Abstract
Generalizations and refinements are given for results of Kozlov and Treschev on non-uniform averagings in the ergodic theorem in the case of operator semigroups on spaces of integrable functions and semigroups of measure-preserving transformations. Conditions on the averaging measures are studied under which the averages converge for broad classes of integrable functions.
Bibliography: 96 items.
Export citation and abstract BibTeX RIS
This research was supported by the Russian Foundation for Basic Research (grants nos. 18-31-20008 and 20-01-00432), Moscow Center of Fundamental and Applied Mathematics, and the Foundation for the Advancement of Theoretical Physics and Mathematics BASIS (grant no. 18-1-6-83-1). |
§ 1. Introduction
This paper is devoted to generalizations and refinements of several results of Kozlov and Treschev in [58]–[60], [56] connected with the study of convergence of non-uniform averagings in the situation of the individual ergodic theorem and its extensions. In its original setting their problem was concerned with classical dynamical systems on manifolds (see [55], [61], and [57]), but both the formulations of the results and the methods of proofs could be carried over directly to a considerably more general context of measure spaces, measurable transformations of them, and functions on them. Moreover, in this more general situation the formulations and proofs turned out to be even slightly simpler. Therefore, our discussion below follows this generality. The cited papers have attracted the attention of other researchers and have been further developed (see [15], [20], [21], [53], [54], and also the recent paper [67] in physics). Below we give a survey of this research along with some new results.
The aforementioned non-uniform averaging is defined by the formula
for integrable functions on a probability space on which a semigroup of measurable transformations , , acts with preservation of the measure , that is, , , and the map is measurable, where is a Borel probability measure on . A more general problem deals with the integrals
where is a semigroup of bounded linear operators on the space . The question is posed of the existence of the limit of the function almost everywhere or in the mean as . In addition, in this situation there arises an interesting dynamics of measures on the space (see §5).
If is Lebesgue measure on the interval , then after the change of variable we arrive at the classical averages
or, in the operator case,
For the former integral the Birkhoff–Khinchin theorem (see [15], [64], [75], [31], and [34]) gives convergence almost everywhere for any integrable function , to the function equal to the conditional expectation of with respect to the -algebra generated by all measurable functions invariant with respect to all the transformations . This conditional expectation is also the projection of on the subspace of functions invariant with respect to all the transformations . For the latter integral a number of convergence results are also known, which we present in §3. The Kozlov–Treschev theorem gives convergence of to the same limit for bounded measurable functions and absolutely continuous measures .
There is a vast literature on individual ergodic theorems for operator semigroups (not necessarily generated by transformations of measure spaces, as in the first ergodic theorems) on spaces of integrable functions, beginning with the classical works [33] and [34] by Dunford and Schwartz, the main results in which will be recalled below. We note the survey [3], the papers [12], [5], [40], [90], [91], [51], [65], [44], [94], [95] investigating convergence of the ratios , which reduce to for and but are more natural in other situations, the papers [35], [77], [80], [38], [39], [36], and [37] on convergence of resolvent averages of the form
which corresponds to the weight and (such averages were considered as early as [33] and [34]), the paper [1] on averaging with monotone weights, and also the papers [78], [79], [81], [82], [71], [13], and [96] on classical averaging or averaging with other special weights. However, it should be noted that the author of [12] already considered non-uniform averages of the form
where is an increasing function of bounded variation satisfying the equation with some probability measure . In this setting the existence of a limit was proved for the ratio , where and is a probability density with respect to . If , then this case is similar to the classical situation with division of an integral by ; the only difference is that the integral over is taken not with respect to Lebesgue measure, but with respect to . In addition, [28] and [29] dealt with non-uniform averages exactly of the form (1.1) with densities having bounded supports, which has some specific features (results in these papers are mentioned below). In [6], sequences of non-uniform averages defined by sequences of different weights were studied. Many works have been devoted to non-uniform discrete averages for single transformations or operators (see [64], [11], [69], and the literature therein). Certain non-uniform averages of a different kind connected with ergodic group actions were considered in [87]–[89], [7], [70], and [42], where there are also additional references. There is an even more extensive literature on convergence of averages in the mean, but here we hardly touch upon this question.
In §3 we discuss convergence of non-uniform averages for unbounded functions mostly in the second, more general, case of semigroups. In §4, non-uniform averaging of stochastic systems is considered. In §5, weak convergence of measures generated by the averaging measure is discussed.
§ 2. Notation, terminology and auxiliary results
Throughout, we use the basic standard notions and facts connected with the Lebesgue integral, the spaces with their classical norms , and linear operators. In some constructions and results we use certain more special concepts introduced below.
In particular, we shall use the Orlicz classes of integrable functions generated by convex functions (see [62] and [76]). An increasing convex function on is called an -function if
In other words,
where is a non-decreasing function on such that
An -function is said to satisfy the -condition for large values if, for some and , we have
For every this implies the estimate
with some . If this condition holds for , then is said to satisfy the global -condition.
For such a function we obtain
To satisfy the -condition for large values it suffices to have the estimate
for the right-hand derivative of the function outside some interval (see [62], Chap. 1, §4).
Typical examples of -functions satisfying the global -condition are with (but not with ) and also .
