal. We denote the length of one of the gaps by \(y_k\). Motivated by some results developed in [17, 18], we conjecture that if \(y_{k+1} \le y_k\) (or \(y_{k+2} \le y_k\) and \(y_{k+3} \le y_k\)) for all \(k\ge 1\) then
Now, we define the following auxiliary dimensions
and
It follows from Proposition 1 that, for all \(q\in \mathbb {R}\)
Theorem 2 is a consequence of the following proposition.
Proposition 2
Let E be a homogeneous Moran set satisfying (SSC) and \(\mu \) be the Moran measure on E. Then for all \(q\in \mathbb {R}\), we have
Proof
Let \(t>\underline{\beta }(q)\), then there exists a subsequence \((k_i)_i\) such that
Which implies that
which clearly implies that \(\underline{\varphi }(q)\le \underline{\beta }(q).\) The proof of the other inequality is identical to the proof of the first statement and is therefore omitted.
We will prove the second assertion. Let \(t>\overline{\beta }(q)\), there exists \(N\in \mathbb {N}\) such that for all \(k\ge N\) we have
Which clearly implies that
It follows that \(\overline{\varphi }(q)\le \overline{\beta }(q).\) The proof of the other inequality is identical to the proof of the first statement and is therefore omitted which yields the desired result. \(\square \)
5 Some examples
In this section, more motivations and examples related to these concepts, will be discussed. In particular, some examples show that the two main results are completely related.
5.1 Example 1
If \(J=[0,1]\), \(n_k=2\) and \(c_k=\frac{1}{3}\) for all \(k\ge 1\) then the set E is the middle-third Cantor set and \(\mu \) is the Bernoulli measure with parameters \(P_1=P_{1,1}\) and \(P_2=P_{1,2}\). Also, Theorem 2 implies that
5.2 Example 2
Let \(A=\{a, b\}\) be a two-letter alphabet, and \(A^{*}\) the free monoid generated by A. Let F be the homomorphism on \(A^{*},\) defined by \(F(a)=ab\) and \(F(b)=a\). It is easy to see that \(F^{n}(a)=F^{n-1}(a) F^{n-2}(a)\). We denote by \(\left| F^{n}(a)\right| \) the length of the word \(F^{n}(a)\), thus
Therefore, as \(n \rightarrow \infty \), we get the infinite sequence
which is called the Fibonacci sequence. For any \(n\ge 1,\) write \(\omega _{n}=\left. \omega \right| _{n}=s_{1} s_{2} \cdots s_{n}\). We denote by \(\left| \omega _{n}\right| _{a}\) the number of the occurrence of the letter \(a \text { in } \omega _{n}, \text { and }\left| \omega _{n}\right| _{b}\) the number of occurrence of b. Then \(\left| \omega _{n}\right| _{a}+\left| \omega _{n}\right| _{b}=n\). It follows from Wu [23] that \(\lim _{n \rightarrow + \infty } \frac{\left| \omega _{n}\right| _{a}}{n}=~\eta ,\) where \(\eta ^{2}+\eta =1\).
Let \(0<r_{a}<\frac{1}{2}, 0<r_{b}<\frac{1}{3}, r_{a}, r_{b} \in \mathbb {R}\). In the above Moran construction, let
and
Then we construct the homogeneous Moran set relating to the Fibonacci sequence and denote it by \(E:=E(\omega )=\left( J,\left\{ n_{k}\right\} ,\left\{ c_{k}\right\} \right) \). By the construction of E, we have
Let \(P_{a}=\left( P_{a_{1}}, P_{a_{2}}\right) , P_{b}=\left( P_{b_{1}}, P_{b_{2}}, P_{b_{3}}\right) \) be probability vectors, i.e.,
For any \(k \ge 1\) and any \(\sigma \in D_{k},\) we know \(\sigma =\sigma _{1} \sigma _{2} \cdots \sigma _{k}\) where
For \(\sigma =\sigma _{1}\sigma _{2} \cdots \sigma _{k},\) we define \(\sigma (a)\) as follows: let \(\omega _{k}=s_{1}s_{2}\cdots {s_{k}}\) and \({e_{1}}<e_{2}<\cdots <e_{\left| \omega _{k}\right| _{a}}\) be the occurrences of the letter a in \(\omega _{k},\) then \(\sigma (a)=\sigma _{e_{1}}\sigma _{e_{2}}\cdots \sigma _{e_{\left| \omega _{k}\right| _{a}}}\). Similarly, let \(\delta _{1}<\delta _{2}<\cdots <\delta _{\left| \omega _{k}\right| _{b}}\) be the occurrences of the letter b in \(\omega _{k},\) then \(\sigma (b)=\sigma _{\delta _{1}}\sigma _{\delta _{2}}\cdots \sigma _{\delta _{|\omega _{k}|_{b}}}\).
