You are currently browsing the category archive for the ‘math.CA’ category.

In the traditional foundations of probability theory, one selects a probability space {(\Omega, {\mathcal B}, {\mathbf P})}, and makes a distinction between deterministic mathematical objects, which do not depend on the sampled state {\omega \in \Omega}, and stochastic (or random) mathematical objects, which do depend (but in a measurable fashion) on the sampled state {\omega \in \Omega}. For instance, a deterministic real number would just be an element {x \in {\bf R}}, whereas a stochastic real number (or real random variable) would be a measurable function {x: \Omega \rightarrow {\bf R}}, where in this post {{\bf R}} will always be endowed with the Borel {\sigma}-algebra. (For readers familiar with nonstandard analysis, the adjectives “deterministic” and “stochastic” will be used here in a manner analogous to the uses of the adjectives “standard” and “nonstandard” in nonstandard analysis. The analogy is particularly close when comparing with the “cheap nonstandard analysis” discussed in this previous blog post. We will also use “relative to {\Omega}” as a synonym for “stochastic”.)

Actually, for our purposes we will adopt the philosophy of identifying stochastic objects that agree almost surely, so if one was to be completely precise, we should define a stochastic real number to be an equivalence class {[x]} of measurable functions {x: \Omega \rightarrow {\bf R}}, up to almost sure equivalence. However, we shall often abuse notation and write {[x]} simply as {x}.

More generally, given any measurable space {X = (X, {\mathcal X})}, we can talk either about deterministic elements {x \in X}, or about stochastic elements of {X}, that is to say equivalence classes {[x]} of measurable maps {x: \Omega \rightarrow X} up to almost sure equivalence. We will use {\Gamma(X|\Omega)} to denote the set of all stochastic elements of {X}. (For readers familiar with sheaves, it may helpful for the purposes of this post to think of {\Gamma(X|\Omega)} as the space of measurable global sections of the trivial {X}-bundle over {\Omega}.) Of course every deterministic element {x} of {X} can also be viewed as a stochastic element {x|\Omega \in \Gamma(X|\Omega)} given by (the equivalence class of) the constant function {\omega \mapsto x}, thus giving an embedding of {X} into {\Gamma(X|\Omega)}. We do not attempt here to give an interpretation of {\Gamma(X|\Omega)} for sets {X} that are not equipped with a {\sigma}-algebra {{\mathcal X}}.

Remark 1 In my previous post on the foundations of probability theory, I emphasised the freedom to extend the sample space {(\Omega, {\mathcal B}, {\mathbf P})} to a larger sample space whenever one wished to inject additional sources of randomness. This is of course an important freedom to possess (and in the current formalism, is the analogue of the important operation of base change in algebraic geometry), but in this post we will focus on a single fixed sample space {(\Omega, {\mathcal B}, {\mathbf P})}, and not consider extensions of this space, so that one only has to consider two types of mathematical objects (deterministic and stochastic), as opposed to having many more such types, one for each potential choice of sample space (with the deterministic objects corresponding to the case when the sample space collapses to a point).

Any (measurable) {k}-ary operation on deterministic mathematical objects then extends to their stochastic counterparts by applying the operation pointwise. For instance, the addition operation {+: {\bf R} \times {\bf R} \rightarrow {\bf R}} on deterministic real numbers extends to an addition operation {+: \Gamma({\bf R}|\Omega) \times \Gamma({\bf R}|\Omega) \rightarrow \Gamma({\bf R}|\Omega)}, by defining the class {[x]+[y]} for {x,y: \Omega \rightarrow {\bf R}} to be the equivalence class of the function {\omega \mapsto x(\omega) + y(\omega)}; this operation is easily seen to be well-defined. More generally, any measurable {k}-ary deterministic operation {O: X_1 \times \dots \times X_k \rightarrow Y} between measurable spaces {X_1,\dots,X_k,Y} extends to an stochastic operation {O: \Gamma(X_1|\Omega) \times \dots \Gamma(X_k|\Omega) \rightarrow \Gamma(Y|\Omega)} in the obvious manner.

There is a similar story for {k}-ary relations {R: X_1 \times \dots \times X_k \rightarrow \{\hbox{true},\hbox{false}\}}, although here one has to make a distinction between a deterministic reading of the relation and a stochastic one. Namely, if we are given stochastic objects {x_i \in \Gamma(X_i|\Omega)} for {i=1,\dots,k}, the relation {R(x_1,\dots,x_k)} does not necessarily take values in the deterministic Boolean algebra {\{ \hbox{true}, \hbox{false}\}}, but only in the stochastic Boolean algebra {\Gamma(\{ \hbox{true}, \hbox{false}\}|\Omega)} – thus {R(x_1,\dots,x_k)} may be true with some positive probability and also false with some positive probability (with the event that {R(x_1,\dots,x_k)} being stochastically true being determined up to null events). Of course, the deterministic Boolean algebra embeds in the stochastic one, so we can talk about a relation {R(x_1,\dots,x_k)} being determinstically true or deterministically false, which (due to our identification of stochastic objects that agree almost surely) means that {R(x_1(\omega),\dots,x_k(\omega))} is almost surely true or almost surely false respectively. For instance given two stochastic objects {x,y}, one can view their equality relation {x=y} as having a stochastic truth value. This is distinct from the way the equality symbol {=} is used in mathematical logic, which we will now call “equality in the deterministic sense” to reduce confusion. Thus, {x=y} in the deterministic sense if and only if the stochastic truth value of {x=y} is equal to {\hbox{true}}, that is to say that {x(\omega)=y(\omega)} for almost all {\omega}.

Any universal identity for deterministic operations (or universal implication between identities) extends to their stochastic counterparts: for instance, addition is commutative, associative, and cancellative on the space of deterministic reals {{\bf R}}, and is therefore commutative, associative, and cancellative on stochastic reals {\Gamma({\bf R}|\Omega)} as well. However, one has to be more careful when working with mathematical laws that are not expressible as universal identities, or implications between identities. For instance, {{\bf R}} is an integral domain: if {x_1,x_2 \in {\bf R}} are deterministic reals such that {x_1 x_2=0}, then one must have {x_1=0} or {x_2=0}. However, if {x_1, x_2 \in \Gamma({\bf R}|\Omega)} are stochastic reals such that {x_1 x_2 = 0} (in the deterministic sense), then it is no longer necessarily the case that {x_1=0} (in the deterministic sense) or that {x_2=0} (in the deterministic sense); however, it is still true that “{x_1=0} or {x_2=0}” is true in the deterministic sense if one interprets the boolean operator “or” stochastically, thus “{x_1(\omega)=0} or {x_2(\omega)=0}” is true for almost all {\omega}. Another way to properly obtain a stochastic interpretation of the integral domain property of {{\bf R}} is to rewrite it as

\displaystyle  x_1,x_2 \in {\bf R}, x_1 x_2 = 0 \implies x_i=0 \hbox{ for some } i \in \{1,2\}

and then make all sets stochastic to obtain the true statement

\displaystyle  x_1,x_2 \in \Gamma({\bf R}|\Omega), x_1 x_2 = 0 \implies x_i=0 \hbox{ for some } i \in \Gamma(\{1,2\}|\Omega),

thus we have to allow the index {i} for which vanishing {x_i=0} occurs to also be stochastic, rather than deterministic. (A technical note: when one proves this statement, one has to select {i} in a measurable fashion; for instance, one can choose {i(\omega)} to equal {1} when {x_1(\omega)=0}, and {2} otherwise (so that in the “tie-breaking” case when {x_1(\omega)} and {x_2(\omega)} both vanish, one always selects {i(\omega)} to equal {1}).)

Similarly, the law of the excluded middle fails when interpreted deterministically, but remains true when interpreted stochastically: if {S} is a stochastic statement, then it is not necessarily the case that {S} is either deterministically true or deterministically false; however the sentence “{S} or not-{S}” is still deterministically true if the boolean operator “or” is interpreted stochastically rather than deterministically.

To avoid having to keep pointing out which operations are interpreted stochastically and which ones are interpreted deterministically, we will use the following convention: if we assert that a mathematical sentence {S} involving stochastic objects is true, then (unless otherwise specified) we mean that {S} is deterministically true, assuming that all relations used inside {S} are interpreted stochastically. For instance, if {x,y} are stochastic reals, when we assert that “Exactly one of {x < y}, {x=y}, or {x>y} is true”, then by default it is understood that the relations {<}, {=}, {>} and the boolean operator “exactly one of” are interpreted stochastically, and the assertion is that the sentence is deterministically true.

In the above discussion, the stochastic objects {x} being considered were elements of a deterministic space {X}, such as the reals {{\bf R}}. However, it can often be convenient to generalise this situation by allowing the ambient space {X} to also be stochastic. For instance, one might wish to consider a stochastic vector {v(\omega)} inside a stochastic vector space {V(\omega)}, or a stochastic edge {e} of a stochastic graph {G(\omega)}. In order to formally describe this situation within the classical framework of measure theory, one needs to place all the ambient spaces {X(\omega)} inside a measurable space. This can certainly be done in many contexts (e.g. when considering random graphs on a deterministic set of vertices, or if one is willing to work up to equivalence and place the ambient spaces inside a suitable moduli space), but is not completely natural in other contexts. For instance, if one wishes to consider stochastic vector spaces of potentially unbounded dimension (in particular, potentially larger than any given cardinal that one might specify in advance), then the class of all possible vector spaces is so large that it becomes a proper class rather than a set (even if one works up to equivalence), making it problematic to give this class the structure of a measurable space; furthermore, even once one does so, one needs to take additional care to pin down what it would mean for a random vector {\omega \mapsto v_\omega} lying in a random vector space {\omega \mapsto V_\omega} to depend “measurably” on {\omega}.

Of course, in any reasonable application one can avoid the set theoretic issues at least by various ad hoc means, for instance by restricting the dimension of all spaces involved to some fixed cardinal such as {2^{\aleph_0}}. However, the measure-theoretic issues can require some additional effort to resolve properly.

In this post I would like to describe a different way to formalise stochastic spaces, and stochastic elements of these spaces, by viewing the spaces as measure-theoretic analogue of a sheaf, but being over the probability space {\Omega} rather than over a topological space; stochastic objects are then sections of such sheaves. Actually, for minor technical reasons it is convenient to work in the slightly more general setting in which the base space {\Omega} is a finite measure space {(\Omega, {\mathcal B}, \mu)} rather than a probability space, thus {\mu(\Omega)} can take any value in {[0,+\infty)} rather than being normalised to equal {1}. This will allow us to easily localise to subevents {\Omega'} of {\Omega} without the need for normalisation, even when {\Omega'} is a null event (though we caution that the map {x \mapsto x|\Omega'} from deterministic objects {x} ceases to be injective in this latter case). We will however still continue to use probabilistic terminology. despite the lack of normalisation; thus for instance, sets {E} in {{\mathcal B}} will be referred to as events, the measure {\mu(E)} of such a set will be referred to as the probability (which is now permitted to exceed {1} in some cases), and an event whose complement is a null event shall be said to hold almost surely. It is in fact likely that almost all of the theory below extends to base spaces which are {\sigma}-finite rather than finite (for instance, by damping the measure to become finite, without introducing any further null events), although we will not pursue this further generalisation here.

The approach taken in this post is “topos-theoretic” in nature (although we will not use the language of topoi explicitly here), and is well suited to a “pointless” or “point-free” approach to probability theory, in which the role of the stochastic state {\omega \in \Omega} is suppressed as much as possible; instead, one strives to always adopt a “relative point of view”, with all objects under consideration being viewed as stochastic objects relative to the underlying base space {\Omega}. In this perspective, the stochastic version of a set is as follows.

Definition 1 (Stochastic set) Unless otherwise specified, we assume that we are given a fixed finite measure space {\Omega = (\Omega, {\mathcal B}, \mu)} (which we refer to as the base space). A stochastic set (relative to {\Omega}) is a tuple {X|\Omega = (\Gamma(X|E)_{E \in {\mathcal B}}, ((|E))_{E \subset F, E,F \in {\mathcal B}})} consisting of the following objects:

  • A set {\Gamma(X|E)} assigned to each event {E \in {\mathcal B}}; and
  • A restriction map {x \mapsto x|E} from {\Gamma(X|F)} to {\Gamma(X|E)} to each pair {E \subset F} of nested events {E,F \in {\mathcal B}}. (Strictly speaking, one should indicate the dependence on {F} in the notation for the restriction map, e.g. using {x \mapsto x|)E \leftarrow F)} instead of {x \mapsto x|E}, but we will abuse notation by omitting the {F} dependence.)

We refer to elements of {\Gamma(X|E)} as local stochastic elements of the stochastic set {X|\Omega}, localised to the event {E}, and elements of {\Gamma(X|\Omega)} as global stochastic elements (or simply elements) of the stochastic set. (In the language of sheaves, one would use “sections” instead of “elements” here, but I prefer to use the latter terminology here, for compatibility with conventional probabilistic notation, where for instance measurable maps from {\Omega} to {{\bf R}} are referred to as real random variables, rather than sections of the reals.)

Furthermore, we impose the following axioms:

  • (Category) The map {x \mapsto x|E} from {\Gamma(X|E)} to {\Gamma(X|E)} is the identity map, and if {E \subset F \subset G} are events in {{\mathcal B}}, then {((x|F)|E) = (x|E)} for all {x \in \Gamma(X|G)}.
  • (Null events trivial) If {E \in {\mathcal B}} is a null event, then the set {\Gamma(X|E)} is a singleton set. (In particular, {\Gamma(X|\emptyset)} is always a singleton set; this is analogous to the convention that {x^0=1} for any number {x}.)
  • (Countable gluing) Suppose that for each natural number {n}, one has an event {E_n \in {\mathcal B}} and an element {x_n \in \Gamma(X|E_n)} such that {x_n|(E_n \cap E_m) = x_m|(E_n \cap E_m)} for all {n,m}. Then there exists a unique {x\in \Gamma(X|\bigcup_{n=1}^\infty E_n)} such that {x_n = x|E_n} for all {n}.

If {\Omega'} is an event in {\Omega}, we define the localisation {X|\Omega'} of the stochastic set {X|\Omega} to {\Omega'} to be the stochastic set

\displaystyle X|\Omega' := (\Gamma(X|E)_{E \in {\mathcal B}; E \subset \Omega'}, ((|E))_{E \subset F \subset \Omega', E,F \in {\mathcal B}})

relative to {\Omega'}. (Note that there is no need to renormalise the measure on {\Omega'}, as we are not demanding that our base space have total measure {1}.)

The following fact is useful for actually verifying that a given object indeed has the structure of a stochastic set:

Exercise 1 Show that to verify the countable gluing axiom of a stochastic set, it suffices to do so under the additional hypothesis that the events {E_n} are disjoint. (Note that this is quite different from the situation with sheaves over a topological space, in which the analogous gluing axiom is often trivial in the disjoint case but has non-trivial content in the overlapping case. This is ultimately because a {\sigma}-algebra is closed under all Boolean operations, whereas a topology is only closed under union and intersection.)

Let us illustrate the concept of a stochastic set with some examples.

Example 1 (Discrete case) A simple case arises when {\Omega} is a discrete space which is at most countable. If we assign a set {X_\omega} to each {\omega \in \Omega}, with {X_\omega} a singleton if {\mu(\{\omega\})=0}. One then sets {\Gamma(X|E) := \prod_{\omega \in E} X_\omega}, with the obvious restriction maps, giving rise to a stochastic set {X|\Omega}. (Thus, a local element {x} of {\Gamma(X|E)} can be viewed as a map {\omega \mapsto x(\omega)} on {E} that takes values in {X_\omega} for each {\omega \in E}.) Conversely, it is not difficult to see that any stochastic set over an at most countable discrete probability space {\Omega} is of this form up to isomorphism. In this case, one can think of {X|\Omega} as a bundle of sets {X_\omega} over each point {\omega} (of positive probability) in the base space {\Omega}. One can extend this bundle interpretation of stochastic sets to reasonably nice sample spaces {\Omega} (such as standard Borel spaces) and similarly reasonable {X}; however, I would like to avoid this interpretation in the formalism below in order to be able to easily work in settings in which {\Omega} and {X} are very “large” (e.g. not separable in any reasonable sense). Note that we permit some of the {X_\omega} to be empty, thus it can be possible for {\Gamma(X|\Omega)} to be empty whilst {\Gamma(X|E)} for some strict subevents {E} of {\Omega} to be non-empty. (This is analogous to how it is possible for a sheaf to have local sections but no global sections.) As such, the space {\Gamma(X|\Omega)} of global elements does not completely determine the stochastic set {X|\Omega}; one sometimes needs to localise to an event {E} in order to see the full structure of such a set. Thus it is important to distinguish between a stochastic set {X|\Omega} and its space {\Gamma(X|\Omega)} of global elements. (As such, it is a slight abuse of the axiom of extensionality to refer to global elements of {X|\Omega} simply as “elements”, but hopefully this should not cause too much confusion.)