Below we use the known fact (see [62], §8 and Lemma 5.1 in §5) that, for every integrable function , there exists an -function satisfying the global -condition for which the function is integrable.
For an -function on the complementary (or complex conjugate) function is defined by
Hence Young's inequality holds. The complementary function is also convex and increasing (and is an -function). For example, if , where , then , where .
The Orlicz class generated by a given -function consists of the equivalence classes of -measurable functions for which . One has . It is known that the Orlicz class is a convex set, but for an atomless measure it becomes a linear space precisely when the function satisfies the -condition for large values. Even if does not satisfy this condition, one can introduce the Banach space of equivalence classes of measurable functions with finite Orlicz norm
where is the complementary function for . Then , and if the -condition holds for large values, then the two sets are equal (in the general case coincides with the linear span of the class ). In addition, in the general case
for and . The Orlicz norm is equivalent to the Luxemburg norm defined without the complementary function by the equality
One has . If the -condition holds for large values, then convergence of a sequence to zero in one of these norms is equivalent to convergence to zero of the integrals of (see [62], Theorem 9.4).
A linear operator on the space is said to be non-negative if for all .
The norm of an operator on will be denoted by , and the norm of an operator between Banach spaces and is denoted by .
If is a probability space and is a sub--algebra of , then the conditional expectation of a -integrable function is the -integrable function which is measurable with respect to and such that
for all bounded -measurable functions . One can also say that this is the projection of with some special properties on the subspace of -measurable functions (for functions in this is the usual orthogonal projection).
Below we need the following Jensen inequality for operators, which can be extracted from the results in [43] but which for the reader's convenience we derive here from the usual Jensen inequality (we even give two derivations).
Lemma 2.1. Let be a probability space and let be a non-negative operator with . Suppose that is an increasing convex function on with . Then
If extends to a continuous operator from to and is such that , then this estimate remains valid.
Proof. Since by the non-negativity of , we can assume that . Moreover, it suffices to prove the assertion for bounded . Fix some -measurable bounded versions of the functions and .
For every , we can find measurable sets such that for -almost all we get that for all and
For example, one can take the sets
Then for a fixed we cover by the measurable disjoint sets
where a finite collection of disjoint Borel sets of diameter less than covers a square containing the values of the map , and we take only those sets which have positive -measure. For every point in such a set we have , hence . Moreover,
which gives the indicated relations.
Now fix a point such that for all and let and be the indicated limits. Consider the non-negative measure
where is the bounded operator on adjoint to the operator on . Thus, the measure is given by the density with respect to the measure . The integral of a bounded measurable function with respect to the measure is given by
It is clear that , since .
By the condition and the convexity of , for the subprobability measure we have Jensen's inequality
By our choice of and the continuity of we obtain on the left-hand side and on the right-hand side as .
Let us consider another justification for the case of . We recall (for example, see [27], §11.6) that the Gelfand transform defines a linear isometry of the complex Banach algebra and the space of continuous complex functions on the compact space of maximal ideals of this Banach algebra; moreover, is a homomorphism of algebras and preserves the non-negativity of elements. Hence, the non-negative operator on generates a non-negative operator on by the formula . By the Riesz theorem, for every there exists a Radon measure on for which
This measure is non-negative because for . In addition, since , as follows from the estimate and the equality . By the usual Jensen's inequality for subprobability measures we get that for . Now it remains to observe that for all non-negative , since for every continuous function on the real line we have , because this is true for all polynomials and remains true for their uniform limits on a closed interval containing the values of . Finally, it is worth noting that in most applications the operators with the stated properties are given in integral form by means of families of measures, so for them the assertion follows directly from Jensen's inequality.
Remark 2.2. If instead of the condition we impose the weaker condition , where , and the function is required to satisfy the global -condition, then from (2.1) we obtain the estimate
Remark 2.3. (i) It is known (see [64], §4.1, Theorem 1.3, for example) that for every bounded operator on there exists a non-negative operator with the same norm and defined on non-negative functions by
such that for all functions in
For a contraction this gives the estimate
(ii) For every bounded operator on there also exists a non-negative operator on with the same norm (and defined by the same formula) for which (2.3) is true. If the operator is a contraction on , then so is . This was proved by Kantorovich [49] for more general lattices, and was later rediscovered by several authors in studying operators on (see [33], [30], and also the comments in [64], §4.1). Thus, any operator with finite norms on and is majorized by a non-negative operator with preservation of both norms. For a semigroup of operators , the operators constructed by the indicated formula do not always form a semigroup, but it was proved in [51] and [65] (see also [64], §7.2, Theorem 2.7) that if is a contractive operator semigroup on , then there exists a contractive semigroup of non-negative operators for which , and moreover, for every contractive semigroup of non-negative operators with the property that . Unlike the case of a single operator, it is important here that the semigroup be contractive (see [51]). If the maps are continuous on for all , then this property is inherited by the semigroup . Finally, it follows from the proofs given in these works that if for all , then .
The following technical lemma will be useful below.
Lemma 2.4. Let and . Then the function
is continuous on . The same is true if and , where and , or if the function is concentrated on a bounded interval and there is an -function such that and are integrable, where is the complementary function for .