Let
Obviously
Let \(\mu \) be a mass distribution on E, such that for any \(\sigma \in D_{k},\)
It follows that
By using Theorem 2 we have
where \(\eta ^{2}+\eta =1\).
Given \(q\in \mathbb {R},\) it follows from Wu [23, Proposition 3.1] that there exists a probability measure \(\nu _q\) supported by E such that for any \(k\ge 1\) and \(\sigma _0\in D_k\),
where
By a simple calculation, we have
which implies that \( \beta (q)-\beta _k(q)=O(\frac{1}{k})\) and
which gives that
By a similar way, we obtain
This implies that
Now, (5.1) gives that \(\nu _q\) is a Gibbs measure for \(\mu \) at \((q, B_\mu (q))\) and then the conditions of Theorem 1 are satisfied.
5.3 Example 3
A particular homogeneous Moran set E satisfying (SSC) and a Moran measure \(\mu \) on E may now be defined as follows: Let
and
where \(0<r_1<\frac{1}{2}\) and \(0<r_2<\frac{1}{3}\). Put
where
We conclude that
It follows from Theorem 2 that
It is obvious that \(k(\beta (q)-\beta _k(q))<+\infty \) for all \(k\ge 1\) which implies, by using similar techniques of Sect. 5.2, that
Wu and Xiao [24, Proposition 3.1] implies that, there exists a probability measure \({\nu }_q\) supported by E such that for any \(k\ge 1\) and \(\sigma _0\in D_k\),
Now, it follows from Theorem 2 that \(\nu _q\) is a Gibbs measure for \(\mu \) at \((q, B_\mu (q))\) which gives that the conditions in Theorem 1 are satisfied.
Next, we give concrete interesting examples related to our main result and we obtain the multifractal Hausdorff and packing dimension functions of Moran measure associated with homogeneous Moran fractals for which \(b_\mu (q)\) and \(B_\mu (q)\) differ for all \(q\ne 1\).
5.4 Example 4
Let \((t_k)_k\) be a sequence of integers such that
In the Moran construction described in Definition 1, we define the family of parameters \(n_i\), \(c_i\) and \(p_{i,j}\) as follows:
For \(0<r_a<\frac{1}{2}\) and \(0<r_b<\frac{1}{3}\), let
Let \((P_{a,j})_{j=1}^2\) and \((P_{b,j})_{j=1}^3\) be two probability vectors. Define
and
If \(N_k\) is the number of integers \(i\le k\) such that \(P_{i,j} = P_{a,j}\), then
and
We can then conclude from Theorem 2 that
and
5.5 Example 5
Let \(A=\{a, b\}\) be a two-letter alphabet, \(\omega =s_1s_2...s_k....\) be a sequence over A, \(s_i\in A\). For any \(n\ge 1,\) write \(\omega _{n}=\left. \omega \right| _{n}=s_{1} s_{2} \cdots s_{n}\). We denote by \(\left| \omega _{n}\right| _{a}\) the number of the occurrence of the letter \(a \text { in } \omega _{n}, \text { and }\left| \omega _{n}\right| _{b}\) the number of occurrence of b. Then \(\left| \omega _{n}\right| _{a}+\left| \omega _{n}\right| _{b}=n\). In the above Moran construction, we take
where \(0<r_a<\frac{1}{2}\), \(0<r_b<\frac{1}{3}\). Let \(P_{a}=\left( P_{a_{1}}, P_{a_{2}}\right) , P_{b}=\left( P_{b_{1}}, P_{b_{2}}, P_{b_{3}}\right) \) be probability vectors such that
By a simple calculation, we get
where \(\tau _k(a)=\frac{\left| \omega _{k}\right| _{a}}{k}\). Write \(\underline{\tau }(a)=\liminf _{k\rightarrow \infty }\tau _k(a)\) and \(\overline{\tau }(a)=\limsup _{k\rightarrow \infty }\tau _k(a)\). Using (5.2), we have
-
(1)
if \(q<1\), then