Example 2 (Measurable spaces as stochastic sets) Returning now to a general base space {\Omega}, any (deterministic) measurable space {X} gives rise to a stochastic set {X|\Omega}, with {\Gamma(X|E)} being defined as in previous discussion as the measurable functions from {E} to {X} modulo almost everywhere equivalence (in particular, {\Gamma(X|E)} a singleton set when {E} is null), with the usual restriction maps. The constraint of measurability on the maps {x: E \rightarrow \Omega}, together with the quotienting by almost sure equivalence, means that {\Gamma(X|E)} is now more complicated than a plain Cartesian product {\prod_{\omega \in E} X_\omega} of fibres, but this still serves as a useful first approximation to what {\Gamma(X|E)} is for the purposes of developing intuition. Indeed, the measurability constraint is so weak (as compared for instance to topological or smooth constraints in other contexts, such as sheaves of continuous or smooth sections of bundles) that the intuition of essentially independent fibres is quite an accurate one, at least if one avoids consideration of an uncountable number of objects simultaneously.

Example 3 (Hilbert modules) This example is the one that motivated this post for me. Suppose that one has an extension {(\tilde \Omega, \tilde {\mathcal B}, \tilde \mu)} of the base space {(\Omega, {\mathcal B},\mu)}, thus we have a measurable factor map {\pi: \tilde \Omega \rightarrow \Omega} such that the pushforward of the measure {\tilde \mu} by {\pi} is equal to {\mu}. Then we have a conditional expectation operator {\pi_*: L^2(\tilde \Omega,\tilde {\mathcal B},\tilde \mu) \rightarrow L^2(\Omega,{\mathcal B},\mu)}, defined as the adjoint of the pullback map {\pi^*: L^2(\Omega,{\mathcal B},\mu) \rightarrow L^2(\tilde \Omega,\tilde {\mathcal B},\tilde \mu)}. As is well known, the conditional expectation operator also extends to a contraction {\pi_*: L^1(\tilde \Omega,\tilde {\mathcal B},\tilde \mu) \rightarrow L^1(\Omega,{\mathcal B}, \mu)}. We then define the Hilbert module {L^2(\tilde \Omega|\Omega)} to be the space of functions {f \in L^2(\tilde \Omega,\tilde {\mathcal B},\tilde \mu)} with {\pi_*(|f|^2) \in L^\infty( \Omega, {\mathcal B}, \mu )}; this is a Hilbert module over {L^\infty(\Omega, {\mathcal B}, \mu)} which is of particular importance in the Furstenberg-Zimmer structure theory of measure-preserving systems. We can then define the stochastic set {L^2_\pi(\tilde \Omega)|\Omega} by setting

\displaystyle  \Gamma(L^2_\pi(\tilde \Omega)|E) := L^2( \pi^{-1}(E) | E )

with the obvious restriction maps. In the case that {\Omega,\Omega'} are standard Borel spaces, one can disintegrate {\mu'} as an integral {\mu' = \int_\Omega \nu_\omega\ d\mu(\omega)} of probability measures {\nu_\omega} (supported in the fibre {\pi^{-1}(\{\omega\})}), in which case this stochastic set can be viewed as having fibres {L^2( \tilde \Omega, \tilde {\mathcal B}, \nu_\omega )} (though if {\Omega} is not discrete, there are still some measurability conditions in {\omega} on the local and global elements that need to be imposed). However, I am interested in the case when {\Omega,\Omega'} are not standard Borel spaces (in fact, I will take them to be algebraic probability spaces, as defined in this previous post), in which case disintegrations are not available. However, it appears that the stochastic analysis developed in this blog post can serve as a substitute for the tool of disintegration in this context.

We make the remark that if {X|\Omega} is a stochastic set and {E, F} are events that are equivalent up to null events, then one can identify {\Gamma(X|E)} with {\Gamma(X|F)} (through their common restriction to {\Gamma(X|(E \cap F))}, with the restriction maps now being bijections). As such, the notion of a stochastic set does not require the full structure of a concrete probability space {(\Omega, {\mathcal B}, {\mathbf P})}; one could also have defined the notion using only the abstract {\sigma}-algebra consisting of {{\mathcal B}} modulo null events as the base space, or equivalently one could define stochastic sets over the algebraic probability spaces defined in this previous post. However, we will stick with the classical formalism of concrete probability spaces here so as to keep the notation reasonably familiar.

As a corollary of the above observation, we see that if the base space {\Omega} has total measure {0}, then all stochastic sets are trivial (they are just points).

Exercise 2 If {X|\Omega} is a stochastic set, show that there exists an event {\Omega'} with the property that for any event {E}, {\Gamma(X|E)} is non-empty if and only if {E} is contained in {\Omega'} modulo null events. (In particular, {\Omega'} is unique up to null events.) Hint: consider the numbers {\mu( E )} for {E} ranging over all events with {\Gamma(X|E)} non-empty, and form a maximising sequence for these numbers. Then use all three axioms of a stochastic set.

One can now start take many of the fundamental objects, operations, and results in set theory (and, hence, in most other categories of mathematics) and establish analogues relative to a finite measure space. Implicitly, what we will be doing in the next few paragraphs is endowing the category of stochastic sets with the structure of an elementary topos. However, to keep things reasonably concrete, we will not explicitly emphasise the topos-theoretic formalism here, although it is certainly lurking in the background.

Firstly, we define a stochastic function {f: X|\Omega \rightarrow Y|\Omega} between two stochastic sets {X|\Omega, Y|\Omega} to be a collection of maps {f: \Gamma(X|E) \rightarrow \Gamma(Y|E)} for each {E \in {\mathcal B}} which form a natural transformation in the sense that {f(x|E) = f(x)|E} for all {x \in \Gamma(X|F)} and nested events {E \subset F}. In the case when {\Omega} is discrete and at most countable (and after deleting all null points), a stochastic function is nothing more than a collection of functions {f_\omega: X_\omega \rightarrow Y_\omega} for each {\omega \in \Omega}, with the function {f: \Gamma(X|E) \rightarrow \Gamma(Y|E)} then being a direct sum of the factor functions {f_\omega}:

\displaystyle  f( (x_\omega)_{\omega \in E} ) = ( f_\omega(x_\omega) )_{\omega \in E}.

Thus (in the discrete, at most countable setting, at least) stochastic functions do not mix together information from different states {\omega} in a sample space; the value of {f(x)} at {\omega} depends only on the value of {x} at {\omega}. The situation is a bit more subtle for continuous probability spaces, due to the identification of stochastic objects that agree almost surely, nevertheness it is still good intuition to think of stochastic functions as essentially being “pointwise” or “local” in nature.

One can now form the stochastic set {\hbox{Hom}(X \rightarrow Y)|\Omega} of functions from {X|\Omega} to {Y|\Omega}, by setting {\Gamma(\hbox{Hom}(X \rightarrow Y)|E)} for any event {E} to be the set of local stochastic functions {f: X|E \rightarrow Y|E} of the localisations of {X|\Omega, Y|\Omega} to {E}; this is a stochastic set if we use the obvious restriction maps. In the case when {\Omega} is discrete and at most countable, the fibre {\hbox{Hom}(X \rightarrow Y)_\omega} at a point {\omega} of positive measure is simply the set {Y_\omega^{X_\omega}} of functions from {X_\omega} to {Y_\omega}.

In a similar spirit, we say that one stochastic set {Y|\Omega} is a (stochastic) subset of another {X|\Omega}, and write {Y|\Omega \subset X|\Omega}, if we have a stochastic inclusion map, thus {\Gamma(Y|E) \subset \Gamma(X|E)} for all events {E}, with the restriction maps being compatible. We can then define the power set {2^X|\Omega} of a stochastic set {X|\Omega} by setting {\Gamma(2^X|E)} for any event {E} to be the set of all stochastic subsets {Y|E} of {X|E} relative to {E}; it is easy to see that {2^X|\Omega} is a stochastic set with the obvious restriction maps (one can also identify {2^X|\Omega} with {\hbox{Hom}(X, \{\hbox{true},\hbox{false}\})|\Omega} in the obvious fashion). Again, when {\Omega} is discrete and at most countable, the fibre of {2^X|\Omega} at a point {\omega} of positive measure is simply the deterministic power set {2^{X_\omega}}.

Note that if {f: X|\Omega \rightarrow Y|\Omega} is a stochastic function and {Y'|\Omega} is a stochastic subset of {Y|\Omega}, then the inverse image {f^{-1}(Y')|\Omega}, defined by setting {\Gamma(f^{-1}(Y')|E)} for any event {E} to be the set of those {x \in \Gamma(X|E)} with {f(x) \in \Gamma(Y'|E)}, is a stochastic subset of {X|\Omega}. In particular, given a {k}-ary relation {R: X_1 \times \dots \times X_k|\Omega \rightarrow \{\hbox{true}, \hbox{false}\}|\Omega}, the inverse image {R^{-1}( \{ \hbox{true} \}|\Omega )} is a stochastic subset of {X_1 \times \dots \times X_k|\Omega}, which by abuse of notation we denote as

\displaystyle  \{ (x_1,\dots,x_k) \in X_1 \times \dots \times X_k: R(x_1,\dots,x_k) \hbox{ is true} \}|\Omega.

In a similar spirit, if {X'|\Omega} is a stochastic subset of {X|\Omega} and {f: X|\Omega \rightarrow Y|\Omega} is a stochastic function, we can define the image {f(X')|\Omega} by setting {\Gamma(f(X')|E)} to be the set of those {f(x)} with {x \in \Gamma(X'|E)}; one easily verifies that this is a stochastic subset of {Y|\Omega}.

Remark 2 One should caution that in the definition of the subset relation {Y|\Omega \subset X|\Omega}, it is important that {\Gamma(Y|E) \subset \Gamma(X|E)} for all events {E}, not just the global event {\Omega}; in particular, just because a stochastic set {X|\Omega} has no global sections, does not mean that it is contained in the stochastic empty set {\emptyset|\Omega}.

Now we discuss Boolean operations on stochastic subsets of a given stochastic set {X|\Omega}. Given two stochastic subsets {X_1|\Omega, X_2|\Omega} of {X|\Omega}, the stochastic intersection {(X_1 \cap X_2)|\Omega} is defined by setting {\Gamma((X_1 \cap X_2)|E)} to be the set of {x \in \Gamma(X|E)} that lie in both {\Gamma(X_1|E)} and {\Gamma(X_2|E)}:

\displaystyle  \Gamma(X_1 \cap X_2)|E) := \Gamma(X_1|E) \cap \Gamma(X_2|E).

This is easily verified to again be a stochastic subset of {X|\Omega}. More generally one may define stochastic countable intersections {(\bigcap_{n=1}^\infty X_n)|\Omega} for any sequence {X_n|\Omega} of stochastic subsets of {X|\Omega}. One could extend this definition to uncountable families if one wished, but I would advise against it, because some of the usual laws of Boolean algebra (e.g. the de Morgan laws) may break down in this setting.

Stochastic unions are a bit more subtle. The set {\Gamma((X_1 \cup X_2)|E)} should not be defined to simply be the union of {\Gamma(X_1|E)} and {\Gamma(X_2|E)}, as this would not respect the gluing axiom. Instead, we define {\Gamma((X_1 \cup X_2)|E)} to be the set of all {x \in \Gamma(X|E)} such that one can cover {E} by measurable subevents {E_1,E_2} such that {x_i|E_i \in \Gamma(X_i|E_i)} for {i=1,2}; then {(X_1 \cup X_2)|\Omega} may be verified to be a stochastic subset of {X|\Omega}. Thus for instance {\{0,1\}|\Omega} is the stochastic union of {\{0\}|\Omega} and {\{1\}|\Omega}. Similarly for countable unions {(\bigcup_{n=1}^\infty X_n)|\Omega} of stochastic subsets {X_n|\Omega} of {X|\Omega}, although for uncountable unions are extremely problematic (they are disliked by both the measure theory and the countable gluing axiom) and will not be defined here. Finally, the stochastic difference set {\Gamma((X_1 \backslash X_2)|E)} is defined as the set of all {x|E} in {\Gamma(X_1|E)} such that {x|F \not \in \Gamma(X_2|F)} for any subevent {F} of {E} of positive probability. One may verify that in the case when {\Omega} is discrete and at most countable, these Boolean operations correspond to the classical Boolean operations applied separately to each fibre {X_{i,\omega}} of the relevant sets {X_i}. We also leave as an exercise to the reader to verify the usual laws of Boolean arithmetic, e.g. the de Morgan laws, provided that one works with at most countable unions and intersections.

One can also consider a stochastic finite union {(\bigcup_{n=1}^N X_n)|\Omega} in which the number {N} of sets in the union is itself stochastic. More precisely, let {X|\Omega} be a stochastic set, let {N \in {\bf N}|\Omega} be a stochastic natural number, and let {n \mapsto X_n|\Omega} be a stochastic function from the stochastic set {\{ n \in {\bf N}: n \leq N\}|\Omega} (defined by setting {\Gamma(\{n \in {\bf N}: n\leq N\}|E) := \{ n \in {\bf N}|E: n \leq N|E\}})) to the stochastic power set {2^X|\Omega}. Here we are considering {0} to be a natural number, to allow for unions that are possibly empty, with {{\bf N}_+ := {\bf N} \backslash \{0\}} used for the positive natural numbers. We also write {(X_n)_{n=1}^N|\Omega} for the stochastic function {n \mapsto X_n|\Omega}. Then we can define the stochastic union {\bigcup_{n=1}^N X_n|\Omega} by setting {\Gamma(\bigcup_{n=1}^N X_n|E)} for an event {E} to be the set of local elements {x \in \Gamma(X|E)} with the property that there exists a covering of {E} by measurable subevents {E_{n_0}} for {n_0 \in {\bf N}_+}, such that one has {n_0 \leq N|E_{n_0}} and {x|E_{n_0} \in \Gamma(X_{n_0}|E_{n_0})}. One can verify that {\bigcup_{n=1}^N X_n|\Omega} is a stochastic set (with the obvious restriction maps). Again, in the model case when {\Omega} is discrete and at most countable, the fibre {(\bigcup_{n=1}^N X_n)_\omega} is what one would expect it to be, namely {\bigcup_{n=1}^{N(\omega)} (X_n)_\omega}.

The Cartesian product {(X \times Y)|\Omega} of two stochastic sets may be defined by setting {\Gamma((X \times Y)|E) := \Gamma(X|E) \times \Gamma(Y|E)} for all events {E}, with the obvious restriction maps; this is easily seen to be another stochastic set. This lets one define the concept of a {k}-ary operation {f: (X_1 \times \dots \times X_k)|\Omega \rightarrow Y|\Omega} from {k} stochastic sets {X_1,\dots,X_k} to another stochastic set {Y}, or a {k}-ary relation {R: (X_1 \times \dots \times X_k)|\Omega \rightarrow \{\hbox{true}, \hbox{false}\}|\Omega}. In particular, given {x_i \in X_i|\Omega} for {i=1,\dots,k}, the relation {R(x_1,\dots,x_k)} may be deterministically true, deterministically false, or have some other stochastic truth value.

Remark 3 In the degenerate case when {\Omega} is null, stochastic logic becomes a bit weird: all stochastic statements are deterministically true, as are their stochastic negations, since every event in {\Omega} (even the empty set) now holds with full probability. Among other pathologies, the empty set now has a global element over {\Omega} (this is analogous to the notorious convention {0^0=1}), and any two deterministic objects {x,y} become equal over {\Omega}: {x|\Omega=y|\Omega}.

The following simple observation is crucial to subsequent discussion. If {(x_n)_{n \in {\bf N}_+}} is a sequence taking values in the global elements {\Gamma(X|\Omega)} of a stochastic space {X|\Omega}, then we may also define global elements {x_n \in \Gamma(X|\Omega)} for stochastic indices {n \in {\bf N}_+|\Omega} as well, by appealing to the countable gluing axiom to glue together {x_{n_0}} restricted to the set {\{ \omega \in \Omega: n(\omega) = n_0\}} for each deterministic natural number {n_0} to form {x_n}. With this definition, the map {n \mapsto x_n} is a stochastic function from {{\bf N}_+|\Omega} to {X|\Omega}; indeed, this creates a one-to-one correspondence between external sequences (maps {n \mapsto x_n} from {{\bf N}_+} to {\Gamma(X|\Omega)}) and stochastic sequences (stochastic functions {n \mapsto x_n} from {{\bf N}_+|\Omega} to {X|\Omega}). Similarly with {{\bf N}_+} replaced by any other at most countable set. This observation will be important in allowing many deterministic arguments involving sequences will be able to be carried over to the stochastic setting.

We now specialise from the extremely broad discipline of set theory to the more focused discipline of real analysis. There are two fundamental axioms that underlie real analysis (and in particular distinguishes it from real algebra). The first is the Archimedean property, which we phrase in the “no infinitesimal” formulation as follows:

Proposition 2 (Archimedean property) Let {x \in {\bf R}} be such that {x \leq 1/n} for all positive natural numbers {n}. Then {x \leq 0}.