If is an arbitrary non-negative measurable function on , then
where infinite values of integrals and suprema are allowed.
Proof. It is well known that for every measurable function on the functions converge to in measure on every compact interval as in . This implies the continuity of the map from to the space and also to the space or the Orlicz space provided that belongs to this space. Hence, the integral of a product with the function is continuous in .
The last assertion of the lemma is valid for bounded functions by the continuity in of the corresponding integrals, which increase pointwise to the (possibly, infinite) integral containing . Then the suprema for increase to the supremum for .
We recall that bounded countably additive measures on the Borel -algebra of a topological space are called Borel measures. A non-negative Borel measure is called a Radon measure if for any Borel set and any , there exists a compact set such that . A signed Borel measure is said to be Radon if its total variation defined by is Radon, where and are the positive and negative parts of , respectively (information about measures can be found in [15]).
We also recall that a Hausdorff topological space said to be Souslin if it is the image of some complete separable metric space under a continuous map. On Souslin spaces all Borel measures are Radon.
A sequence of Borel measures on a topological space is said to be weakly convergent to a Borel measure if for every bounded continuous function on
This convergence is weaker than the convergence for every Borel set. However, in the case of Radon probability measures on completely regular spaces weak convergence is equivalent to convergence on all Borel sets such that the topological boundary of (the difference between the closure and the interior) has zero -measure. About weak convergence of measures see [15] and [19].
§ 3. Non-uniform Kozlov–Treschev averagings and the ergodic theorem
We have already stated the classical ergodic theorem on convergence of averages almost everywhere, that is, the so-called individual ergodic theorem (which differs from ergodic theorems on convergence in some norm). This theorem has interesting and non-trivial generalizations (proved by other means) to the case of a semigroup of bounded linear operators on . We shall present several such generalizations. First we recall that a semigroup of bounded linear operators on a Banach space is said to be strongly measurable if for every the map with values in is Lebesgue measurable. Such a semigroup is said to be strongly integrable on compact intervals if the indicated maps are Bochner integrable on every compact interval. Strong measurability implies the continuity of the map for , but it does not imply the strong continuity of the semigroup, which is continuity of the map also at zero (see [4]). It is known (see [34], Theorem III.11.17) that for any Bochner integrable map there exists a real function on which is integrable with respect to the product of Lebesgue measure and (and which is measurable with respect to the product of the corresponding -algebras) such that for almost every the function is a representative of the equivalence class of the element . In [23] (Exercise 1.8.14) a simple construction is described that, in the case of a separable measure (that is, a measure with separable ) and a family of -integrable functions with and a measurable space such that the integrals of the functions over sets in are measurable in , enables one to construct an -measurable function with almost everywhere for every fixed . In place of separability of the measure it suffices to have separability of the subspace generated by all the , so this construction applies to any continuous map from to . With the aid of the indicated version of one can give meaning to the integral of with respect to for any fixed . The situation is similar for general Borel measures on instead of Lebesgue measure. Throughout, in integrals we mean such versions. It is useful to note that if a function on is continuous in the first argument and -measurable in the second argument, then it is -measurable (see [15], Lemma 6.4.6). In addition, Lemma 2.4 enables us, under the conditions we are considering, to compute the maximal functions introduced below by using rational .
The following result was proved in the paper [33] by Dunford and Schwartz (which they included in §11 of Chap. IV and §7 of Chap. VIII of their well-known monograph [34]).
Theorem 3.1. Let be a strongly measurable operator semigroup on such that
Then for every function the averages
converge -almost everywhere as .
The limit function is some projection of the function onto the closed subspace of all functions in that are invariant with respect to the operators . However, one should bear in mind that, unlike the Hilbert space , a projection onto a closed subspace in need not exist or may be non-unique. If the are Markov operators, which means that for and , and the measure is invariant with respect to them, that is,
which also gives the equality , then coincides with the conditional expectation of with respect to the -algebra generated by all functions in that are invariant with respect to the . If only constants have this property, then the limit in the theorem equals the integral of . For general semigroups the limit function can differ from the conditional expectation. For example, if , then the limit is zero.
The same authors (see [34], Theorem VIII.7.7) obtained the following useful estimate. It employs the following convention. It is known (see [15], Theorem 4.7.1) that, given a family of - measurable functions for which there exists a -measurable function with the property that for each we have almost everywhere (such a function is called a lattice upper bound of the given family), there exists an at most countable set of indices such that for each we have almost everywhere. Any other lattice upper bound of the given family is almost everywhere. This enables one to define the lattice least upper bound (lattice supremum) as a measurable function by taking the supremum over a countable subset. Of course, with this convention one can obtain a function that is smaller than the usual , and moreover, the usual supremum can give a non-measurable function. Nevertheless, if and the function is measurable in both arguments, then the usual supremum with respect to will also give a -measurable function (see [15], Corollary 2.12.8 and Exercise 6.10.42), but it is not always the lattice supremum (for example, if for and for on an interval with Lebesgue measure, then the usual supremum equals while the lattice supremum is ).