$$\begin{aligned}{} & {} \log (r_a)\log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log (r_b)\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \ge 0,\\{} & {} {b}_{\mu }(q)=\frac{\underline{\tau }(a)\left( \log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \right) -\log \left( \sum _{1}^{3} P_{b_{j}}^{q}\right) }{\underline{\tau }(a)\left( \log r_{a}-\log r_b\right) + \log r_{b}} \end{aligned}$$and
$$\begin{aligned} {B}_{\mu }(q)=\frac{\overline{\tau }(a)\left( \log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \right) -\log \left( \sum _{1}^{3} P_{b_{j}}^{q}\right) }{\overline{\tau }(a)\left( \log r_{a}-\log r_b\right) + \log r_{b}}.\end{aligned}$$ -
(2)
If \(q\ge 1\), then
$$\begin{aligned}{} & {} \log (r_a)\log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log (r_b)\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \le 0,\\{} & {} {b}_{\mu }(q)=\frac{\overline{\tau }(a)\left( \log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \right) -\log \left( \sum _{1}^{3} P_{b_{j}}^{q}\right) }{\overline{\tau }(a)\left( \log r_{a}-\log r_b\right) + \log r_{b}} \end{aligned}$$and
$$\begin{aligned}{B}_{\mu }(q)=\frac{\underline{\tau }(a)\left( \log \left( \sum _{i}^{3} P_{b_{j}}^{q}\right) -\log \left( \sum _{i}^{2} P_{a_{i}}^{q}\right) \right) -\log \left( \sum _{1}^{3} P_{b_{j}}^{q}\right) }{\underline{\tau }(a)\left( \log r_{a}-\log r_b\right) + \log r_{b}}.\end{aligned}$$
5.6 Example 6
Let \(J=[0,1]\), \(n_i=2\) and \(\mathscr {N}:=\{N_k\}_{k\in \mathbb {N}}\) be an increasing sequence of integers with \(N_0=0\) and \({\lim _{k\rightarrow +\infty }\frac{N_{k+1}}{N_{k}}=+\infty }\). Fix four real numbers \(A,B,p,\widetilde{p}\) with \(A>B>2\) and \(0<p,\widetilde{p}\le 1/2\). Now for every \(i\in \mathbb {N}\), we define \(c_i\) and \(\{P_{i,j}\}_{1\le j\le n_i}\) as follows:
Now, we can define a homogeneous Moran set E satisfying (SSC) and a Moran measure \(\mu \) on it. Define the functions
and
We can conclude that
If \(-\frac{\log (1-\widetilde{p})}{\log B}<-\frac{\log p}{\log A}\), the method in [2, 26] can follows that:
and
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
Baek, H.K.: Packing dimension and measure of homogeneous Cantor sets. Bull. Aust. Math. Soc. 74, 443–448 (2006)
Ben Nasr, F., Peyrière, J.: Revisiting the multifractal analysis of measures. Rev. Math. Ibro. 25, 315–328 (2013)
Ben Nasr, F., Bhouri, I., Heurteaux, Y.: The validity of the multifractal formalism: results and examples. Adv. Math. 165, 264–284 (2002)
Bhouri, I.: Une condition de validité du formalisme multifractal pour les mesures. Thèse de Doctorat (1999)
Douzi, Z., Selmi, B.: On the mutual singularity of Hewitt–Stromberg measures for which the multifractal functions do not necessarily coincide. Ric. Mat. (to appear). https://doi.org/10.1007/s11587-021-00572-6
Douzi, Z., Selmi, B.: On the mutual singularity of multifractal measures. Electron. Res. Arch. 28, 423–432 (2020)
Douzi, Z., Samti, A., Selmi, B.: Another example of the mutual singularity of multifractal measures. Proyecciones 40, 17–33 (2021)
Falconer, K.J.: Techniques in Fractal Geometry. Wiley, New York (1997)
Feng, D.-J., Wen, Z.-Y., Wu, J.: Dimension of the homogeneous Moran sets. Sci. China (Ser. A) 40, 475–482 (1997)
Feng, D.-J., Rao, H., Wu, J.: The net measure properties of one dimensional homogeneous Cantor set and its applications. Progr. Nat. Sci. 7, 673–678 (1997)