The other is the least upper bound axiom:

Proposition 3 (Least upper bound axiom) Let {S} be a non-empty subset of {{\bf R}} which has an upper bound {M \in {\bf R}}, thus {x \leq M} for all {x \in S}. Then there exists a unique real number {\sup S \in {\bf R}} with the following properties:

  • {x \leq \sup S} for all {x \in S}.
  • For any real {L < \sup S}, there exists {x \in S} such that {L < x \leq \sup S}.
  • {\sup S \leq M}.

Furthermore, {\sup S} does not depend on the choice of {M}.

The Archimedean property extends easily to the stochastic setting:

Proposition 4 (Stochastic Archimedean property) Let {x \in \Gamma({\bf R}|\Omega)} be such that {x \leq 1/n} for all deterministic natural numbers {n}. Then {x \leq 0}.

Remark 4 Here, incidentally, is one place in which this stochastic formalism deviates from the nonstandard analysis formalism, as the latter certainly permits the existence of infinitesimal elements. On the other hand, we caution that stochastic real numbers are permitted to be unbounded, so that formulation of Archimedean property is not valid in the stochastic setting.

The proof is easy and is left to the reader. The least upper bound axiom also extends nicely to the stochastic setting, but the proof requires more work (in particular, our argument uses the monotone convergence theorem):

Theorem 5 (Stochastic least upper bound axiom) Let {S|\Omega} be a stochastic subset of {{\bf R}|\Omega} which has a global upper bound {M \in {\bf R}|\Omega}, thus {x \leq M} for all {x \in \Gamma(S|\Omega)}, and is globally non-empty in the sense that there is at least one global element {x \in \Gamma(S|\Omega)}. Then there exists a unique stochastic real number {\sup S \in \Gamma({\bf R}|\Omega)} with the following properties:

  • {x \leq \sup S} for all {x \in \Gamma(S|\Omega)}.
  • For any stochastic real {L < \sup S}, there exists {x \in \Gamma(S|\Omega)} such that {L < x \leq \sup S}.
  • {\sup S \leq M}.

Furthermore, {\sup S} does not depend on the choice of {M}.

For future reference, we note that the same result holds with {{\bf R}} replaced by {{\bf N} \cup \{+\infty\}} throughout, since the latter may be embedded in the former, for instance by mapping {n} to {1 - \frac{1}{n+1}} and {+\infty} to {1}. In applications, the above theorem serves as a reasonable substitute for the countable axiom of choice, which does not appear to hold in unrestricted generality relative to a measure space; in particular, it can be used to generate various extremising sequences for stochastic functionals on various stochastic function spaces.

Proof: Uniqueness is clear (using the Archimedean property), as well as the independence on {M}, so we turn to existence. By using an order-preserving map from {{\bf R}} to {(-1,1)} (e.g. {x \mapsto \frac{2}{\pi} \hbox{arctan}(x)}) we may assume that {S|\Omega} is a subset of {(-1,1)|\Omega}, and that {M < 1}.

We observe that {\Gamma(S|\Omega)} is a lattice: if {x, y \in \Gamma(S|\Omega)}, then {\max(x,y)} and {\min(x,y)} also lie in {\Gamma(S|\Omega)}. Indeed, {\max(x,y)} may be formed by appealing to the countable gluing axiom to glue {y} (restricted the set {\{ \omega \in \Omega: x(\omega) < y(\omega) \}}) with {x} (restricted to the set {\{ \omega \in \Omega: x(\omega) \geq y(\omega) \}}), and similarly for {\min(x,y)}. (Here we use the fact that relations such as {<} are Borel measurable on {{\bf R}}.)

Let {A \in {\bf R}} denote the deterministic quantity

\displaystyle  A := \sup \{ \int_\Omega x(\omega)\ d\mu(\omega): x \in \Gamma(S|\Omega) \}

then (by Proposition 3!) {A} is well-defined; here we use the hypothesis that {\mu(\Omega)} is finite. Thus we may find a sequence {(x_n)_{n \in {\bf N}}} of elements {x_n} of {\Gamma(S|\Omega)} such that

\displaystyle  \int_\Omega x_n(\omega)\ d\mu(\omega) \rightarrow A \hbox{ as } n \rightarrow \infty. \ \ \ \ \ (1)

Using the lattice property, we may assume that the {x_n} are non-decreasing: {x_n \leq x_m} whenever {n \leq m}. If we then define {\sup S(\omega) := \sup_n x_n(\omega)} (after choosing measurable representatives of each equivalence class {x_n}), then {\sup S} is a stochastic real with {\sup S \leq M}.

If {x \in \Gamma(S|\Omega)}, then {\max(x,x_n) \in \Gamma(S|\Omega)}, and so

\displaystyle  \int_\Omega \max(x,x_n)\ d\mu(\omega) \leq A.

From this and (1) we conclude that

\displaystyle  \int_\Omega \max(x-x_n,0) \rightarrow 0 \hbox{ as } n \rightarrow \infty.

From monotone convergence, we conclude that

\displaystyle  \int_\Omega \max(x-\sup S,0) = 0

and so {x \leq \sup S}, as required.

Now let {L < \sup S} be a stochastic real. After choosing measurable representatives of each relevant equivalence class, we see that for almost every {\omega \in \Omega}, we can find a natural number {n(\omega)} with {x_{n(\omega)} > L}. If we choose {n(\omega)} to be the first such positive natural number when it exists, and (say) {1} otherwise, then {n} is a stochastic positive natural number and {L < x_n}. The claim follows. \Box

Using Proposition 4 and Theorem 5, one can then revisit many of the other foundational results of deterministic real analysis, and develop stochastic analogues; we give some examples of this below the fold (focusing on the Heine-Borel theorem and a case of the spectral theorem). As an application of this formalism, we revisit some of the Furstenberg-Zimmer structural theory of measure-preserving systems, particularly that of relatively compact and relatively weakly mixing systems, and interpret them in this framework, basically as stochastic versions of compact and weakly mixing systems (though with the caveat that the shift map is allowed to act non-trivially on the underlying probability space). As this formalism is “point-free”, in that it avoids explicit use of fibres and disintegrations, it will be well suited for generalising this structure theory to settings in which the underlying probability spaces are not standard Borel, and the underlying groups are uncountable; I hope to discuss such generalisations in future blog posts.

Remark 5 Roughly speaking, stochastic real analysis can be viewed as a restricted subset of classical real analysis in which all operations have to be “measurable” with respect to the base space. In particular, indiscriminate application of the axiom of choice is not permitted, and one should largely restrict oneself to performing countable unions and intersections rather than arbitrary unions or intersections. Presumably one can formalise this intuition with a suitable “countable transfer principle”, but I was not able to formulate a clean and general principle of this sort, instead verifying various assertions about stochastic objects by hand rather than by direct transfer from the deterministic setting. However, it would be desirable to have such a principle, since otherwise one is faced with the tedious task of redoing all the foundations of real analysis (or whatever other base theory of mathematics one is going to be working in) in the stochastic setting by carefully repeating all the arguments.

More generally, topos theory is a good formalism for capturing precisely the informal idea of performing mathematics with certain operations, such as the axiom of choice, the law of the excluded middle, or arbitrary unions and intersections, being somehow “prohibited” or otherwise “restricted”.

Read the rest of this entry »

Due to some requests, I’m uploading to my blog the slides for my recent talk in Segovia (for the birthday conference of Michael Cowling) on “Hilbert’s fifth problem and approximate groups“.  The slides cover essentially the same range of topics in this series of lecture notes, or in this text of mine, though of course in considerably less detail, given that the slides are meant to be presented in an hour.

As laid out in the foundational work of Kolmogorov, a classical probability space (or probability space for short) is a triplet {(X, {\mathcal X}, \mu)}, where {X} is a set, {{\mathcal X}} is a {\sigma}-algebra of subsets of {X}, and {\mu: {\mathcal X} \rightarrow [0,1]} is a countably additive probability measure on {{\mathcal X}}. Given such a space, one can form a number of interesting function spaces, including

  • the (real) Hilbert space {L^2(X, {\mathcal X}, \mu)} of square-integrable functions {f: X \rightarrow {\bf R}}, modulo {\mu}-almost everywhere equivalence, and with the positive definite inner product {\langle f, g\rangle_{L^2(X, {\mathcal X}, \mu)} := \int_X f g\ d\mu}; and
  • the unital commutative Banach algebra {L^\infty(X, {\mathcal X}, \mu)} of essentially bounded functions {f: X \rightarrow {\bf R}}, modulo {\mu}-almost everywhere equivalence, with {\|f\|_{L^\infty(X, {\mathcal X}, \mu)}} defined as the essential supremum of {|f|}.

There is also a trace {\tau = \tau_\mu: L^\infty(X, {\mathcal X}, \mu) \rightarrow {\bf C}} on {L^\infty} defined by integration: {\tau(f) := \int_X f\ d\mu}.

One can form the category {\mathbf{Prb}} of classical probability spaces, by defining a morphism {\phi: (X, {\mathcal X}, \mu) \rightarrow (Y, {\mathcal Y}, \nu)} between probability spaces to be a function {\phi: X \rightarrow Y} which is measurable (thus {\phi^{-1}(E) \in {\mathcal X}} for all {E \in {\mathcal Y}}) and measure-preserving (thus {\mu(\phi^{-1}(E)) = \nu(E)} for all {E \in {\mathcal Y}}).

Let us now abstract the algebraic features of these spaces as follows; for want of a better name, I will refer to this abstraction as an algebraic probability space, and is very similar to the non-commutative probability spaces studied in this previous post, except that these spaces are now commutative (and real).

Definition 1 An algebraic probability space is a pair {({\mathcal A}, \tau)} where

  • {{\mathcal A}} is a unital commutative real algebra;
  • {\tau: {\mathcal A} \rightarrow {\bf R}} is a homomorphism such that {\tau(1)=1} and {\tau( f^2 ) \geq 0} for all {f \in {\mathcal A}};
  • Every element {f} of {{\mathcal A}} is bounded in the sense that {\sup_{k \geq 1} \tau( f^{2k} )^{1/2k} < \infty}. (Technically, this isn’t an algebraic property, but I need it for technical reasons.)

A morphism {\phi: ({\mathcal A}_1, \tau_1) \rightarrow ({\mathcal A}_2, \tau_2)} is a homomorphism {\phi^*: {\mathcal A}_2 \rightarrow {\mathcal A}_1} which is trace-preserving, in the sense that {\tau_1(\phi^*(f)) = \tau_2(f)} for all {f \in {\mathcal A}_2}.

For want of a better name, I’ll denote the category of algebraic probability spaces as {\mathbf{AlgPrb}}. One can view this category as the opposite category to that of (a subcategory of) the category of tracial commutative real algebras. One could emphasise this opposite nature by denoting the algebraic probability space as {({\mathcal A}, \tau)^{op}} rather than {({\mathcal A},\tau)}; another suggestive (but slightly inaccurate) notation, inspired by the language of schemes, would be {\hbox{Spec}({\mathcal A},\tau)} rather than {({\mathcal A},\tau)}. However, we will not adopt these conventions here, and refer to algebraic probability spaces just by the pair {({\mathcal A},\tau)}.

By the previous discussion, we have a covariant functor {F: \textbf{Prb} \rightarrow \textbf{AlgPrb}} that takes a classical probability space {(X, {\mathcal X}, \mu)} to its algebraic counterpart {(L^\infty(X, {\mathcal X},\mu), \tau_\mu)}, with a morphism {\phi: (X, {\mathcal X}, \mu) \rightarrow (Y, {\mathcal Y}, \nu)} of classical probability spaces mapping to a morphism {F(\phi): (L^\infty(X, {\mathcal X},\mu), \tau_\mu) \rightarrow (L^\infty(Y, {\mathcal Y},\nu), \tau_\nu)} of the corresponding algebraic probability spaces by the formula

\displaystyle  F(\phi)^* f := f \circ \phi

for {f \in L^\infty(Y, {\mathcal Y}, \nu)}. One easily verifies that this is a functor.

In this post I would like to describe a functor {G: \textbf{AlgPrb} \rightarrow \textbf{Prb}} which partially inverts {F} (up to natural isomorphism), that is to say a recipe for starting with an algebraic probability space {({\mathcal A}, \tau)} and producing a classical probability space {(X, {\mathcal X}, \mu)}. This recipe is not new – it is basically the (commutative) Gelfand-Naimark-Segal construction (discussed in this previous post) combined with the Loomis-Sikorski theorem (discussed in this previous post). However, I wanted to put the construction in a single location for sake of reference. I also wanted to make the point that {F} and {G} are not complete inverses; there is a bit of information in the algebraic probability space (e.g. topological information) which is lost when passing back to the classical probability space. In some future posts, I would like to develop some ergodic theory using the algebraic foundations of probability theory rather than the classical foundations; this turns out to be convenient in the ergodic theory arising from nonstandard analysis (such as that described in this previous post), in which the groups involved are uncountable and the underlying spaces are not standard Borel spaces.

Let us describe how to construct the functor {G}, with details postponed to below the fold.

  1. Starting with an algebraic probability space {({\mathcal A}, \tau)}, form an inner product on {{\mathcal A}} by the formula {\langle f, g \rangle := \tau(fg)}, and also form the spectral radius {\rho(f) :=\lim_{k \rightarrow \infty} \tau(f^{2^k})^{1/2^k}}.
  2. The inner product is clearly positive semi-definite. Quotienting out the null vectors and taking completions, we arrive at a real Hilbert space {L^2 = L^2({\mathcal A},\tau)}, to which the trace {\tau} may be extended.
  3. Somewhat less obviously, the spectral radius is well-defined and gives a norm on {{\mathcal A}}. Taking {L^2} limits of sequences in {{\mathcal A}} of bounded spectral radius gives us a subspace {L^\infty = L^\infty({\mathcal A},\tau)} of {L^2} that has the structure of a real commutative Banach algebra.
  4. The idempotents {1_E} of the Banach algebra {L^\infty} may be indexed by elements {E} of an abstract {\sigma}-algebra {{\mathcal B}}.
  5. The Boolean algebra homomorphisms {\delta_x: {\mathcal B} \rightarrow \{0,1\}} (or equivalently, the real algebra homomorphisms {\iota_x: L^\infty \rightarrow {\bf R}}) may be indexed by elements {x} of a space {X}.
  6. Let {{\mathcal X}} denote the {\sigma}-algebra on {X} generated by the basic sets {\overline{E} := \{ x \in X: \delta_x(E) = 1 \}} for every {E \in {\mathcal B}}.
  7. Let {{\mathcal N}} be the {\sigma}-ideal of {{\mathcal X}} generated by the sets {\bigcap_n \overline{E_n}}, where {E_n \in {\mathcal B}} is a sequence with {\bigcap_n E_n = \emptyset}.
  8. One verifies that {{\mathcal B}} is isomorphic to {{\mathcal X}/{\mathcal N}}. Using this isomorphism, the trace {\tau} on {L^\infty} can be used to construct a countably additive measure {\mu} on {{\mathcal X}}. The classical probability space {(X, {\mathcal X}, \mu)} is then {G( {\mathcal A}, \tau )}, and the abstract spaces {L^2, L^\infty} may now be identified with their concrete counterparts {L^2(X, {\mathcal X}, \mu)}, {L^\infty(X, {\mathcal X}, \mu)}.
  9. Every algebraic probability space morphism {\phi: ({\mathcal A}_1,\tau_1) \rightarrow ({\mathcal A}_2,\tau_2)} generates a classical probability morphism {G(\phi): (X_1, {\mathcal X}_1, \mu_1) \rightarrow (X_2, {\mathcal X}_2, \mu_2)} via the formula

    \displaystyle  \delta_{G(\phi)(x_1)}( E_2 ) = \delta_{x_1}( \phi^*(E_2) )

    using a pullback operation {\phi^*} on the abstract {\sigma}-algebras {{\mathcal B}_1, {\mathcal B}_2} that can be defined by density.

Remark 1 The classical probability space {X} constructed by the functor {G} has some additional structure; namely {X} is a {\sigma}-Stone space (a Stone space with the property that the closure of any countable union of clopen sets is clopen), {{\mathcal X}} is the Baire {\sigma}-algebra (generated by the clopen sets), and the null sets are the meager sets. However, we will not use this additional structure here.

The partial inversion relationship between the functors {F: \textbf{Prb} \rightarrow \textbf{AlgPrb}} and {G: \textbf{AlgPrb} \rightarrow \textbf{Prb}} is given by the following assertion:

  1. There is a natural transformation from {F \circ G: \textbf{AlgPrb} \rightarrow \textbf{AlgPrb}} to the identity functor {I: \textbf{AlgPrb} \rightarrow \textbf{AlgPrb}}.

More informally: if one starts with an algebraic probability space {({\mathcal A},\tau)} and converts it back into a classical probability space {(X, {\mathcal X}, \mu)}, then there is a trace-preserving algebra homomorphism of {{\mathcal A}} to {L^\infty( X, {\mathcal X}, \mu )}, which respects morphisms of the algebraic probability space. While this relationship is far weaker than an equivalence of categories (which would require that {F \circ G} and {G \circ F} are both natural isomorphisms), it is still good enough to allow many ergodic theory problems formulated using classical probability spaces to be reformulated instead as an equivalent problem in algebraic probability spaces.