Theorem 3.2. Suppose that in the situation of the previous theorem , where , is a family of functions such that there exists a function with for all . Let
(in the sense of the lattice supremum). Then in the case
In the case , and under the additional assumption that ,
Note that in the case of a single function (which is our interest here), to compute we can take the supremum over rational by the continuity of in , and for this we do not need the general remark made above.
It is known that the estimates and imply the inequality for all . It was noted in [37] (see also [2]) that in the case of non-negative operators the last inequality gives the estimate .
In [41] estimates for were studied in the weighted space for a strongly continuous semigroup of non-negative operators on , and it was shown that the boundedness of in is equivalent to the uniform boundedness of the norms of the operators themselves on .
In the study of convergence almost everywhere the following theorem of Banach [8] (see also [9]) has proved useful. In this theorem is the space of equivalence classes of -measurable functions equipped with the topology of convergence in measure, which is generated by the metric
making a complete metric topological vector space.
Theorem 3.3. Let be a Banach space and let be a sequence of continuous linear operators such that for every in some dense set the limit exists for -almost all . Suppose also that for every
Then for every the limit
exists almost everywhere, and is a continuous linear operator.
In addition, an analogous assertion is true for a family of continuous linear operators , , with replaced by , under the additional condition of the uniform continuity of in measure on compact intervals, that is, under the condition that for every and every pair of numbers and there exists a such that for every vector with one has the estimate for all .
The last condition of uniform continuity in measure is satisfied in all the situations of interest for us, because for this condition to hold it suffices that the norm be bounded by some number on the interval , since by the Chebyshev inequality this gives us that
If in this theorem of Banach one takes everywhere finite versions of the measurable functions , then can be replaced by .
In [32], [33], and [34], Theorem IV.11.3, the reader can find the following generalization of Banach's result.
Theorem 3.4. Let be countable sets. Suppose that for every a continuous linear operator from a Banach space to is given such that for every
Suppose also that for every in some dense subset of ,
Then (3.1) is true for all .
In our situation this can be applied to the countable set of times and its subsets .
Below we consider applications of these theorems to generalizations of the Kozlov–Treschev theorem, but we note at once that even without these rather subtle facts one can easily obtain the assertion (i) from the next proposition. The assertion (ii) follows from the Banach theorem (it is easy to verify that its hypotheses are satisfied).
Proposition 3.5. (i) Let be a Banach space continuously embedded in , let be a Banach space continuously embedded in , and let be a family of continuous linear operators from to such that the functions belong to for -almost all when , , and . Suppose that for every pair and with the estimate
holds for all and almost all . If the limit exists almost everywhere for elements and of sets dense in and , respectively, then this is true for all and .
(ii) This assertion remains valid if the estimate above is replaced by the weaker estimate
where is a non-negative function on that is continuous in and measurable in .
Example 3.6. The Kozlov–Treschev theorem follows from this proposition by the classical ergodic theorem if we take and , since contains a dense subset of step functions with bounded support, that is, finite linear combinations of the indicator functions of intervals, and for the indicator function of an interval the assertion is true by the classical ergodic theorem (since it is true for the indicator functions of and , which can be verified by a change of variable, as in the case of the interval ). Moreover, it is obvious that
This observation clearly remains in force in the more general situation of Theorem 3.1. Actually, we only need here to have the conclusion of Theorem 3.1 for bounded functions together with the estimate .
For and one can take suitable Orlicz spaces, provided that we have managed to estimate the maximal function .
We note straightaway that convergence of in is not a problem, because if we have the estimate for some probability measure , then we get the estimate
which implies the convergence of in for all provided there is convergence for all elements in some dense set, say, for bounded functions or functions with finitely many values. Therefore, for every sequence of points tending to infinity there is a subsequence for which the functions converge -almost everywhere. If and , then the same estimate implies convergence in .
Here we are interested not in methods for giving an independent proof of the existence of a pointwise limit for non-uniform averages (although we will say something about such methods), but rather in ways to derive it from the case of the usual averaging. One such way was already pointed out by Dunford and Schwartz [33], who observed that for a monotonically decreasing positive integrable function on ,
where the classical maximal function generated by is defined by
In our situation we have the same inequality:
It follows from the case using the change of variable . We explain the justification in this case. For a continuously differentiable function , we get by integrating by parts that
which can be estimated from above by the quantity
Letting , we obtain the desired estimate for a continuously differentiable decreasing function , and the general case is deduced by means of approximations (though in the previous calculations we could use the Lebesgue–Stieltjes integral for a general monotone function ). The inequality for gives (3.3) with the absolute value.
Thus, the result of Banach stated above implies the following assertion.
Theorem 3.7. Suppose that for a strongly measurable semigroup of operators on the classical theorem on the existence of limits of the functions almost everywhere holds for all . If the density of the measure is estimated from above by an integrable monotonically decreasing function, , and for each the function is integrable on , then for every function
almost everywhere. If the measure is invariant with respect to the operators and there are no non-constant functions which are invariant with respect to the semigroup , then
for -almost all .
Proof. The integrability of ensures the existence of for almost all , since the integral of with respect to the measure turns out to be finite. The inequality (3.3) ensures that almost everywhere for all . In order to apply Banach's theorem, we have to verify that the assertion of the theorem is true for bounded . This follows from Proposition 3.5 applied to and , where as a dense subset one should take the set of step functions with bounded support (for them convergence is ensured by the classical theorem, the validity of which for the given semigroup is assumed in our hypotheses).