Halsey, T.C., Jensen, M.H., Kadanof, L.P., Procaccia, I., Shraiman, B.J.: Fractal measures and their singularities: the characterization of strange sets. Phys. Rev. A 33, 1141–1151 (1986)
Huang, L., Liu, Q., Wang, G.: Multifractal analysis of Bernoulli measures on a class of homogeneous Cantor sets. J. Math. Anal. Appl. 491, 124362 (2020)
Olsen, L.: A multifractal formalism. Adv. Math. 116, 82–196 (1995)
Olsen, L.: Self-affine multifractal Sierpinski sponges in \(\mathbb{R} ^d\). Pac. J. Math. 183, 143–199 (1998)
Pesin, Y.: Dimension theory in dynamical systems, contemporary views and applications. Chicago Lectures in Mathematics, University of Chicago Press, Chicago (1997)
Peyrière, J.: Multifractal measures. Proc. NATO Adv. Study Inst. Prob. Stoch. Methods Anal. Appl., Il Ciocco, NATO ASI Ser., Ser. C: Math. Phys. Sci. 372, 175–186 (1992)
Qu, C.Q., Rao, H., Su, W.Y.: Hausdorff measure of homogeneous Cantor set. Acta Math. Sin., English Seri. 17, 15–20 (2001)
Qu, C.Q.: Hausdorff measures for a class of homogeneous Cantor Sets. Acta Math. Sin., English Ser. 29, 117–122 (2013)
Selmi, B.: Some new characterizations of Olsen’s multifractal functions. RM 75(147), 1–16 (2020)
Selmi, B.: Remarks on the mutual singularity of multifractal measures. Proyecciones 40, 71–82 (2021)
Shen, S.: Multifractal analysis of some inhomogeneous multinomial measures with distinct analytic Olsen’s \(b\) and \(B\) functions. J. Stat. Phys. 159, 1216–1235 (2015)
Shengyou, W., Wu, M.: Relations between packing premeasure and measure on metric space. Acta Math. Sci. 27, 137–144 (2007)
Wu, M.: The singularity spectrum \(f(\alpha )\) of some Moran fractals. Monatsh. Math. 144, 141–55 (2005)
Wu, M., Xiao, J.: The singularity spectrum of some non-regularity Moran fractals. Chaos, Solitons Fractals 44, 548–557 (2011)
Xiao, J., Wu, M.: The multifractal dimension functions of homogeneous Moran measure. Fractals 16, 175–185 (2008)
Yuan, Z.: Multifractal spectra of Moran measures without local dimension. Nonlinearity 32, 5060–5086 (2019)
Acknowledgements
The referees’ constructive criticism and recommendations on the text are appreciated by the authors. The authors would like to thank Professor De-Jun Feng for the reference [10] and Professors Lars Olsen and Jinjun Li for useful discussions while writing this manuscript. The first author is supported by Analysis, Probability & Fractals Laboratory (No: LR18ES17). The second author is supported by the National Natural Science Foundation of China (Grant No. 12061006) and Natural Science Foundation of Jiangxi (Grant No. 20212BAB201002).
Funding
There are no funding sources for this manuscript.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no conflict of interest to declare.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Selmi, B., Yuan, Z. On the multifractal measures: proportionality and dimensions of Moran sets. Rend. Circ. Mat. Palermo, II. Ser 72, 3949–3969 (2023). https://doi.org/10.1007/s12215-023-00873-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12215-023-00873-9
Keywords
- Multifractal analysis
- Homogeneous Cantor sets
- Hausdorff dimension
- Packing dimension
- Homogeneous Moran measures