Remark 2 The opposite composition {G \circ F: \textbf{Prb} \rightarrow \textbf{Prb}} is a little odd: it takes an arbitrary probability space {(X, {\mathcal X}, \mu)} and returns a more complicated probability space {(X', {\mathcal X}', \mu')}, with {X'} being the space of homomorphisms {\iota_x: L^\infty(X, {\mathcal X}, \mu) \rightarrow {\bf R}}. while there is “morally” an embedding of {X} into {X'} using the evaluation map, this map does not exist in general because points in {X} may well have zero measure. However, if one takes a “pointless” approach and focuses just on the measure algebras {({\mathcal X}, \mu)}, {({\mathcal X}', \mu')}, then these algebras become naturally isomorphic after quotienting out by null sets.

Remark 3 An algebraic probability space captures a bit more structure than a classical probability space, because {{\mathcal A}} may be identified with a proper subset of {L^\infty} that describes the “regular” functions (or random variables) of the space. For instance, starting with the unit circle {{\bf R}/{\bf Z}} (with the usual Haar measure and the usual trace {\tau(f) = \int_{{\bf R}/{\bf Z}} f}), any unital subalgebra {{\mathcal A}} of {L^\infty({\bf R}/{\bf Z})} that is dense in {L^2({\bf R}/{\bf Z})} will generate the same classical probability space {G( {\mathcal A}, \tau )} on applying the functor {G}, namely one will get the space {({\bf R}/{\bf Z})'} of homomorphisms from {L^\infty({\bf R}/{\bf Z})} to {{\bf R}} (with the measure induced from {\tau}). Thus for instance {{\mathcal A}} could be the continuous functions {C( {\bf R}/{\bf Z} )}, the Wiener algebra {A({\bf R}/{\bf Z})} or the full space {L^\infty({\bf R}/{\bf Z})}, but the classical space {G( {\mathcal A}, \tau )} will be unable to distinguish these spaces from each other. In particular, the functor {F \circ G} loses information (roughly speaking, this functor takes an algebraic probability space and completes it to a von Neumann algebra, but then forgets exactly what algebra was initially used to create this completion). In ergodic theory, this sort of “extra structure” is traditionally encoded in topological terms, by assuming that the underlying probability space {X} has a nice topological structure (e.g. a standard Borel space); however, with the algebraic perspective one has the freedom to have non-topological notions of extra structure, by choosing {{\mathcal A}} to be something other than an algebra {C(X)} of continuous functions on a topological space. I hope to discuss one such example of extra structure (coming from the Gowers-Host-Kra theory of uniformity seminorms) in a later blog post (this generalises the example of the Wiener algebra given previously, which is encoding “Fourier structure”).

A small example of how one could use the functors {F, G} is as follows. Suppose one has a classical probability space {(X, {\mathcal X}, \mu)} with a measure-preserving action of an uncountable group {\Gamma}, which is only defined (and an action) up to almost everywhere equivalence; thus for instance for any set {E} and any {g, h \in \Gamma}, {T^{gh} E} and {T^g T^h E} might not be exactly equal, but only equal up to a null set. For similar reasons, an element {E} of the invariant factor {{\mathcal X}^\Gamma} might not be exactly invariant with respect to {\Gamma}, but instead one only has {T^g E} and {E} equal up to null sets for each {g \in \Gamma}. One might like to “clean up” the action of {\Gamma} to make it defined everywhere, and a genuine action everywhere, but this is not immediately achievable if {\Gamma} is uncountable, since the union of all the null sets where something bad occurs may cease to be a null set. However, by applying the functor {F}, each shift {T^g: X \rightarrow X} defines a morphism {T^g: L^\infty(X, {\mathcal X}, \mu) \rightarrow L^\infty(X, {\mathcal X}, \mu)} on the associated algebraic probability space (i.e. the Koopman operator), and then applying {G}, we obtain a shift {T^g: X' \rightarrow X'} on a new classical probability space {(X', {\mathcal X}', \mu')} which now gives a genuine measure-preserving action of {\Gamma}, and which is equivalent to the original action from a measure algebra standpoint. The invariant factor {{\mathcal X}^\Gamma} now consists of those sets in {{\mathcal X}'} which are genuinely {\Gamma}-invariant, not just up to null sets. (Basically, the classical probability space {(X', {\mathcal X}', \mu')} contains a Boolean algebra {\overline{\mathcal B}} with the property that every measurable set {A \in {\mathcal X}'} is equivalent up to null sets to precisely one set in {\overline{\mathcal B}}, allowing for a canonical “retraction” onto {\overline{\mathcal B}} that eliminates all null set issues.)

More indirectly, the functors {F, G} suggest that one should be able to develop a “pointless” form of ergodic theory, in which the underlying probability spaces are given algebraically rather than classically. I hope to give some more specific examples of this in later posts.

Read the rest of this entry »

There are a number of ways to construct the real numbers {{\bf R}}, for instance

  • as the metric completion of {{\bf Q}} (thus, {{\bf R}} is defined as the set of Cauchy sequences of rationals, modulo Cauchy equivalence);
  • as the space of Dedekind cuts on the rationals {{\bf Q}};
  • as the space of quasimorphisms {\phi: {\bf Z} \rightarrow {\bf Z}} on the integers, quotiented by bounded functions. (I believe this construction first appears in this paper of Street, who credits the idea to Schanuel, though the germ of this construction arguably goes all the way back to Eudoxus.)

There is also a fourth family of constructions that proceeds via nonstandard analysis, as a special case of what is known as the nonstandard hull construction. (Here I will assume some basic familiarity with nonstandard analysis and ultraproducts, as covered for instance in this previous blog post.) Given an unbounded nonstandard natural number {N \in {}^* {\bf N} \backslash {\bf N}}, one can define two external additive subgroups of the nonstandard integers {{}^* {\bf Z}}:

  • The group {O(N) := \{ n \in {}^* {\bf Z}: |n| \leq CN \hbox{ for some } C \in {\bf N} \}} of all nonstandard integers of magnitude less than or comparable to {N}; and
  • The group {o(N) := \{ n \in {}^* {\bf Z}: |n| \leq C^{-1} N \hbox{ for all } C \in {\bf N} \}} of nonstandard integers of magnitude infinitesimally smaller than {N}.

The group {o(N)} is a subgroup of {O(N)}, so we may form the quotient group {O(N)/o(N)}. This space is isomorphic to the reals {{\bf R}}, and can in fact be used to construct the reals:

Proposition 1 For any coset {n + o(N)} of {O(N)/o(N)}, there is a unique real number {\hbox{st} \frac{n}{N}} with the property that {\frac{n}{N} = \hbox{st} \frac{n}{N} + o(1)}. The map {n + o(N) \mapsto \hbox{st} \frac{n}{N}} is then an isomorphism between the additive groups {O(N)/o(N)} and {{\bf R}}.

Proof: Uniqueness is clear. For existence, observe that the set {\{ x \in {\bf R}: Nx \leq n + o(N) \}} is a Dedekind cut, and its supremum can be verified to have the required properties for {\hbox{st} \frac{n}{N}}. \Box

In a similar vein, we can view the unit interval {[0,1]} in the reals as the quotient

\displaystyle  [0,1] \equiv [N] / o(N) \ \ \ \ \ (1)

where {[N]} is the nonstandard (i.e. internal) set {\{ n \in {\bf N}: n \leq N \}}; of course, {[N]} is not a group, so one should interpret {[N]/o(N)} as the image of {[N]} under the quotient map {{}^* {\bf Z} \rightarrow {}^* {\bf Z} / o(N)} (or {O(N) \rightarrow O(N)/o(N)}, if one prefers). Or to put it another way, (1) asserts that {[0,1]} is the image of {[N]} with respect to the map {\pi: n \mapsto \hbox{st} \frac{n}{N}}.

In this post I would like to record a nice measure-theoretic version of the equivalence (1), which essentially appears already in standard texts on Loeb measure (see e.g. this text of Cutland). To describe the results, we must first quickly recall the construction of Loeb measure on {[N]}. Given an internal subset {A} of {[N]}, we may define the elementary measure {\mu_0(A)} of {A} by the formula

\displaystyle  \mu_0(A) := \hbox{st} \frac{|A|}{N}.

This is a finitely additive probability measure on the Boolean algebra of internal subsets of {[N]}. We can then construct the Loeb outer measure {\mu^*(A)} of any subset {A \subset [N]} in complete analogy with Lebesgue outer measure by the formula

\displaystyle  \mu^*(A) := \inf \sum_{n=1}^\infty \mu_0(A_n)

where {(A_n)_{n=1}^\infty} ranges over all sequences of internal subsets of {[N]} that cover {A}. We say that a subset {A} of {[N]} is Loeb measurable if, for any (standard) {\epsilon>0}, one can find an internal subset {B} of {[N]} which differs from {A} by a set of Loeb outer measure at most {\epsilon}, and in that case we define the Loeb measure {\mu(A)} of {A} to be {\mu^*(A)}. It is a routine matter to show (e.g. using the Carathéodory extension theorem) that the space {{\mathcal L}} of Loeb measurable sets is a {\sigma}-algebra, and that {\mu} is a countably additive probability measure on this space that extends the elementary measure {\mu_0}. Thus {[N]} now has the structure of a probability space {([N], {\mathcal L}, \mu)}.

Now, the group {o(N)} acts (Loeb-almost everywhere) on the probability space {[N]} by the addition map, thus {T^h n := n+h} for {n \in [N]} and {h \in o(N)} (excluding a set of Loeb measure zero where {n+h} exits {[N]}). This action is clearly seen to be measure-preserving. As such, we can form the invariant factor {Z^0_{o(N)}([N]) = ([N], {\mathcal L}^{o(N)}, \mu\downharpoonright_{{\mathcal L}^{o(N)}})}, defined by restricting attention to those Loeb measurable sets {A \subset [N]} with the property that {T^h A} is equal {\mu}-almost everywhere to {A} for each {h \in o(N)}.

The claim is then that this invariant factor is equivalent (up to almost everywhere equivalence) to the unit interval {[0,1]} with Lebesgue measure {m} (and the trivial action of {o(N)}), by the same factor map {\pi: n \mapsto \hbox{st} \frac{n}{N}} used in (1). More precisely:

Theorem 2 Given a set {A \in {\mathcal L}^{o(N)}}, there exists a Lebesgue measurable set {B \subset [0,1]}, unique up to {m}-a.e. equivalence, such that {A} is {\mu}-a.e. equivalent to the set {\pi^{-1}(B) := \{ n \in [N]: \hbox{st} \frac{n}{N} \in B \}}. Conversely, if {B \in [0,1]} is Lebesgue measurable, then {\pi^{-1}(B)} is in {{\mathcal L}^{o(N)}}, and {\mu( \pi^{-1}(B) ) = m( B )}.

More informally, we have the measure-theoretic version

\displaystyle  [0,1] \equiv Z^0_{o(N)}( [N] )

of (1).

Proof: We first prove the converse. It is clear that {\pi^{-1}(B)} is {o(N)}-invariant, so it suffices to show that {\pi^{-1}(B)} is Loeb measurable with Loeb measure {m(B)}. This is easily verified when {B} is an elementary set (a finite union of intervals). By countable subadditivity of outer measure, this implies that Loeb outer measure of {\pi^{-1}(E)} is bounded by the Lebesgue outer measure of {E} for any set {E \subset [0,1]}; since every Lebesgue measurable set differs from an elementary set by a set of arbitrarily small Lebesgue outer measure, the claim follows.

Now we establish the forward claim. Uniqueness is clear from the converse claim, so it suffices to show existence. Let {A \in {\mathcal L}^{o(N)}}. Let {\epsilon>0} be an arbitrary standard real number, then we can find an internal set {A_\epsilon \subset [N]} which differs from {A} by a set of Loeb measure at most {\epsilon}. As {A} is {o(N)}-invariant, we conclude that for every {h \in o(N)}, {A_\epsilon} and {T^h A_\epsilon} differ by a set of Loeb measure (and hence elementary measure) at most {2\epsilon}. By the (contrapositive of the) underspill principle, there must exist a standard {\delta>0} such that {A_\epsilon} and {T^h A_\epsilon} differ by a set of elementary measure at most {2\epsilon} for all {|h| \leq \delta N}. If we then define the nonstandard function {f_\epsilon: [N] \rightarrow {}^* {\bf R}} by the formula

\displaystyle  f(n) := \hbox{st} \frac{1}{\delta N} \sum_{m \in [N]: m \leq n \leq m+\delta N} 1_{A_\epsilon}(m),

then from the (nonstandard) triangle inequality we have

\displaystyle  \frac{1}{N} \sum_{n \in [N]} |f(n) - 1_{A_\epsilon}(n)| \leq 3\epsilon

(say). On the other hand, {f} has the Lipschitz continuity property

\displaystyle  |f(n)-f(m)| \leq \frac{2|n-m|}{\delta N}

and so in particular we see that

\displaystyle  \hbox{st} f(n) = \tilde f( \hbox{st} \frac{n}{N} )

for some Lipschitz continuous function {\tilde f: [0,1] \rightarrow [0,1]}. If we then let {E_\epsilon} be the set where {\tilde f \geq 1 - \sqrt{\epsilon}}, one can check that {A_\epsilon} differs from {\pi^{-1}(E_\epsilon)} by a set of Loeb outer measure {O(\sqrt{\epsilon})}, and hence {A} does so also. Sending {\epsilon} to zero, we see (from the converse claim) that {1_{E_\epsilon}} is a Cauchy sequence in {L^1} and thus converges in {L^1} for some Lebesgue measurable {E}. The sets {A_\epsilon} then converge in Loeb outer measure to {\pi^{-1}(E)}, giving the claim. \Box

Thanks to the Lebesgue differentiation theorem, the conditional expectation {{\bf E}( f | Z^0_{o(N)}([N]))} of a bounded Loeb-measurable function {f: [N] \rightarrow {\bf R}} can be expressed (as a function on {[0,1]}, defined {m}-a.e.) as

\displaystyle  {\bf E}( f | Z^0_{o(N)}([N]))(x) := \lim_{\epsilon \rightarrow 0} \frac{1}{2\epsilon} \int_{[x-\epsilon N,x+\epsilon N]} f\ d\mu.

By the abstract ergodic theorem from the previous post, one can also view this conditional expectation as the element in the closed convex hull of the shifts {T^h f}, {h = o(N)} of minimal {L^2} norm. In particular, we obtain a form of the von Neumann ergodic theorem in this context: the averages {\frac{1}{H} \sum_{h=1}^H T^h f} for {H=O(N)} converge (as a net, rather than a sequence) in {L^2} to {{\bf E}( f | Z^0_{o(N)}([N]))}.

If {f: [N] \rightarrow [-1,1]} is (the standard part of) an internal function, that is to say the ultralimit of a sequence {f_n: [N_n] \rightarrow [-1,1]} of finitary bounded functions, one can view the measurable function {F := {\bf E}( f | Z^0_{o(N)}([N]))} as a limit of the {f_n} that is analogous to the “graphons” that emerge as limits of graphs (see e.g. the recent text of Lovasz on graph limits). Indeed, the measurable function {F: [0,1] \rightarrow [-1,1]} is related to the discrete functions {f_n: [N_n] \rightarrow [-1,1]} by the formula

\displaystyle  \int_a^b F(x)\ dx = \hbox{st} \lim_{n \rightarrow p} \frac{1}{N_n} \sum_{a N_n \leq m \leq b N_n} f_n(m)

for all {0 \leq a < b \leq 1}, where {p} is the nonprincipal ultrafilter used to define the nonstandard universe. In particular, from the Arzela-Ascoli diagonalisation argument there is a subsequence {n_j} such that

\displaystyle  \int_a^b F(x)\ dx = \lim_{j \rightarrow \infty} \frac{1}{N_{n_j}} \sum_{a N_{n_j} \leq m \leq b N_{n_j}} f_n(m),

thus {F} is the asymptotic density function of the {f_n}. For instance, if {f_n} is the indicator function of a randomly chosen subset of {[N_n]}, then the asymptotic density function would equal {1/2} (almost everywhere, at least).

I’m continuing to look into understanding the ergodic theory of {o(N)} actions, as I believe this may allow one to apply ergodic theory methods to the “single-scale” or “non-asymptotic” setting (in which one averages only over scales comparable to a large parameter {N}, rather than the traditional asymptotic approach of letting the scale go to infinity). I’m planning some further posts in this direction, though this is still a work in progress.