Corollary 3.8. Let be a strongly measurable semigroup of operators on such that and . If the density of the measure is estimated from above by an integrable monotonically decreasing function, then the conclusion of the previous theorem holds for every function .
Remark 3.9. The Banach theorem cited above leads to the following somewhat more general conclusion: if under the conditions on the semigroup indicated in Theorem 3.7 the equality (3.4) holds for all for some density , then it remains valid for every density , where is a constant. Note also that one can combine different conditions on the density by writing it as a sum or by partitioning the half-line.
Unlike the case of the classical averaging, where the convergence is for all integrable functions , in the Kozlov–Treschev construction it is necessary to impose certain restrictions on or on the connection between and . Let us consider a modification of the example in [20].
Example 3.10. Let be the unit circle with normalized Lebesgue measure , and let the transformation be rotation by the angle , that is, let for , where . Also, let
There is an absolutely continuous probability measure with support in such that for all . Moreover, the density of can be taken in with some , and the function belongs to with .
Indeed, let and . We take the probability measure with density concentrated on the set
by specifying
At all other points in we set , and the constant is taken so that is a probability measure. Then
It is known (see [50], §8) that for every there exists an infinite set of pairs of natural numbers and such that
For such triples , , we find that
Therefore, by the equality we have
Hence . Note that for .
Remark 3.11. The following result of Stein ([86], p. 73) is well known. Let be a strongly continuous operator semigroup on such that and and such that these operators are self-adjoint on . Then the maximal function
has the property that for every there exists a number such that for all
In our situation we obtain from this estimate the inequalities
Thus, for every absolutely continuous measure and every function with one has convergence of almost everywhere. This follows from the Banach theorem (for which one only needs the estimate almost everywhere), but is easily verified directly. However, the operators generated by transformations of the space are usually not symmetric on . On the contrary, general operator semigroups arising in applications are frequently symmetric. For example, the Ornstein–Uhlenbeck semigroup (see [18], [14], [16], and [17]) is defined by the formula
on the space with respect to the standard Gaussian measure with density on the real line. The Ornstein–Uhlenbeck semigroup is non-negative and symmetric on the space , and moreover, it is contractive on all the spaces . The generator of this semigroup on is the Ornstein–Uhlenbeck operator given by on the space of Sobolev functions with respect to . The measure is invariant with respect to this semigroup, that is,
Note that in the results presented here we do not assume that the measure is invariant with respect to the operators (of course, for the operators generated by measure-preserving transformations this will be automatically the case), but if is nevertheless invariant for non-negative operators , then . If, in addition, (that is, the operators are Markov), then .
The case of a measure with bounded support has some features making it possible to apply the classical ergodic theorem. Suppose that for a -integrable function we choose some -function on satisfying the global -condition and such that the function is also integrable. Such a choice is always possible, as explained in §2. Suppose in addition that the following condition holds for a strongly measurable semigroup of bounded operators on : for some
This condition holds with by Lemma 2.1 if the operators are continuous and non-negative on and are contractions on . It holds with some if the operators are continuous and non-negative on and their norms on are bounded by .
Lemma 3.12. Suppose that the condition (3.6) is satisfied and the operators are non-negative. Let be the complementary function for . If the density on the interval is such that the function is integrable on , then -almost everywhere
Moreover, this inequality holds if and .
Proof. By Young's inequality, for every
The first integral on the right-hand side does not exceed
which gives the desired estimate. In the case of a contractive semigroup on and we can take the dominating semigroup of non-negative operators indicated in Remark 2.3. These are contractions on and on , and hence they satisfy (3.6), so the inequality (3.7) holds for them, and then it holds also for the original semigroup.
The Banach theorem stated above and the convergence of the classical averages for bounded functions imply the following assertion in view of Lemma 3.12.
Theorem 3.13. Let be a strongly measurable operator semigroup on such that
Suppose that the density is concentrated on the interval and the function is integrable on , where is the complementary function for . Then the equality (3.4) holds for -almost all .
Below we will consider the maximal function also for non-uniform averagings.
Example 3.14. Let be a strongly measurable operator semigroup on such that
Suppose that for some , and the density is concentrated on the interval , where . Then (3.4) holds for -almost all .
This assertion is contained in [53], where the condition (3.6) was used without justification in the proof, but according to Lemma 2.1 this is legitimate (yet another inaccuracy in the proof in [53] is that it employs the estimate (3.8) below, for which the condition is needed, but according to what we proved above the result is true for as well).
It was shown above that the condition (3.6) enables us to obtain an estimate that can then be combined with the Banach theorem. However, the convergence we are interested in can be proved directly on the basis of (3.6), without using the maximal function (see below). For semigroups generated by transformations of the space Theorem 3.13 was proved in [20] (as noted above, it also follows from results in [28] and [29]).
Remark 3.15. (i) Suppose that the operators are Markov and the measure is invariant for them. We observe that in the proof of convergence the general case reduces to the case . Indeed, for convergence is trivial, since by the invariance of -measurable integrable functions with respect to . Moreover, we have .