The von Neumann ergodic theorem (the Hilbert space version of the mean ergodic theorem) asserts that if {U: H \rightarrow H} is a unitary operator on a Hilbert space {H}, and {v \in H} is a vector in that Hilbert space, then one has

\displaystyle  \lim_{N \rightarrow \infty} \frac{1}{N} \sum_{n=1}^N U^n v = \pi_{H^U} v

in the strong topology, where {H^U := \{ w \in H: Uw = w \}} is the {U}-invariant subspace of {H}, and {\pi_{H^U}} is the orthogonal projection to {H^U}. (See e.g. these previous lecture notes for a proof.) The same proof extends to more general amenable groups: if {G} is a countable amenable group acting on a Hilbert space {H} by unitary transformations {g: H \rightarrow H}, and {v \in H} is a vector in that Hilbert space, then one has

\displaystyle  \lim_{N \rightarrow \infty} \frac{1}{|\Phi_N|} \sum_{g \in \Phi_N} gv = \pi_{H^G} v \ \ \ \ \ (1)

for any Folner sequence {\Phi_N} of {G}, where {H^G := \{ w \in H: gw = w \hbox{ for all }g \in G \}} is the {G}-invariant subspace. Thus one can interpret {\pi_{H^G} v} as a certain average of elements of the orbit {Gv := \{ gv: g \in G \}} of {v}.

I recently discovered that there is a simple variant of this ergodic theorem that holds even when the group {G} is not amenable (or not discrete), using a more abstract notion of averaging:

Theorem 1 (Abstract ergodic theorem) Let {G} be an arbitrary group acting unitarily on a Hilbert space {H}, and let {v} be a vector in {H}. Then {\pi_{H^G} v} is the element in the closed convex hull of {Gv := \{ gv: g \in G \}} of minimal norm, and is also the unique element of {H^G} in this closed convex hull.

Proof: As the closed convex hull of {Gv} is closed, convex, and non-empty in a Hilbert space, it is a classical fact (see e.g. Proposition 1 of this previous post) that it has a unique element {F} of minimal norm. If {T_g F \neq F} for some {g}, then the midpoint of {T_g F} and {F} would be in the closed convex hull and be of smaller norm, a contradiction; thus {F} is {G}-invariant. To finish the first claim, it suffices to show that {v-F} is orthogonal to every element {h} of {H^G}. But if this were not the case for some such {h}, we would have {\langle T_g v - F, h \rangle = \langle v-F,h\rangle \neq 0} for all {g \in G}, and thus on taking convex hulls {\langle F-F,h\rangle = \langle f-F,f\rangle \neq 0}, a contradiction.

Finally, since {T_g v - F} is orthogonal to {H^G}, the same is true for {F'-F} for any {F'} in the closed convex hull of {Gv}, and this gives the second claim. \Box

This result is due to Alaoglu and Birkhoff. It implies the amenable ergodic theorem (1); indeed, given any {\epsilon>0}, Theorem 1 implies that there is a finite convex combination {v_\epsilon} of shifts {gv} of {v} which lies within {\epsilon} (in the {H} norm) to {\pi_{H^G} v}. By the triangle inequality, all the averages {\frac{1}{|\Phi_N|} \sum_{g \in \Phi_N} gv_\epsilon} also lie within {\epsilon} of {\pi_{H^G} v}, but by the Folner property this implies that the averages {\frac{1}{|\Phi_N|} \sum_{g \in \Phi_N} gv} are eventually within {2\epsilon} (say) of {\pi_{H^G} v}, giving the claim.

It turns out to be possible to use Theorem 1 as a substitute for the mean ergodic theorem in a number of contexts, thus removing the need for an amenability hypothesis. Here is a basic application:

Corollary 2 (Relative orthogonality) Let {G} be a group acting unitarily on a Hilbert space {H}, and let {V} be a {G}-invariant subspace of {H}. Then {V} and {H^G} are relatively orthogonal over their common subspace {V^G}, that is to say the restrictions of {V} and {H^G} to the orthogonal complement of {V^G} are orthogonal to each other.

Proof: By Theorem 1, we have {\pi_{H^G} v = \pi_{V^G} v} for all {v \in V}, and the claim follows. (Thanks to Gergely Harcos for this short argument.) \Box

Now we give a more advanced application of Theorem 1, to establish some “Mackey theory” over arbitrary groups {G}. Define a {G}-system {(X, {\mathcal X}, \mu, (T_g)_{g \in G})} to be a probability space {X = (X, {\mathcal X}, \mu)} together with a measure-preserving action {(T_g)_{g \in G}} of {G} on {X}; this gives an action of {G} on {L^2(X) = L^2(X,{\mathcal X},\mu)}, which by abuse of notation we also call {T_g}:

\displaystyle  T_g f := f \circ T_{g^{-1}}.

(In this post we follow the usual convention of defining the {L^p} spaces by quotienting out by almost everywhere equivalence.) We say that a {G}-system is ergodic if {L^2(X)^G} consists only of the constants.

(A technical point: the theory becomes slightly cleaner if we interpret our measure spaces abstractly (or “pointlessly“), removing the underlying space {X} and quotienting {{\mathcal X}} by the {\sigma}-ideal of null sets, and considering maps such as {T_g} only on this quotient {\sigma}-algebra (or on the associated von Neumann algebra {L^\infty(X)} or Hilbert space {L^2(X)}). However, we will stick with the more traditional setting of classical probability spaces here to keep the notation familiar, but with the understanding that many of the statements below should be understood modulo null sets.)

A factor {Y = (Y, {\mathcal Y}, \nu, (S_g)_{g \in G})} of a {G}-system {X = (X,{\mathcal X},\mu, (T_g)_{g \in G})} is another {G}-system together with a factor map {\pi: X \rightarrow Y} which commutes with the {G}-action (thus {T_g \pi = \pi S_g} for all {g \in G}) and respects the measure in the sense that {\mu(\pi^{-1}(E)) = \nu(E)} for all {E \in {\mathcal Y}}. For instance, the {G}-invariant factor {Z^0_G(X) := (X, {\mathcal X}^G, \mu\downharpoonright_{{\mathcal X}^G}, (T_g)_{g \in G})}, formed by restricting {X} to the invariant algebra {{\mathcal X}^G := \{ E \in {\mathcal X}: T_g E = E \hbox{ a.e. for all } g \in G \}}, is a factor of {X}. (This factor is the first factor in an important hierachy, the next element of which is the Kronecker factor {Z^1_G(X)}, but we will not discuss higher elements of this hierarchy further here.) If {Y} is a factor of {X}, we refer to {X} as an extension of {Y}.

From Corollary 2 we have

Corollary 3 (Relative independence) Let {X} be a {G}-system for a group {G}, and let {Y} be a factor of {X}. Then {Y} and {Z^0_G(X)} are relatively independent over their common factor {Z^0(Y)}, in the sense that the spaces {L^2(Y)} and {L^2(Z^0_G(X))} are relatively orthogonal over {L^2(Z^0_G(Y))} when all these spaces are embedded into {L^2(X)}.

This has a simple consequence regarding the product {X \times Y = (X \times Y, {\mathcal X} \times {\mathcal Y}, \mu \times \nu, (T_g \oplus S_g)_{g \in G})} of two {G}-systems {X = (X, {\mathcal X}, \mu, (T_g)_{g \in G})} and {Y = (Y, {\mathcal Y}, \nu, (S_g)_{g \in G})}, in the case when the {Y} action is trivial:

Lemma 4 If {X,Y} are two {G}-systems, with the action of {G} on {Y} trivial, then {Z^0_G(X \times Y)} is isomorphic to {Z^0_G(X) \times Y} in the obvious fashion.

This lemma is immediate for countable {G}, since for a {G}-invariant function {f}, one can ensure that {T_g f = f} holds simultaneously for all {g \in G} outside of a null set, but is a little trickier for uncountable {G}.

Proof: It is clear that {Z^0_G(X) \times Y} is a factor of {Z^0_G(X \times Y)}. To obtain the reverse inclusion, suppose that it fails, thus there is a non-zero {f \in L^2(Z^0_G(X \times Y))} which is orthogonal to {L^2(Z^0_G(X) \times Y)}. In particular, we have {fg} orthogonal to {L^2(Z^0_G(X))} for any {g \in L^\infty(Y)}. Since {fg} lies in {L^2(Z^0_G(X \times Y))}, we conclude from Corollary 3 (viewing {X} as a factor of {X \times Y}) that {fg} is also orthogonal to {L^2(X)}. Since {g} is an arbitrary element of {L^\infty(Y)}, we conclude that {f} is orthogonal to {L^2(X \times Y)} and in particular is orthogonal to itself, a contradiction. (Thanks to Gergely Harcos for this argument.) \Box

Now we discuss the notion of a group extension.

Definition 5 (Group extension) Let {G} be an arbitrary group, let {Y = (Y, {\mathcal Y}, \nu, (S_g)_{g \in G})} be a {G}-system, and let {K} be a compact metrisable group. A {K}-extension of {Y} is an extension {X = (X, {\mathcal X}, \mu, (T_g)_{g \in G})} whose underlying space is {X = Y \times K} (with {{\mathcal X}} the product of {{\mathcal Y}} and the Borel {\sigma}-algebra on {K}), the factor map is {\pi: (y,k) \mapsto y}, and the shift maps {T_g} are given by

\displaystyle  T_g ( y, k ) = (S_g y, \rho_g(y) k )

where for each {g \in G}, {\rho_g: Y \rightarrow K} is a measurable map (known as the cocycle associated to the {K}-extension {X}).

An important special case of a {K}-extension arises when the measure {\mu} is the product of {\nu} with the Haar measure {dk} on {K}. In this case, {X} also has a {K}-action {k': (y,k) \mapsto (y,k(k')^{-1})} that commutes with the {G}-action, making {X} a {G \times K}-system. More generally, {\mu} could be the product of {\nu} with the Haar measure {dh} of some closed subgroup {H} of {K}, with {\rho_g} taking values in {H}; then {X} is now a {G \times H} system. In this latter case we will call {X} {H}-uniform.

If {X} is a {K}-extension of {Y} and {U: Y \rightarrow K} is a measurable map, we can define the gauge transform {X_U} of {X} to be the {K}-extension of {Y} whose measure {\mu_U} is the pushforward of {\mu} under the map {(y,k) \mapsto (y, U(y) k)}, and whose cocycles {\rho_{g,U}: Y \rightarrow K} are given by the formula

\displaystyle  \rho_{g,U}(y) := U(gy) \rho_g(y) U(y)^{-1}.

It is easy to see that {X_U} is a {K}-extension that is isomorphic to {X} as a {K}-extension of {Y}; we will refer to {X_U} and {X} as equivalent systems, and {\rho_{g,U}} as cohomologous to {\rho_g}. We then have the following fundamental result of Mackey and of Zimmer:

Theorem 6 (Mackey-Zimmer theorem) Let {G} be an arbitrary group, let {Y} be an ergodic {G}-system, and let {K} be a compact metrisable group. Then every ergodic {K}-extension {X} of {Y} is equivalent to an {H}-uniform extension of {Y} for some closed subgroup {H} of {K}.

This theorem is usually stated for amenable groups {G}, but by using Theorem 1 (or more precisely, Corollary 3) the result is in fact also valid for arbitrary groups; we give the proof below the fold. (In the usual formulations of the theorem, {X} and {Y} are also required to be Lebesgue spaces, or at least standard Borel, but again with our abstract approach here, such hypotheses will be unnecessary.) Among other things, this theorem plays an important role in the Furstenberg-Zimmer structural theory of measure-preserving systems (as well as subsequent refinements of this theory by Host and Kra); see this previous blog post for some relevant discussion. One can obtain similar descriptions of non-ergodic extensions via the ergodic decomposition, but the result becomes more complicated to state, and we will not do so here.

Read the rest of this entry »

Given two unit vectors {v,w} in a real inner product space, one can define the correlation between these vectors to be their inner product {\langle v, w \rangle}, or in more geometric terms, the cosine of the angle {\angle(v,w)} subtended by {v} and {w}. By the Cauchy-Schwarz inequality, this is a quantity between {-1} and {+1}, with the extreme positive correlation {+1} occurring when {v,w} are identical, the extreme negative correlation {-1} occurring when {v,w} are diametrically opposite, and the zero correlation {0} occurring when {v,w} are orthogonal. This notion is closely related to the notion of correlation between two non-constant square-integrable real-valued random variables {X,Y}, which is the same as the correlation between two unit vectors {v,w} lying in the Hilbert space {L^2(\Omega)} of square-integrable random variables, with {v} being the normalisation of {X} defined by subtracting off the mean {\mathbf{E} X} and then dividing by the standard deviation of {X}, and similarly for {w} and {Y}.

One can also define correlation for complex (Hermitian) inner product spaces by taking the real part {\hbox{Re} \langle , \rangle} of the complex inner product to recover a real inner product.

While reading the (highly recommended) recent popular maths book “How not to be wrong“, by my friend and co-author Jordan Ellenberg, I came across the (important) point that correlation is not necessarily transitive: if {X} correlates with {Y}, and {Y} correlates with {Z}, then this does not imply that {X} correlates with {Z}. A simple geometric example is provided by the three unit vectors

\displaystyle  u := (1,0); v := (\frac{1}{\sqrt{2}}, \frac{1}{\sqrt{2}}); w := (0,1)

in the Euclidean plane {{\bf R}^2}: {u} and {v} have a positive correlation of {\frac{1}{\sqrt{2}}}, as does {v} and {w}, but {u} and {w} are not correlated with each other. Or: for a typical undergraduate course, it is generally true that good exam scores are correlated with a deep understanding of the course material, and memorising from flash cards are correlated with good exam scores, but this does not imply that memorising flash cards is correlated with deep understanding of the course material.

However, there are at least two situations in which some partial version of transitivity of correlation can be recovered. The first is in the “99%” regime in which the correlations are very close to {1}: if {u,v,w} are unit vectors such that {u} is very highly correlated with {v}, and {v} is very highly correlated with {w}, then this does imply that {u} is very highly correlated with {w}. Indeed, from the identity

\displaystyle  \| u-v \| = 2^{1/2} (1 - \langle u,v\rangle)^{1/2}

(and similarly for {v-w} and {u-w}) and the triangle inequality

\displaystyle  \|u-w\| \leq \|u-v\| + \|v-w\|,

we see that

\displaystyle  (1 - \langle u,w \rangle)^{1/2} \leq (1 - \langle u,v\rangle)^{1/2} + (1 - \langle v,w\rangle)^{1/2}. \ \ \ \ \ (1)

Thus, for instance, if {\langle u, v \rangle \geq 1-\epsilon} and {\langle v,w \rangle \geq 1-\epsilon}, then {\langle u,w \rangle \geq 1-4\epsilon}. This is of course closely related to (though slightly weaker than) the triangle inequality for angles:

\displaystyle  \angle(u,w) \leq \angle(u,v) + \angle(v,w).

Remark 1 (Thanks to Andrew Granville for conversations leading to this observation.) The inequality (1) also holds for sub-unit vectors, i.e. vectors {u,v,w} with {\|u\|, \|v\|, \|w\| \leq 1}. This comes by extending {u,v,w} in directions orthogonal to all three original vectors and to each other in order to make them unit vectors, enlarging the ambient Hilbert space {H} if necessary. More concretely, one can apply (1) to the unit vectors

\displaystyle  (u, \sqrt{1-\|u\|^2}, 0, 0), (v, 0, \sqrt{1-\|v\|^2}, 0), (w, 0, 0, \sqrt{1-\|w\|^2})

in {H \times {\bf R}^3}.

But even in the “{1\%}” regime in which correlations are very weak, there is still a version of transitivity of correlation, known as the van der Corput lemma, which basically asserts that if a unit vector {v} is correlated with many unit vectors {u_1,\dots,u_n}, then many of the pairs {u_i,u_j} will then be correlated with each other. Indeed, from the Cauchy-Schwarz inequality

\displaystyle  |\langle v, \sum_{i=1}^n u_i \rangle|^2 \leq \|v\|^2 \| \sum_{i=1}^n u_i \|^2

we see that

\displaystyle  (\sum_{i=1}^n \langle v, u_i \rangle)^2 \leq \sum_{1 \leq i,j \leq n} \langle u_i, u_j \rangle. \ \ \ \ \ (2)

Thus, for instance, if {\langle v, u_i \rangle \geq \epsilon} for at least {\epsilon n} values of {i=1,\dots,n}, then {\sum_{1 \leq i,j \leq n} \langle u_i, u_j \rangle} must be at least {\epsilon^3 n^2}, which implies that {\langle u_i, u_j \rangle \geq \epsilon^3/2} for at least {\epsilon^3 n^2/2} pairs {(i,j)}. Or as another example: if a random variable {X} exhibits at least {1\%} positive correlation with {n} other random variables {Y_1,\dots,Y_n}, then if {n > 10,000}, at least two distinct {Y_i,Y_j} must have positive correlation with each other (although this argument does not tell you which pair {Y_i,Y_j} are so correlated). Thus one can view this inequality as a sort of `pigeonhole principle” for correlation.

A similar argument (multiplying each {u_i} by an appropriate sign {\pm 1}) shows the related van der Corput inequality

\displaystyle  (\sum_{i=1}^n |\langle v, u_i \rangle|)^2 \leq \sum_{1 \leq i,j \leq n} |\langle u_i, u_j \rangle|, \ \ \ \ \ (3)

and this inequality is also true for complex inner product spaces. (Also, the {u_i} do not need to be unit vectors for this inequality to hold.)

Geometrically, the picture is this: if {v} positively correlates with all of the {u_1,\dots,u_n}, then the {u_1,\dots,u_n} are all squashed into a somewhat narrow cone centred at {v}. The cone is still wide enough to allow a few pairs {u_i, u_j} to be orthogonal (or even negatively correlated) with each other, but (when {n} is large enough) it is not wide enough to allow all of the {u_i,u_j} to be so widely separated. Remarkably, the bound here does not depend on the dimension of the ambient inner product space; while increasing the number of dimensions should in principle add more “room” to the cone, this effect is counteracted by the fact that in high dimensions, almost all pairs of vectors are close to orthogonal, and the exceptional pairs that are even weakly correlated to each other become exponentially rare. (See this previous blog post for some related discussion; in particular, Lemma 2 from that post is closely related to the van der Corput inequality presented here.)

A particularly common special case of the van der Corput inequality arises when {v} is a unit vector fixed by some unitary operator {T}, and the {u_i} are shifts {u_i = T^i u} of a single unit vector {u}. In this case, the inner products {\langle v, u_i \rangle} are all equal, and we arrive at the useful van der Corput inequality

\displaystyle  |\langle v, u \rangle|^2 \leq \frac{1}{n^2} \sum_{1 \leq i,j \leq n} |\langle T^i u, T^j u \rangle|. \ \ \ \ \ (4)

(In fact, one can even remove the absolute values from the right-hand side, by using (2) instead of (4).) Thus, to show that {v} has negligible correlation with {u}, it suffices to show that the shifts of {u} have negligible correlation with each other.

Here is a basic application of the van der Corput inequality:

Proposition 1 (Weyl equidistribution estimate) Let {P: {\bf Z} \rightarrow {\bf R}/{\bf Z}} be a polynomial with at least one non-constant coefficient irrational. Then one has

\displaystyle  \lim_{N \rightarrow \infty} \frac{1}{N} \sum_{n=1}^N e( P(n) ) = 0,

where {e(x) := e^{2\pi i x}}.

Note that this assertion implies the more general assertion

\displaystyle  \lim_{N \rightarrow \infty} \frac{1}{N} \sum_{n=1}^N e( kP(n) ) = 0

for any non-zero integer {k} (simply by replacing {P} by {kP}), which by the Weyl equidistribution criterion is equivalent to the sequence {P(1), P(2),\dots} being asymptotically equidistributed in {{\bf R}/{\bf Z}}.

Proof: We induct on the degree {d} of the polynomial {P}, which must be at least one. If {d} is equal to one, the claim is easily established from the geometric series formula, so suppose that {d>1} and that the claim has already been proven for {d-1}. If the top coefficient {a_d} of {P(n) = a_d n^d + \dots + a_0} is rational, say {a_d = \frac{p}{q}}, then by partitioning the natural numbers into residue classes modulo {q}, we see that the claim follows from the induction hypothesis; so we may assume that the top coefficient {a_d} is irrational.

In order to use the van der Corput inequality as stated above (i.e. in the formalism of inner product spaces) we will need a non-principal ultrafilter {p} (see e.g this previous blog post for basic theory of ultrafilters); we leave it as an exercise to the reader to figure out how to present the argument below without the use of ultrafilters (or similar devices, such as Banach limits). The ultrafilter {p} defines an inner product {\langle, \rangle_p} on bounded complex sequences {z = (z_1,z_2,z_3,\dots)} by setting

\displaystyle  \langle z, w \rangle_p := \hbox{st} \lim_{N \rightarrow p} \frac{1}{N} \sum_{n=1}^N z_n \overline{w_n}.

Strictly speaking, this inner product is only positive semi-definite rather than positive definite, but one can quotient out by the null vectors to obtain a positive-definite inner product. To establish the claim, it will suffice to show that

\displaystyle  \langle 1, e(P) \rangle_p = 0

for every non-principal ultrafilter {p}.

Note that the space of bounded sequences (modulo null vectors) admits a shift {T}, defined by

\displaystyle  T (z_1,z_2,\dots) := (z_2,z_3,\dots).

This shift becomes unitary once we quotient out by null vectors, and the constant sequence {1} is clearly a unit vector that is invariant with respect to the shift. So by the van der Corput inequality, we have

\displaystyle  |\langle 1, e(P) \rangle_p| \leq \frac{1}{n^2} \sum_{1 \leq i,j \leq n} |\langle T^i e(P), T^j e(P) \rangle_p|

for any {n \geq 1}. But we may rewrite {\langle T^i e(P), T^j e(P) \rangle = \langle 1, e(T^j P - T^i P) \rangle_p}. Then observe that if {i \neq j}, {T^j P - T^i P} is a polynomial of degree {d-1} whose {d-1} coefficient is irrational, so by induction hypothesis we have {\langle T^i e(P), T^j e(P) \rangle_p = 0} for {i \neq j}. For {i=j} we of course have {\langle T^i e(P), T^j e(P) \rangle_p = 1}, and so

\displaystyle  |\langle 1, e(P) \rangle_p| \leq \frac{1}{n^2} \times n

for any {n}. Letting {n \rightarrow \infty}, we obtain the claim. \Box

Let {V} be a quasiprojective variety defined over a finite field {{\bf F}_q}, thus for instance {V} could be an affine variety

\displaystyle  V = \{ x \in {\bf A}^d: P_1(x) = \dots = P_m(x) = 0\} \ \ \ \ \ (1)

where {{\bf A}^d} is {d}-dimensional affine space and {P_1,\dots,P_m: {\bf A}^d \rightarrow {\bf A}} are a finite collection of polynomials with coefficients in {{\bf F}_q}. Then one can define the set {V[{\bf F}_q]} of {{\bf F}_q}-rational points, and more generally the set {V[{\bf F}_{q^n}]} of {{\bf F}_{q^n}}-rational points for any {n \geq 1}, since {{\bf F}_{q^n}} can be viewed as a field extension of {{\bf F}_q}. Thus for instance in the affine case (1) we have

\displaystyle  V[{\bf F}_{q^n}] := \{ x \in {\bf F}_{q^n}^d: P_1(x) = \dots = P_m(x) = 0\}.

The Weil conjectures are concerned with understanding the number

\displaystyle  S_n := |V[{\bf F}_{q^n}]| \ \ \ \ \ (2)

of {{\bf F}_{q^n}}-rational points over a variety {V}. The first of these conjectures was proven by Dwork, and can be phrased as follows.

Theorem 1 (Rationality of the zeta function) Let {V} be a quasiprojective variety defined over a finite field {{\bf F}_q}, and let {S_n} be given by (2). Then there exist a finite number of algebraic integers {\alpha_1,\dots,\alpha_k, \beta_1,\dots,\beta_{k'} \in O_{\overline{{\bf Q}}}} (known as characteristic values of {V}), such that

\displaystyle  S_n = \alpha_1^n + \dots + \alpha_k^n - \beta_1^n - \dots - \beta_{k'}^n

for all {n \geq 1}.

After cancelling, we may of course assume that {\alpha_i \neq \beta_j} for any {i=1,\dots,k} and {j=1,\dots,k'}, and then it is easy to see (as we will see below) that the {\alpha_1,\dots,\alpha_k,\beta_1,\dots,\beta_{k'}} become uniquely determined up to permutations of the {\alpha_1,\dots,\alpha_k} and {\beta_1,\dots,\beta_{k'}}. These values are known as the characteristic values of {V}. Since {S_n} is a rational integer (i.e. an element of {{\bf Z}}) rather than merely an algebraic integer (i.e. an element of the ring of integers {O_{\overline{{\bf Q}}}} of the algebraic closure {\overline{{\bf Q}}} of {{\bf Q}}), we conclude from the above-mentioned uniqueness that the set of characteristic values are invariant with respect to the Galois group {Gal(\overline{{\bf Q}} / {\bf Q} )}. To emphasise this Galois invariance, we will not fix a specific embedding {\iota_\infty: \overline{{\bf Q}} \rightarrow {\bf C}} of the algebraic numbers into the complex field {{\bf C} = {\bf C}_\infty}, but work with all such embeddings simultaneously. (Thus, for instance, {\overline{{\bf Q}}} contains three cube roots of {2}, but which of these is assigned to the complex numbers {2^{1/3}}, {e^{2\pi i/3} 2^{1/3}}, {e^{4\pi i/3} 2^{1/3}} will depend on the choice of embedding {\iota_\infty}.)

An equivalent way of phrasing Dwork’s theorem is that the ({T}-form of the) zeta function

\displaystyle \zeta_V(T) := \exp( \sum_{n=1}^\infty \frac{S_n}{n} T^n )

associated to {V} (which is well defined as a formal power series in {T}, at least) is equal to a rational function of {T} (with the {\alpha_1,\dots,\alpha_k} and {\beta_1,\dots,\beta_{k'}} being the poles and zeroes of {\zeta_V} respectively). Here, we use the formal exponential

\displaystyle  \exp(X) := 1 + X + \frac{X^2}{2!} + \frac{X^3}{3!} + \dots.

Equivalently, the ({s}-form of the) zeta-function {s \mapsto \zeta_V(q^{-s})} is a meromorphic function on the complex numbers {{\bf C}} which is also periodic with period {2\pi i/\log q}, and which has only finitely many poles and zeroes up to this periodicity.

Dwork’s argument relies primarily on {p}-adic analysis – an analogue of complex analysis, but over an algebraically complete (and metrically complete) extension {{\bf C}_p} of the {p}-adic field {{\bf Q}_p}, rather than over the Archimedean complex numbers {{\bf C}}. The argument is quite effective, and in particular gives explicit upper bounds for the number {k+k'} of characteristic values in terms of the complexity of the variety {V}; for instance, in the affine case (1) with {V} of degree {D}, Bombieri used Dwork’s methods (in combination with Deligne’s theorem below) to obtain the bound {k+k' \leq (4D+9)^{2d+1}}, and a subsequent paper of Hooley established the slightly weaker bound {k+k' \leq (11D+11)^{d+m+2}} purely from Dwork’s methods (a similar bound had also been pointed out in unpublished work of Dwork). In particular, one has bounds that are uniform in the field {{\bf F}_q}, which is an important fact for many analytic number theory applications.

These {p}-adic arguments stand in contrast with Deligne’s resolution of the last (and deepest) of the Weil conjectures:

Theorem 2 (Riemann hypothesis) Let {V} be a quasiprojective variety defined over a finite field {{\bf F}_q}, and let {\lambda \in \overline{{\bf Q}}} be a characteristic value of {V}. Then there exists a natural number {w} such that {|\iota_\infty(\lambda)|_\infty = q^{w/2}} for every embedding {\iota_\infty: \overline{{\bf Q}} \rightarrow {\bf C}}, where {| |_\infty} denotes the usual absolute value on the complex numbers {{\bf C} = {\bf C}_\infty}. (Informally: {\lambda} and all of its Galois conjugates have complex magnitude {q^{w/2}}.)

To put it another way that closely resembles the classical Riemann hypothesis, all the zeroes and poles of the {s}-form {s \mapsto \zeta_V(q^{-s})} lie on the critical lines {\{ s \in {\bf C}: \hbox{Re}(s) = \frac{w}{2} \}} for {w=0,1,2,\dots}. (See this previous blog post for further comparison of various instantiations of the Riemann hypothesis.) Whereas Dwork uses {p}-adic analysis, Deligne uses the essentially orthogonal technique of ell-adic cohomology to establish his theorem. However, ell-adic methods can be used (via the Grothendieck-Lefschetz trace formula) to establish rationality, and conversely, in this paper of Kedlaya p-adic methods are used to establish the Riemann hypothesis. As pointed out by Kedlaya, the ell-adic methods are tied to the intrinsic geometry of {V} (such as the structure of sheaves and covers over {V}), while the {p}-adic methods are more tied to the extrinsic geometry of {V} (how {V} sits inside its ambient affine or projective space).

In this post, I would like to record my notes on Dwork’s proof of Theorem 1, drawing heavily on the expositions of Serre, Hooley, Koblitz, and others.

The basic strategy is to control the rational integers {S_n} both in an “Archimedean” sense (embedding the rational integers inside the complex numbers {{\bf C}_\infty} with the usual norm {||_\infty}) as well as in the “{p}-adic” sense, with {p} the characteristic of {{\bf F}_q} (embedding the integers now in the “complexification” {{\bf C}_p} of the {p}-adic numbers {{\bf Q}_p}, which is equipped with a norm {||_p} that we will recall later). (This is in contrast to the methods of ell-adic cohomology, in which one primarily works over an {\ell}-adic field {{\bf Q}_\ell} with {\ell \neq p,\infty}.) The Archimedean control is trivial:

Proposition 3 (Archimedean control of {S_n}) With {S_n} as above, and any embedding {\iota_\infty: \overline{{\bf Q}} \rightarrow {\bf C}}, we have

\displaystyle  |\iota_\infty(S_n)|_\infty \leq C q^{A n}

for all {n} and some {C, A >0} independent of {n}.

Proof: Since {S_n} is a rational integer, {|\iota_\infty(S_n)|_\infty} is just {|S_n|_\infty}. By decomposing {V} into affine pieces, we may assume that {V} is of the affine form (1), then we trivially have {|S_n|_\infty \leq q^{nd}}, and the claim follows. \Box

Another way of thinking about this Archimedean control is that it guarantees that the zeta function {T \mapsto \zeta_V(T)} can be defined holomorphically on the open disk in {{\bf C}_\infty} of radius {q^{-A}} centred at the origin.

The {p}-adic control is significantly more difficult, and is the main component of Dwork’s argument:

Proposition 4 ({p}-adic control of {S_n}) With {S_n} as above, and using an embedding {\iota_p: \overline{{\bf Q}} \rightarrow {\bf C}_p} (defined later) with {p} the characteristic of {{\bf F}_q}, we can find for any real {A > 0} a finite number of elements {\alpha_1,\dots,\alpha_k,\beta_1,\dots,\beta_{k'} \in {\bf C}_p} such that

\displaystyle  |\iota_p(S_n) - (\alpha_1^n + \dots + \alpha_k^n - \beta_1^n - \dots - \beta_{k'}^n)|_p \leq q^{-An}

for all {n}.

Another way of thinking about this {p}-adic control is that it guarantees that the zeta function {T \mapsto \zeta_V(T)} can be defined meromorphically on the entire {p}-adic complex field {{\bf C}_p}.

Proposition 4 is ostensibly much weaker than Theorem 1 because of (a) the error term of {p}-adic magnitude at most {Cq^{-An}}; (b) the fact that the number {k+k'} of potential characteristic values here may go to infinity as {A \rightarrow \infty}; and (c) the potential characteristic values {\alpha_1,\dots,\alpha_k,\beta_1,\dots,\beta_{k'}} only exist inside the complexified {p}-adics {{\bf C}_p}, rather than in the algebraic integers {O_{\overline{{\bf Q}}}}. However, it turns out that by combining {p}-adic control on {S_n} in Proposition 4 with the trivial control on {S_n} in Proposition 3, one can obtain Theorem 1 by an elementary argument that does not use any further properties of {S_n} (other than the obvious fact that the {S_n} are rational integers), with the {A} in Proposition 4 chosen to exceed the {A} in Proposition 3. We give this argument (essentially due to Borel) below the fold.

The proof of Proposition 4 can be split into two pieces. The first piece, which can be viewed as the number-theoretic component of the proof, uses external descriptions of {V} such as (1) to obtain the following decomposition of {S_n}:

Proposition 5 (Decomposition of {S_n}) With {\iota_p} and {S_n} as above, we can decompose {\iota_p(S_n)} as a finite linear combination (over the integers) of sequences {S'_n \in {\bf C}_p}, such that for each such sequence {n \mapsto S'_n}, the zeta functions

\displaystyle  \zeta'(T) := \exp( \sum_{n=1}^\infty \frac{S'_n}{n} T^n ) = \sum_{n=0}^\infty c_n T^n

are entire in {{\bf C}_p}, by which we mean that

\displaystyle  |c_n|_p^{1/n} \rightarrow 0

as {n \rightarrow \infty}.

This proposition will ultimately be a consequence of the properties of the Teichmuller lifting {\tau: \overline{{\bf F}_p}^\times \rightarrow {\bf C}_p^\times}.

The second piece, which can be viewed as the “{p}-adic complex analytic” component of the proof, relates the {p}-adic entire nature of a zeta function with control on the associated sequence {S'_n}, and can be interpreted (after some manipulation) as a {p}-adic version of the Weierstrass preparation theorem:

Proposition 6 ({p}-adic Weierstrass preparation theorem) Let {S'_n} be a sequence in {{\bf C}_p}, such that the zeta function

\displaystyle  \zeta'(T) := \exp( \sum_{n=1}^\infty \frac{S'_n}{n} T^n )

is entire in {{\bf C}_p}. Then for any real {A > 0}, there exist a finite number of elements {\beta_1,\dots,\beta_{k'} \in {\bf C}_p} such that

\displaystyle  |\iota_p(S'_n) + \beta_1^n + \dots + \beta_{k'}^n|_p \leq q^{-An}

for all {n} and some {C>0}.