When we use a convex function satisfying the global -condition and such that the function is integrable, we find that once is replaced by the difference the function will also be integrable, because
by the -condition, and the function is estimated from above by the integrable function according to Jensen's inequality (see [15], Proposition 10.1.9).
(ii) Let us see how Theorem 3.13 can be proved directly in this situation without using the Banach theorem. Recall that the chosen function satisfies the -condition. As indicated in (i), it suffices to consider the case where . For each we introduce the bounded cut-off functions
which converge to pointwise and in , and moreover, the compositions converge to in the same way. We write as
The functions converge pointwise to zero and do not exceed in absolute value. Hence, the integrals of the functions converge to zero. Therefore, the integrals of the conditional expectations also converge to zero. Passing to a subsequence, we can assume that almost everywhere. The functions converge in to , that is, to zero in our situation. Passing to a subsequence once again, we can assume that almost everywhere.
For any the bounded function is such that almost everywhere as . Let us take a set of unit measure for all of whose points there is such convergence for each and also the convergence and the convergence .
We show that for all . Let . By the -condition there exists a such that if the integral of is less than (see [62], Theorem 9.4). Fix such that
Next we take such that
which is possible by the boundedness of . Hence
Further, we have the estimate
As , the right-hand side tends to . Hence, there exists a such that the left-hand side is less than for all . Therefore, for such the function satisfies . We finally obtain
which completes the proof.
In [53] there is a proof of the following assertion giving convergence for families of functions in the case of a measurable semiflow of measure-preserving transformations of the space .
Theorem 3.16. Let be some set of -measurable functions on the space such that . Also, suppose that the function is measurable with respect to , the density has support in an interval , , where and , and for . Let be a positive function on such that and . If almost everywhere as , then
for -almost all .
We mention a result in [53] for maximal functions generated by a weight (it was assumed there that the operators are non-negative, but according to Remark 2.3 the estimate remains valid even without this assumption). We return to these functions below.
Theorem 3.17. Let be a strongly measurable operator semigroup on such that
In addition, suppose that , where , and that there exists a decreasing function such that for some
Then for any function with
Note that by a condition of the theorem, and hence the weight need not belong to .
According to what we said above, the supremum on the left-hand side of (3.8) can be taken over a countable dense set.
Now it is appropriate to mention a number of results in the papers [28] and [29] involving convergence of averages of the form
defined by a probability density with bounded support on the half-line in the case of a semiflow of transformations preserving the measure , and also convergence of more general averages of the form
for transformations preserving the measure and indexed by points such that and .
In the cited papers convergence of such averages was investigated using the Lorentz class defined for any probability density on and consisting of equivalence classes of -measurable functions with finite quantity
where is the decreasing equimeasurable rearrangement of the function , that is, the decreasing right-continuous function on for which
(where is Lebesgue measure). On Lorentz classes of the indicated form see [72] and [63], Chap. II, §5. In the case where the function is decreasing (and only in this case), the Lorentz class is a normed space with the norm (the fact that this is indeed a norm is not obvious and needs verification). Then this space is a Banach space.
The following results were proved in [28].
Theorem 3.18. Let be a probability density on the cube with Lebesgue measure, let be its decreasing equimeasurable rearrangement on , and let . Then exists -almost everywhere.
The justification employs some estimates for the maximal function
(the same as in Theorem 3.17).
Theorem 3.19. There is a number depending only on , with , such that for all non-negative functions and all
In place of the unit cube in one can use any cube with normalized Lebesgue measure, and this also lets us use the function . It is interesting to consider the case of a weight with non-compact support (there are also Lorentz classes for such weights; see [72] and [63], Chap. II, §5).
In addition, it was shown in [28] that if and is an increasing weight on , then for the existence of the limit the inclusion is necessary in the following sense: if and , where is Lebesgue measure, then there exists a function on which is equimeasurable with and such that, for the flow of shifts by (mod ), convergence of fails for every .
It would be interesting to obtain analogues of these results for operator semigroups on . It is clear that Theorem 3.18 implies Theorem 3.13 in the case of the semigroup generated by the transformations , because in the situation of Theorem 3.13 the integral of is estimated by the sum of the integrals of and over the interval , where the integral of over with respect to Lebesgue measure equals the integral of with respect to the measure . One can derive Theorem 3.18 from Theorem 3.19 by means of the Banach theorem. Indeed, the estimate for the maximal function shows that the measure of the set of points with is zero, for if it is equal to a number , then the integral on the right-hand side is not less than , but this integral is not greater than the integral of because is decreasing, and is integrable since , all of which leads to a contradiction as . Thus, the function is bounded with respect to for almost all , and the bounded functions (for which the limit exists, as we know) are dense in the Banach space (see [63], Chap. II, §5). It would be interesting to continue the study of the maximal function (3.9), which is more naturally connected with the summation method for general densities. This is also important for obtaining conditions for convergence without using the classical averaging, which undoubtedly should lead to more general and sharper results.
The next assertion gives conditions enabling us to get rid of boundedness of the support or monotonicity of by requiring some estimates at infinity which admit arbitrarily high peaks (see [20] for the proof).