Clearly, the combination of Proposition 5 and Proposition 6 (and the non-Archimedean nature of the {||_p} norm) imply Proposition 4.

Read the rest of this entry »

This is a blog version of a talk I recently gave at the IPAM workshop on “The Kakeya Problem, Restriction Problem, and Sum-product Theory”.

Note: the discussion here will be highly non-rigorous in nature, being extremely loose in particular with asymptotic notation and with the notion of dimension. Caveat emptor.

One of the most infamous unsolved problems at the intersection of geometric measure theory, incidence combinatorics, and real-variable harmonic analysis is the Kakeya set conjecture. We will focus on the following three-dimensional case of the conjecture, stated informally as follows:

Conjecture 1 (Kakeya conjecture) Let {E} be a subset of {{\bf R}^3} that contains a unit line segment in every direction. Then {\hbox{dim}(E) = 3}.

This conjecture is not precisely formulated here, because we have not specified exactly what type of set {E} is (e.g. measurable, Borel, compact, etc.) and what notion of dimension we are using. We will deliberately ignore these technical details in this post. It is slightly more convenient for us here to work with lines instead of unit line segments, so we work with the following slight variant of the conjecture (which is essentially equivalent):

Conjecture 2 (Kakeya conjecture, again) Let {{\cal L}} be a family of lines in {{\bf R}^3} that meet {B(0,1)} and contain a line in each direction. Let {E} be the union of the restriction {\ell \cap B(0,2)} to {B(0,2)} of every line {\ell} in {{\cal L}}. Then {\hbox{dim}(E) = 3}.

As the space of all directions in {{\bf R}^3} is two-dimensional, we thus see that {{\cal L}} is an (at least) two-dimensional subset of the four-dimensional space of lines in {{\bf R}^3} (actually, it lies in a compact subset of this space, since we have constrained the lines to meet {B(0,1)}). One could then ask if this is the only property of {{\cal L}} that is needed to establish the Kakeya conjecture, that is to say if any subset of {B(0,2)} which contains a two-dimensional family of lines (restricted to {B(0,2)}, and meeting {B(0,1)}) is necessarily three-dimensional. Here we have an easy counterexample, namely a plane in {B(0,2)} (passing through the origin), which contains a two-dimensional collection of lines. However, we can exclude this case by adding an additional axiom, leading to what one might call a “strong” Kakeya conjecture:

Conjecture 3 (Strong Kakeya conjecture) Let {{\cal L}} be a two-dimensional family of lines in {{\bf R}^3} that meet {B(0,1)}, and assume the Wolff axiom that no (affine) plane contains more than a one-dimensional family of lines in {{\cal L}}. Let {E} be the union of the restriction {\ell \cap B(0,2)} of every line {\ell} in {{\cal L}}. Then {\hbox{dim}(E) = 3}.

Actually, to make things work out we need a more quantitative version of the Wolff axiom in which we constrain the metric entropy (and not just dimension) of lines that lie close to a plane, rather than exactly on the plane. However, for the informal discussion here we will ignore these technical details. Families of lines that lie in different directions will obey the Wolff axiom, but the converse is not true in general.

In 1995, Wolff established the important lower bound {\hbox{dim}(E) \geq 5/2} (for various notions of dimension, e.g. Hausdorff dimension) for sets {E} in Conjecture 3 (and hence also for the other forms of the Kakeya problem). However, there is a key obstruction to going beyond the {5/2} barrier, coming from the possible existence of half-dimensional (approximate) subfields of the reals {{\bf R}}. To explain this problem, it easiest to first discuss the complex version of the strong Kakeya conjecture, in which all relevant (real) dimensions are doubled:

Conjecture 4 (Strong Kakeya conjecture over {{\bf C}}) Let {{\cal L}} be a four (real) dimensional family of complex lines in {{\bf C}^3} that meet the unit ball {B(0,1)} in {{\bf C}^3}, and assume the Wolff axiom that no four (real) dimensional (affine) subspace contains more than a two (real) dimensional family of complex lines in {{\cal L}}. Let {E} be the union of the restriction {\ell \cap B(0,2)} of every complex line {\ell} in {{\cal L}}. Then {E} has real dimension {6}.

The argument of Wolff can be adapted to the complex case to show that all sets {E} occuring in Conjecture 4 have real dimension at least {5}. Unfortunately, this is sharp, due to the following fundamental counterexample:

Proposition 5 (Heisenberg group counterexample) Let {H \subset {\bf C}^3} be the Heisenberg group

\displaystyle  H = \{ (z_1,z_2,z_3) \in {\bf C}^3: \hbox{Im}(z_1) = \hbox{Im}(z_2 \overline{z_3}) \}

and let {{\cal L}} be the family of complex lines

\displaystyle  \ell_{s,t,\alpha} := \{ (\overline{\alpha} z + t, z, sz + \alpha): z \in {\bf C} \}

with {s,t \in {\bf R}} and {\alpha \in {\bf C}}. Then {H} is a five (real) dimensional subset of {{\bf C}^3} that contains every line in the four (real) dimensional set {{\cal L}}; however each four real dimensional (affine) subspace contains at most a two (real) dimensional set of lines in {{\cal L}}. In particular, the strong Kakeya conjecture over the complex numbers is false.

This proposition is proven by a routine computation, which we omit here. The group structure on {H} is given by the group law

\displaystyle  (z_1,z_2,z_3) \cdot (w_1,w_2,w_3) = (z_1 + w_1 + z_2 \overline{w_3} - z_3 \overline{w_2}, z_2 +w_2, z_3+w_3),

giving {E} the structure of a {2}-step simply-connected nilpotent Lie group, isomorphic to the usual Heisenberg group over {{\bf R}^2}. Note that while the Heisenberg group is a counterexample to the complex strong Kakeya conjecture, it is not a counterexample to the complex form of the original Kakeya conjecture, because the complex lines {{\cal L}} in the Heisenberg counterexample do not point in distinct directions, but instead only point in a three (real) dimensional subset of the four (real) dimensional space of available directions for complex lines. For instance, one has the one real-dimensional family of parallel lines

\displaystyle  \ell_{0,t,0} = \{ (t, z, 0): z \in {\bf C}\}

with {t \in {\bf R}}; multiplying this family of lines on the right by a group element in {H} gives other families of parallel lines, which in fact sweep out all of {{\cal L}}.

The Heisenberg counterexample ultimately arises from the “half-dimensional” (and hence degree two) subfield {{\bf R}} of {{\bf C}}, which induces an involution {z \mapsto \overline{z}} which can then be used to define the Heisenberg group {H} through the formula

\displaystyle  H = \{ (z_1,z_2,z_3) \in {\bf C}^3: z_1 - \overline{z_1} = z_2 \overline{z_3} - z_3 \overline{z_2} \}.

Analogous Heisenberg counterexamples can also be constructed if one works over finite fields {{\bf F}_{q^2}} that contain a “half-dimensional” subfield {{\bf F}_q}; we leave the details to the interested reader. Morally speaking, if {{\bf R}} in turn contained a subfield of dimension {1/2} (or even a subring or “approximate subring”), then one ought to be able to use this field to generate a counterexample to the strong Kakeya conjecture over the reals. Fortunately, such subfields do not exist; this was a conjecture of Erdos and Volkmann that was proven by Edgar and Miller, and more quantitatively by Bourgain (answering a question of Nets Katz and myself). However, this fact is not entirely trivial to prove, being a key example of the sum-product phenomenon.

We thus see that to go beyond the {5/2} dimension bound of Wolff for the 3D Kakeya problem over the reals, one must do at least one of two things:

  • (a) Exploit the distinct directions of the lines in {{\mathcal L}} in a way that goes beyond the Wolff axiom; or
  • (b) Exploit the fact that {{\bf R}} does not contain half-dimensional subfields (or more generally, intermediate-dimensional approximate subrings).

(The situation is more complicated in higher dimensions, as there are more obstructions than the Heisenberg group; for instance, in four dimensions quadric surfaces are an important obstruction, as discussed in this paper of mine.)

Various partial or complete results on the Kakeya problem over various fields have been obtained through route (a) or route (b). For instance, in 2000, Nets Katz, Izabella Laba and myself used route (a) to improve Wolff’s lower bound of {5/2} for Kakeya sets very slightly to {5/2+10^{-10}} (for a weak notion of dimension, namely upper Minkowski dimension). In 2004, Bourgain, Katz, and myself established a sum-product estimate which (among other things) ruled out approximate intermediate-dimensional subrings of {{\bf F}_p}, and then pursued route (b) to obtain a corresponding improvement {5/2+\epsilon} to the Kakeya conjecture over finite fields of prime order. The analogous (discretised) sum-product estimate over the reals was established by Bourgain in 2003, which in principle would allow one to extend the result of Katz, Laba and myself to the strong Kakeya setting, but this has not been carried out in the literature. Finally, in 2009, Dvir used route (a) and introduced the polynomial method (as discussed previously here) to completely settle the Kakeya conjecture in finite fields.

Below the fold, I present a heuristic argument of Nets Katz and myself, which in principle would use route (b) to establish the full (strong) Kakeya conjecture. In broad terms, the strategy is as follows:

  1. Assume that the (strong) Kakeya conjecture fails, so that there are sets {E} of the form in Conjecture 3 of dimension {3-\sigma} for some {\sigma>0}. Assume that {E} is “optimal”, in the sense that {\sigma} is as large as possible.
  2. Use the optimality of {E} (and suitable non-isotropic rescalings) to establish strong forms of standard structural properties expected of such sets {E}, namely “stickiness”, “planiness”, “local graininess” and “global graininess” (we will roughly describe these properties below). Heuristically, these properties are constraining {E} to “behave like” a putative Heisenberg group counterexample.
  3. By playing all these structural properties off of each other, show that {E} can be parameterised locally by a one-dimensional set which generates a counterexample to Bourgain’s sum-product theorem. This contradiction establishes the Kakeya conjecture.

Nets and I have had an informal version of argument for many years, but were never able to make a satisfactory theorem (or even a partial Kakeya result) out of it, because we could not rigorously establish anywhere near enough of the necessary structural properties (stickiness, planiness, etc.) on the optimal set {E} for a large number of reasons (one of which being that we did not have a good notion of dimension that did everything that we wished to demand of it). However, there is beginning to be movement in these directions (e.g. in this recent result of Guth using the polynomial method obtaining a weak version of local graininess on certain Kakeya sets). In view of this (and given that neither Nets or I have been actively working in this direction for some time now, due to many other projects), we’ve decided to distribute these ideas more widely than before, and in particular on this blog.

Read the rest of this entry »

Roth’s theorem on arithmetic progressions asserts that every subset of the integers {{\bf Z}} of positive upper density contains infinitely many arithmetic progressions of length three. There are many versions and variants of this theorem. Here is one of them:

Theorem 1 (Roth’s theorem) Let {G = (G,+)} be a compact abelian group, with Haar probability measure {\mu}, which is {2}-divisible (i.e. the map {x \mapsto 2x} is surjective) and let {A} be a measurable subset of {G} with {\mu(A) \geq \alpha} for some {0 < \alpha < 1}. Then we have

\displaystyle  \int_G \int_G 1_A(x) 1_A(x+r) 1_A(x+2r)\ d\mu(x) d\mu(r) \gg_\alpha 1,

where {X \gg_\alpha Y} denotes the bound {X \geq c_\alpha Y} for some {c_\alpha > 0} depending only on {\alpha}.

This theorem is usually formulated in the case that {G} is a finite abelian group of odd order (in which case the result is essentially due to Meshulam) or more specifically a cyclic group {G = {\bf Z}/N{\bf Z}} of odd order (in which case it is essentially due to Varnavides), but is also valid for the more general setting of {2}-divisible compact abelian groups, as we shall shortly see. One can be more precise about the dependence of the implied constant {c_\alpha} on {\alpha}, but to keep the exposition simple we will work at the qualitative level here, without trying at all to get good quantitative bounds. The theorem is also true without the {2}-divisibility hypothesis, but the proof we will discuss runs into some technical issues due to the degeneracy of the {2r} shift in that case.

We can deduce Theorem 1 from the following more general Khintchine-type statement. Let {\hat G} denote the Pontryagin dual of a compact abelian group {G}, that is to say the set of all continuous homomorphisms {\xi: x \mapsto \xi \cdot x} from {G} to the (additive) unit circle {{\bf R}/{\bf Z}}. Thus {\hat G} is a discrete abelian group, and functions {f \in L^2(G)} have a Fourier transform {\hat f \in \ell^2(\hat G)} defined by

\displaystyle  \hat f(\xi) := \int_G f(x) e^{-2\pi i \xi \cdot x}\ d\mu(x).

If {G} is {2}-divisible, then {\hat G} is {2}-torsion-free in the sense that the map {\xi \mapsto 2 \xi} is injective. For any finite set {S \subset \hat G} and any radius {\rho>0}, define the Bohr set

\displaystyle  B(S,\rho) := \{ x \in G: \sup_{\xi \in S} \| \xi \cdot x \|_{{\bf R}/{\bf Z}} < \rho \}

where {\|\theta\|_{{\bf R}/{\bf Z}}} denotes the distance of {\theta} to the nearest integer. We refer to the cardinality {|S|} of {S} as the rank of the Bohr set. We record a simple volume bound on Bohr sets:

Lemma 2 (Volume packing bound) Let {G} be a compact abelian group with Haar probability measure {\mu}. For any Bohr set {B(S,\rho)}, we have

\displaystyle  \mu( B( S, \rho ) ) \gg_{|S|, \rho} 1.

Proof: We can cover the torus {({\bf R}/{\bf Z})^S} by {O_{|S|,\rho}(1)} translates {\theta+Q} of the cube {Q := \{ (\theta_\xi)_{\xi \in S} \in ({\bf R}/{\bf Z})^S: \sup_{\xi \in S} \|\theta_\xi\|_{{\bf R}/{\bf Z}} < \rho/2 \}}. Then the sets {\{ x \in G: (\xi \cdot x)_{\xi \in S} \in \theta + Q \}} form an cover of {G}. But all of these sets lie in a translate of {B(S,\rho)}, and the claim then follows from the translation invariance of {\mu}. \Box

Given any Bohr set {B(S,\rho)}, we define a normalised “Lipschitz” cutoff function {\nu_{B(S,\rho)}: G \rightarrow {\bf R}} by the formula

\displaystyle  \nu_{B(S,\rho)}(x) = c_{B(S,\rho)} (1 - \frac{1}{\rho} \sup_{\xi \in S} \|\xi \cdot x\|_{{\bf R}/{\bf Z}})_+ \ \ \ \ \ (1)

where {c_{B(S,\rho)}} is the constant such that

\displaystyle  \int_G \nu_{B(S,\rho)}\ d\mu = 1,

thus

\displaystyle c_{B(S,\rho)} = \left( \int_{B(S,\rho)} (1 - \frac{1}{\rho} \sup_{\xi \in S} \|\xi \cdot x\|_{{\bf R}/{\bf Z}})\ d\mu(x) \right)^{-1}.

The function {\nu_{B(S,\rho)}} should be viewed as an {L^1}-normalised “tent function” cutoff to {B(S,\rho)}. Note from Lemma 2 that

\displaystyle  1 \ll_{|S|,\rho} c_{B(S,\rho)} \ll_{|S|,\rho} 1. \ \ \ \ \ (2)

We then have the following sharper version of Theorem 1:

Theorem 3 (Roth-Khintchine theorem) Let {G = (G,+)} be a {2}-divisible compact abelian group, with Haar probability measure {\mu}, and let {\epsilon>0}. Then for any measurable function {f: G \rightarrow [0,1]}, there exists a Bohr set {B(S,\rho)} with {|S| \ll_\epsilon 1} and {\rho \gg_\epsilon 1} such that

\displaystyle  \int_G \int_G f(x) f(x+r) f(x+2r) \nu_{B(S,\rho)}*\nu_{B(S,\rho)}(r)\ d\mu(x) d\mu(r) \ \ \ \ \ (3)

\displaystyle  \geq (\int_G f\ d\mu)^3 - O(\epsilon)

where {*} denotes the convolution operation

\displaystyle  f*g(x) := \int_G f(y) g(x-y)\ d\mu(y).

A variant of this result (expressed in the language of ergodic theory) appears in this paper of Bergelson, Host, and Kra; a combinatorial version of the Bergelson-Host-Kra result that is closer to Theorem 3 subsequently appeared in this paper of Ben Green and myself, but this theorem arguably appears implicitly in a much older paper of Bourgain. To see why Theorem 3 implies Theorem 1, we apply the theorem with {f := 1_A} and {\epsilon} equal to a small multiple of {\alpha^3} to conclude that there is a Bohr set {B(S,\rho)} with {|S| \ll_\alpha 1} and {\rho \gg_\alpha 1} such that

\displaystyle  \int_G \int_G 1_A(x) 1_A(x+r) 1_A(x+2r) \nu_{B(S,\rho)}*\nu_{B(S,\rho)}(r)\ d\mu(x) d\mu(r) \gg \alpha^3.