Theorem 3.20. Suppose that the semigroup is generated by a semigroup of measure-preserving transformations of the space . Suppose also that the density of the measure satisfies the following condition: there exist positive numbers such that for , , and . Let be a -integrable function. Then the equality
holds for -almost all , and if the semiflow is ergodic, then (3.5) holds.
Is it possible to omit the absolute continuity of the averaging measure ? It is readily seen that in many cases the presence of atoms of the measure excludes convergence of . For example, if is just the Dirac measure at the point , then . For the group of rotations of the circle by the angle there is no limit of as , so it is easy to find even a continuous function for which has no limit for any . The case of an atomless measure is more interesting, but also here the limit can fail to exist even for bounded continuous functions in the same example of the group of rotations by the angle on the circle with Lebesgue measure . Indeed, it is known (see [73], §2.2) that there exists an atomless singular Borel probability measure on such that its Fourier transform has no limit at infinity. Then for the simplest function we get that has no limit as for any . However, it is not clear whether there exists a singular probability measure on such that for the group of rotations under consideration the limit of exists -almost everywhere as for every bounded Borel function (or every function ). If we admit only continuous functions in this setting, then the answer is affirmative. Indeed, then it suffices to have a limit for all polynomials in and , that is, it suffices that the Fourier transform of the measure have zero limit at infinity. It is well known that there exist singular measures with this property (see §2.2 in [73]). In this situation the existence of the limit of for all turns out to be equivalent to the property that the maximal function is finite -almost everywhere for such . The case of continuous is discussed in the next section. Of course, convergence of to in can hold for singular measures too. For example, this is the case for the group of rotations of the circle with Lebesgue measure and every measure whose Fourier transform tends to zero at infinity. This is clear from the convergence already noted on continuous functions along with the estimate , which holds for all probability measures .
Korolev [54] obtained an analogue of the well-known Wiener–Wintner theorem for the case of non-uniform averaging. The latter theorem [93] asserts the following for the usual averages: for any ergodic semigroup of measure-preserving transformations and every integrable function , there exists a set with such that for all and the averages
have a limit as , and if the semigroup is weakly mixing and , then this limit is .
We recall that a semigroup of measure-preserving transformations of the space is said to be weakly mixing (see [31], Chap. 1, §6) if for every two functions
This property implies ergodicity. An analogous property is introduced for operator semigroups.
Consider the non-uniform averages
Korolev [54] proved the following assertion.
Theorem 3.21. Let be a weakly mixing semigroup. Then for every function and every probability density on with bounded support, where and , there exists a set such that and for all and
In connection with the Wiener–Wintner theorem we mention the papers [83] and [10], and also [68] and [74], where singular averagings were considered. It would be interesting to extend these results to the case of operator semigroups. It is also of interest to study the rate of convergence of non-uniform averages, which has already been investigated for the usual averages (see [47], [48]). Finally, the question arises of possible analogues of results in [45] and [46] in the situation under consideration.
§ 4. Non-uniform averagings for stochastic systems
In [21] non-uniform averagings were considered for stochastic equations. The main distinction of the stochastic case from the deterministic case is the absence of the semigroup property with respect to the time.
Let be a continuous map on with values in the space of linear operators on , let be a Borel vector field on , and let , , be a -dimensional Wiener process. We shall assume that , , is the coordinate process on , where is the space of all continuous trajectories on the half-line with the topology of uniform convergence on compact sets, is the Borel -algebra, and is the Wiener measure.
We consider the stochastic differential equation
The formal generator of the diffusion defined by this equation has the form
Suppose that the following conditions are satisfied:
(i) , where is the local Sobolev class of functions in with first-order generalized derivatives in , and moreover, , , where is a positive constant, and is a locally bounded map;
(ii) there exists a function such that the sets are compact and
With the aid of results in [92] one can show that there exists a strong solution of the indicated equation. It is also known (see [23] and [22]) that the process obtained has a unique invariant probability measure with a positive continuous density of class with respect to Lebesgue measure, and the strongly continuous semigroup on generated by the process is strongly Feller, that is, it takes functions in to continuous functions, and the measure is ergodic with respect to this semigroup, that is, if for all , then the function coincides with a constant almost everywhere. For what follows, only the listed properties of the process and its transition semigroup will be essential.
We denote by the image of the Wiener measure under the map defined by . It is readily seen that the measures are weakly continuous in , and hence the map is measurable for every set . We define a probability measure on the space by
On the path space the semigroup of shifts acts by the formula . It is known (see [85], Chap. 1, §1.2) that the measure on is ergodic if and only if is ergodic, and this holds under our assumptions.
Let , where . It follows from the Birkhoff–Khinchin ergodic theorem that for -almost all
Let be the class of all -invariant sets. The ergodicity of implies that the sets in have -measure or .
The following ergodic theorem with the usual averaging holds (variants of this theorem can be found in [85], Chap. 1, §1.2, Theorem 7, and [66], §1.3, and [21] contains a short proof in precisely the given formulation).
Theorem 4.1. Let the conditions (i) and (ii) hold. Then for every Borel function in
for -almost all and for every .
Since , the previous theorem gives the following assertion.
Corollary 4.2. Let the conditions (i) and (ii) hold. Then for any Borel function in and any
for -almost all .
We proceed to Kozlov–Treschev type averaging in the stochastic case. Let be an absolutely continuous probability measure on with density with respect to Lebesgue measure. For every function the averages
are defined for -almost all , where is the Wiener measure. For these averages we obtain the following assertion.
Corollary 4.3. Let the conditions (i) and (ii) hold. Then for every bounded Borel function on and for every
for -almost all .
For functions in the following assertion is valid.
Theorem 4.4. Let the conditions (i) and (ii) hold and let and , where is a probability density and . Assume one of the conditions
(i) the density has bounded support in the interval ,
(ii) and there exists a non-decreasing function on such that , , and on for some .
Then for any and for -almost all
Besides non-uniform averages of stochastic systems, it would be interesting to apply analogous ideas to convergence of solutions of non-linear parabolic equations to stationary distributions (see [24]–[26], [84]), and also to analysis of attractors (see [52]).
§ 5. Dynamics of measures
Non-uniform Kozlov–Treschev averagings generated by transformations of preserving the measure motivate the consideration of a family of probability measures on defined in the following way: the measure is the image of the measure under the map
By the definition of the image of a measure, the integral of a -measurable bounded function with respect to the measure is given by the formula
Here we discuss the character of convergence of the measures to the measure . In the case of an ergodic system the result due to Kozlov and Treschev has some similarity to convergence of the measures to in the topology of convergence on every set. However, this convergence can fail literally, because convergence for almost all appears only after the integration of every fixed function , and therefore the corresponding measure-zero set can depend on . For example, consider the ergodic group of shifts along trajectories of the differential equation
(with incommensurable and ) on the two-dimensional torus with the normalized Lebesgue measure , which is invariant with respect to . For every Borel probability measure on , all the measures are singular with respect to , since every measure is concentrated on a curve of -measure zero. Hence for no can one have convergence of these measures to on every set (although for an absolutely continuous measure one has weak convergence, as we shall see below). In this circle of questions it turns out to be natural to employ the topological structure of the space , because we shall be discussing weak convergence of measures with respect to the duality with the space of bounded continuous functions (see §2). As above, we assume the joint measurability of with respect to .
The next results were proved in [20].
Theorem 5.1. Let be a Radon probability measure on a completely regular topological space such that all compact sets in are metrizable (for example, is itself metrizable or is a Souslin space). Suppose that the semiflow is ergodic. Let the measure be absolutely continuous with respect to Lebesgue measure on . Then for -almost all the measures converge weakly to as .
We recall that a family of Radon measures on a space is said to be uniformly tight if for any there exists a compact set such that
According to the Prohorov theorem, every weakly convergent sequence of Borel measures on a complete separable metric space is uniformly tight. It is also known that every uniformly tight family of Radon measures that is bounded in variation on a completely regular space has a compact closure in the weak topology (see [15] or [19]). If the space is Souslin, then every sequence of measures in such a family contains a weakly convergent subsequence.
Theorem 5.2. Suppose that is a Souslin (or metric) space and is a Radon probability measure such that the semiflow is ergodic. Let the measure be absolutely continuous. Then for any there exists a compact set such that and the family of measures with and is uniformly tight.
Since is the Dirac measure at the point , in the case of a non-compact space one cannot manage without removing some part of the space and making an indentation from zero with respect to . However, it is not clear whether it is enough just to remove some part of the space.
We now discuss analogous questions for singular atomless measures on with Fourier transform given by
Example 5.3. Let be the unit circle with the normalized Lebesgue measure and let the transformation be rotation by the angle . If the Fourier transform of tends to zero at infinity, then the measures converge weakly to for all as .
Conversely, if one has weak convergence for some , then the Fourier transform of tends to zero at infinity.
Indeed, for the function or on we have, respectively,
If for some we have weak convergence of the measures , then has some limits and at and . We show that both limits are zero. It is known (see [73], Theorem 3.2.3) that the limit
equals the size of the jump at the point of the distribution function of the probability measure , which equals zero in our case due to the absence of atoms of . Taking , we see that this is possible only in the case when . The convolution has no atom at zero either, so for its Fourier transform the indicated limit at also equals zero. Hence , and thus .
If we are given that as , then for the functions of the form with non-zero integer the integrals with respect to the measures tend to zero. The integral of equals . Then convergence holds for finite linear combinations of such functions, but they uniformly approximate all continuous functions on the circle.
In the stochastic case considered in §4, as in the case of the deterministic semigroup, there arise the measures on defined as the images of the measure under the maps
that is, the integral of a bounded Borel function with respect to the measure is given by the formula
We mention two results obtained in [21]. As in §4, we assume that the probability space for our system is the path space with the Wiener measure .
Theorem 5.4. Suppose that the conditions (i) and (ii) after (4.1) hold, is the solution of (4.1), and is the corresponding invariant probability measure for the process. Let be an absolutely continuous Borel probability measure on . Then for any and for -almost all the measures converge weakly to the measure as .
There is also an analogue of the theorem on uniform tightness.
Theorem 5.5. Under the hypotheses of the previous theorem, for any there exists a compact set such that and the family of measures
is uniformly tight.
The author thanks F.-Y. Wang and S. V. Shaposhnikov for useful discussions.