But from (2) we have the pointwise bound {\nu_{B(S,\rho)}*\nu_{B(S,\rho)} \ll_\alpha 1}, and Theorem 1 follows.

Below the fold, we give a short proof of Theorem 3, using an “energy pigeonholing” argument that essentially dates back to the 1986 paper of Bourgain mentioned previously (not to be confused with a later 1999 paper of Bourgain on Roth’s theorem that was highly influential, for instance in emphasising the importance of Bohr sets). The idea is to use the pigeonhole principle to choose the Bohr set {B(S,\rho)} to capture all the “large Fourier coefficients” of {f}, but such that a certain “dilate” of {B(S,\rho)} does not capture much more Fourier energy of {f} than {B(S,\rho)} itself. The bound (3) may then be obtained through elementary Fourier analysis, without much need to explicitly compute things like the Fourier transform of an indicator function of a Bohr set. (However, the bound obtained by this argument is going to be quite poor – of tower-exponential type.) To do this we perform a structural decomposition of {f} into “structured”, “small”, and “highly pseudorandom” components, as is common in the subject (e.g. in this previous blog post), but even though we crucially need to retain non-negativity of one of the components in this decomposition, we can avoid recourse to conditional expectation with respect to a partition (or “factor”) of the space, using instead convolution with one of the {\nu_{B(S,\rho)}} considered above to achieve a similar effect.

Read the rest of this entry »

Throughout this post, we will work only at the formal level of analysis, ignoring issues of convergence of integrals, justifying differentiation under the integral sign, and so forth. (Rigorous justification of the conservation laws and other identities arising from the formal manipulations below can usually be established in an a posteriori fashion once the identities are in hand, without the need to rigorously justify the manipulations used to come up with these identities).

It is a remarkable fact in the theory of differential equations that many of the ordinary and partial differential equations that are of interest (particularly in geometric PDE, or PDE arising from mathematical physics) admit a variational formulation; thus, a collection {\Phi: \Omega \rightarrow M} of one or more fields on a domain {\Omega} taking values in a space {M} will solve the differential equation of interest if and only if {\Phi} is a critical point to the functional

\displaystyle  J[\Phi] := \int_\Omega L( x, \Phi(x), D\Phi(x) )\ dx \ \ \ \ \ (1)

involving the fields {\Phi} and their first derivatives {D\Phi}, where the Lagrangian {L: \Sigma \rightarrow {\bf R}} is a function on the vector bundle {\Sigma} over {\Omega \times M} consisting of triples {(x, q, \dot q)} with {x \in \Omega}, {q \in M}, and {\dot q: T_x \Omega \rightarrow T_q M} a linear transformation; we also usually keep the boundary data of {\Phi} fixed in case {\Omega} has a non-trivial boundary, although we will ignore these issues here. (We also ignore the possibility of having additional constraints imposed on {\Phi} and {D\Phi}, which require the machinery of Lagrange multipliers to deal with, but which will only serve as a distraction for the current discussion.) It is common to use local coordinates to parameterise {\Omega} as {{\bf R}^d} and {M} as {{\bf R}^n}, in which case {\Sigma} can be viewed locally as a function on {{\bf R}^d \times {\bf R}^n \times {\bf R}^{dn}}.

Example 1 (Geodesic flow) Take {\Omega = [0,1]} and {M = (M,g)} to be a Riemannian manifold, which we will write locally in coordinates as {{\bf R}^n} with metric {g_{ij}(q)} for {i,j=1,\dots,n}. A geodesic {\gamma: [0,1] \rightarrow M} is then a critical point (keeping {\gamma(0),\gamma(1)} fixed) of the energy functional

\displaystyle  J[\gamma] := \frac{1}{2} \int_0^1 g_{\gamma(t)}( D\gamma(t), D\gamma(t) )\ dt

or in coordinates (ignoring coordinate patch issues, and using the usual summation conventions)

\displaystyle  J[\gamma] = \frac{1}{2} \int_0^1 g_{ij}(\gamma(t)) \dot \gamma^i(t) \dot \gamma^j(t)\ dt.

As discussed in this previous post, both the Euler equations for rigid body motion, and the Euler equations for incompressible inviscid flow, can be interpreted as geodesic flow (though in the latter case, one has to work really formally, as the manifold {M} is now infinite dimensional).

More generally, if {\Omega = (\Omega,h)} is itself a Riemannian manifold, which we write locally in coordinates as {{\bf R}^d} with metric {h_{ab}(x)} for {a,b=1,\dots,d}, then a harmonic map {\Phi: \Omega \rightarrow M} is a critical point of the energy functional

\displaystyle  J[\Phi] := \frac{1}{2} \int_\Omega h(x) \otimes g_{\gamma(x)}( D\gamma(x), D\gamma(x) )\ dh(x)

or in coordinates (again ignoring coordinate patch issues)

\displaystyle  J[\Phi] = \frac{1}{2} \int_{{\bf R}^d} h_{ab}(x) g_{ij}(\Phi(x)) (\partial_a \Phi^i(x)) (\partial_b \Phi^j(x))\ \sqrt{\det(h(x))}\ dx.

If we replace the Riemannian manifold {\Omega} by a Lorentzian manifold, such as Minkowski space {{\bf R}^{1+3}}, then the notion of a harmonic map is replaced by that of a wave map, which generalises the scalar wave equation (which corresponds to the case {M={\bf R}}).

Example 2 ({N}-particle interactions) Take {\Omega = {\bf R}} and {M = {\bf R}^3 \otimes {\bf R}^N}; then a function {\Phi: \Omega \rightarrow M} can be interpreted as a collection of {N} trajectories {q_1,\dots,q_N: {\bf R} \rightarrow {\bf R}^3} in space, which we give a physical interpretation as the trajectories of {N} particles. If we assign each particle a positive mass {m_1,\dots,m_N > 0}, and also introduce a potential energy function {V: M \rightarrow {\bf R}}, then it turns out that Newton’s laws of motion {F=ma} in this context (with the force {F_i} on the {i^{th}} particle being given by the conservative force {-\nabla_{q_i} V}) are equivalent to the trajectories {q_1,\dots,q_N} being a critical point of the action functional

\displaystyle  J[\Phi] := \int_{\bf R} \sum_{i=1}^N \frac{1}{2} m_i |\dot q_i(t)|^2 - V( q_1(t),\dots,q_N(t) )\ dt.

Formally, if {\Phi = \Phi_0} is a critical point of a functional {J[\Phi]}, this means that

\displaystyle  \frac{d}{ds} J[ \Phi[s] ]|_{s=0} = 0

whenever {s \mapsto \Phi[s]} is a (smooth) deformation with {\Phi[0]=\Phi_0} (and with {\Phi[s]} respecting whatever boundary conditions are appropriate). Interchanging the derivative and integral, we (formally, at least) arrive at

\displaystyle  \int_\Omega \frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0}\ dx = 0. \ \ \ \ \ (2)

Write {\delta \Phi := \frac{d}{ds} \Phi[s]|_{s=0}} for the infinitesimal deformation of {\Phi_0}. By the chain rule, {\frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0}} can be expressed in terms of {x, \Phi_0(x), \delta \Phi(x), D\Phi_0(x), D \delta \Phi(x)}. In coordinates, we have

\displaystyle  \frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0} = \delta \Phi^i(x) L_{q^i}(x,\Phi_0(x), D\Phi_0(x)) \ \ \ \ \ (3)

\displaystyle  + \partial_{x^a} \delta \Phi^i(x) L_{\partial_{x^a} q^i} (x,\Phi_0(x), D\Phi_0(x)),

where we parameterise {\Sigma} by {x, (q^i)_{i=1,\dots,n}, (\partial_{x^a} q^i)_{a=1,\dots,d; i=1,\dots,n}}, and we use subscripts on {L} to denote partial derivatives in the various coefficients. (One can of course work in a coordinate-free manner here if one really wants to, but the notation becomes a little cumbersome due to the need to carefully split up the tangent space of {\Sigma}, and we will not do so here.) Thus we can view (2) as an integral identity that asserts the vanishing of a certain integral, whose integrand involves {x, \Phi_0(x), \delta \Phi(x), D\Phi_0(x), D \delta \Phi(x)}, where {\delta \Phi} vanishes at the boundary but is otherwise unconstrained.

A general rule of thumb in PDE and calculus of variations is that whenever one has an integral identity of the form {\int_\Omega F(x)\ dx = 0} for some class of functions {F} that vanishes on the boundary, then there must be an associated differential identity {F = \hbox{div} X} that justifies this integral identity through Stokes’ theorem. This rule of thumb helps explain why integration by parts is used so frequently in PDE to justify integral identities. The rule of thumb can fail when one is dealing with “global” or “cohomologically non-trivial” integral identities of a topological nature, such as the Gauss-Bonnet or Kazhdan-Warner identities, but is quite reliable for “local” or “cohomologically trivial” identities, such as those arising from calculus of variations.

In any case, if we apply this rule to (2), we expect that the integrand {\frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0}} should be expressible as a spatial divergence. This is indeed the case:

Proposition 1 (Formal) Let {\Phi = \Phi_0} be a critical point of the functional {J[\Phi]} defined in (1). Then for any deformation {s \mapsto \Phi[s]} with {\Phi[0] = \Phi_0}, we have

\displaystyle  \frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0} = \hbox{div} X \ \ \ \ \ (4)

where {X} is the vector field that is expressible in coordinates as

\displaystyle  X^a := \delta \Phi^i(x) L_{\partial_{x^a} q^i}(x,\Phi_0(x), D\Phi_0(x)). \ \ \ \ \ (5)

Proof: Comparing (4) with (3), we see that the claim is equivalent to the Euler-Lagrange equation

\displaystyle  L_{q^i}(x,\Phi_0(x), D\Phi_0(x)) - \partial_{x^a} L_{\partial_{x^a} q^i}(x,\Phi_0(x), D\Phi_0(x)) = 0. \ \ \ \ \ (6)

The same computation, together with an integration by parts, shows that (2) may be rewritten as

\displaystyle  \int_\Omega ( L_{q^i}(x,\Phi_0(x), D\Phi_0(x)) - \partial_{x^a} L_{\partial_{x^a} q^i}(x,\Phi_0(x), D\Phi_0(x)) ) \delta \Phi^i(x)\ dx = 0.

Since {\delta \Phi^i(x)} is unconstrained on the interior of {\Omega}, the claim (6) follows (at a formal level, at least). \Box

Many variational problems also enjoy one-parameter continuous symmetries: given any field {\Phi_0} (not necessarily a critical point), one can place that field in a one-parameter family {s \mapsto \Phi[s]} with {\Phi[0] = \Phi_0}, such that

\displaystyle  J[ \Phi[s] ] = J[ \Phi[0] ]

for all {s}; in particular,

\displaystyle  \frac{d}{ds} J[ \Phi[s] ]|_{s=0} = 0,

which can be written as (2) as before. Applying the previous rule of thumb, we thus expect another divergence identity

\displaystyle  \frac{d}{ds} L( x, \Phi[s](x), D\Phi[s](x) )|_{s=0} = \hbox{div} Y \ \ \ \ \ (7)

whenever {s \mapsto \Phi[s]} arises from a continuous one-parameter symmetry. This expectation is indeed the case in many examples. For instance, if the spatial domain {\Omega} is the Euclidean space {{\bf R}^d}, and the Lagrangian (when expressed in coordinates) has no direct dependence on the spatial variable {x}, thus

\displaystyle  L( x, \Phi(x), D\Phi(x) ) = L( \Phi(x), D\Phi(x) ), \ \ \ \ \ (8)

then we obtain {d} translation symmetries

\displaystyle  \Phi[s](x) := \Phi(x - s e^a )

for {a=1,\dots,d}, where {e^1,\dots,e^d} is the standard basis for {{\bf R}^d}. For a fixed {a}, the left-hand side of (7) then becomes

\displaystyle  \frac{d}{ds} L( \Phi(x-se^a), D\Phi(x-se^a) )|_{s=0} = -\partial_{x^a} [ L( \Phi(x), D\Phi(x) ) ]

\displaystyle  = \hbox{div} Y

where {Y(x) = - L(\Phi(x), D\Phi(x)) e^a}. Another common type of symmetry is a pointwise symmetry, in which

\displaystyle  L( x, \Phi[s](x), D\Phi[s](x) ) = L( x, \Phi[0](x), D\Phi[0](x) ) \ \ \ \ \ (9)

for all {x}, in which case (7) clearly holds with {Y=0}.

If we subtract (4) from (7), we obtain the celebrated theorem of Noether linking symmetries with conservation laws:

Theorem 2 (Noether’s theorem) Suppose that {\Phi_0} is a critical point of the functional (1), and let {\Phi[s]} be a one-parameter continuous symmetry with {\Phi[0] = \Phi_0}. Let {X} be the vector field in (5), and let {Y} be the vector field in (7). Then we have the pointwise conservation law

\displaystyle  \hbox{div}(X-Y) = 0.

In particular, for one-dimensional variational problems, in which {\Omega \subset {\bf R}}, we have the conservation law {(X-Y)(t) = (X-Y)(0)} for all {t \in \Omega} (assuming of course that {\Omega} is connected and contains {0}).

Noether’s theorem gives a systematic way to locate conservation laws for solutions to variational problems. For instance, if {\Omega \subset {\bf R}} and the Lagrangian has no explicit time dependence, thus

\displaystyle  L(t, \Phi(t), \dot \Phi(t)) = L(\Phi(t), \dot \Phi(t)),

then by using the time translation symmetry {\Phi[s](t) := \Phi(t-s)}, we have

\displaystyle  Y(t) = - L( \Phi(t), \dot\Phi(t) )

as discussed previously, whereas we have {\delta \Phi(t) = - \dot \Phi(t)}, and hence by (5)

\displaystyle  X(t) := - \dot \Phi^i(x) L_{\dot q^i}(\Phi(t), \dot \Phi(t)),

and so Noether’s theorem gives conservation of the Hamiltonian

\displaystyle  H(t) := \dot \Phi^i(x) L_{\dot q^i}(\Phi(t), \dot \Phi(t))- L(\Phi(t), \dot \Phi(t)). \ \ \ \ \ (10)

For instance, for geodesic flow, the Hamiltonian works out to be

\displaystyle  H(t) = \frac{1}{2} g_{ij}(\gamma(t)) \dot \gamma^i(t) \dot \gamma^j(t),

so we see that the speed of the geodesic is conserved over time.

For pointwise symmetries (9), {Y} vanishes, and so Noether’s theorem simplifies to {\hbox{div} X = 0}; in the one-dimensional case {\Omega \subset {\bf R}}, we thus see from (5) that the quantity

\displaystyle  \delta \Phi^i(t) L_{\dot q^i}(t,\Phi_0(t), \dot \Phi_0(t)) \ \ \ \ \ (11)

is conserved in time. For instance, for the {N}-particle system in Example 2, if we have the translation invariance

\displaystyle  V( q_1 + h, \dots, q_N + h ) = V( q_1, \dots, q_N )

for all {q_1,\dots,q_N,h \in {\bf R}^3}, then we have the pointwise translation symmetry

\displaystyle  q_i[s](t) := q_i(t) + s e^j

for all {i=1,\dots,N}, {s \in{\bf R}} and some {j=1,\dots,3}, in which case {\dot q_i(t) = e^j}, and the conserved quantity (11) becomes

\displaystyle  \sum_{i=1}^n m_i \dot q_i^j(t);

as {j=1,\dots,3} was arbitrary, this establishes conservation of the total momentum

\displaystyle  \sum_{i=1}^n m_i \dot q_i(t).

Similarly, if we have the rotation invariance

\displaystyle  V( R q_1, \dots, Rq_N ) = V( q_1, \dots, q_N )

for any {q_1,\dots,q_N \in {\bf R}^3} and {R \in SO(3)}, then we have the pointwise rotation symmetry

\displaystyle  q_i[s](t) := \exp( s A ) q_i(t)

for any skew-symmetric real {3 \times 3} matrix {A}, in which case {\dot q_i(t) = A q_i(t)}, and the conserved quantity (11) becomes

\displaystyle  \sum_{i=1}^n m_i \langle A q_i(t), \dot q_i(t) \rangle;

since {A} is an arbitrary skew-symmetric matrix, this establishes conservation of the total angular momentum

\displaystyle  \sum_{i=1}^n m_i q_i(t) \wedge \dot q_i(t).

Below the fold, I will describe how Noether’s theorem can be used to locate all of the conserved quantities for the Euler equations of inviscid fluid flow, discussed in this previous post, by interpreting that flow as geodesic flow in an infinite dimensional manifold.

Read the rest of this entry »

Archives

RSS Google+ feed

  • An error has occurred; the feed is probably down. Try again later.
Follow

Get every new post delivered to your Inbox.

Join 3,576 other followers