Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

November 11, 2020

The Categorical Origins of Lebesgue Integration, Revisited

Posted by Tom Leinster

I’ve just arXived a new paper: The categorical origins of Lebesgue integration (arXiv:2011.00412). Longtime Café readers may remember that I blogged about this stuff back in 2014, but I’ve only just written it up.

What’s it all about? There are two main theorems, which loosely are as follows:

Theorem A   The Banach space L 1[0,1]L^1[0, 1] has a simple universal property. This leads to a unique characterization of integration.

Theorem B   The functor L 1:L^1: (finite measure spaces) \to (Banach spaces) has a simple universal property. This leads to a unique characterization of integration on finite measure spaces.

But there’s more! The mist has cleared on some important things since that last post back in 2014. I’ll give you the highlights.

Universal property of L p[0,1]L^p[0, 1]   Let 1p<1 \leq p \lt \infty. The Banach space L p[0,1]L^p[0, 1], with a little bit of extra structure, has a simple universal property.

To say what that universal property is, I need to put a norm on the direct sum VWV \oplus W of two Banach spaces, and the one I’ll use is this:

(v,w)=(12(v p+w p)) 1/p. \| (v, w) \| = \Bigl( \tfrac{1}{2} ( \|v\|^p + \|w\|^p ) \Bigr)^{1/p}.

In words, it’s the power mean of order pp of the norms of vv and ww.

Now consider the category 𝒜 p\mathcal{A}^p of triples (V,v,δ)(V, v, \delta), where VV is a Banach space, vv is a point in the closed unit ball of VV, and δ\delta is a linear contraction V pVVV \oplus_p V \to V such that δ(v,v)=v\delta(v, v) = v. The maps are the linear contractions that preserve the structure in the obvious sense.

One object of 𝒜 p\mathcal{A}^p is the triple (L p[0,1],I,γ)(L^p[0, 1], I, \gamma), where II is the constant function 11 and γ\gamma takes two functions on [0,1][0, 1], juxtaposes them, and scales the domain by a factor of 1/21/2:

the function gamma

The theorem is that (L p[0,1],I,γ)(L^p[0, 1], I, \gamma) is, in fact, the initial object of 𝒜 p\mathcal{A}^p. Among other things, this characterizes the Banach space L p[0,1]L^p[0, 1] uniquely up to isometric isomorphism.

When p=1p = 1, another object (V,v,δ)(V, v, \delta) of 𝒜 1\mathcal{A}^1 is the ground field 𝔽\mathbb{F} (either \mathbb{R} or \mathbb{C}) together with the element 1𝔽1 \in \mathbb{F} and the arithmetic mean operation m:𝔽𝔽𝔽m : \mathbb{F} \oplus \mathbb{F} \to \mathbb{F}. Since (L 1[0,1],I,γ)(L^1[0, 1], I, \gamma) is initial in 𝒜 1\mathcal{A}^1, there’s a unique map (L 1[0,1],I,γ)(𝔽,1,m)(L^1[0, 1], I, \gamma) \to (\mathbb{F}, 1, m). It’s nothing but integration.

Continuous functions on the Cantor set   The result I just stated characterized L p[0,1]L^p[0, 1] for 1p<1 \leq p \lt \infty, but excluded p=p = \infty. What about that case?

The category 𝒜 \mathcal{A}^\infty does have an initial object, but it’s not L [0,1]L^\infty[0, 1]. In fact, it’s C({0,1} )C(\{0, 1\}^\mathbb{N}), the continuous functions on the Cantor set.

This isn’t as outlandish as it might seem. First of all, the interval [0,1][0, 1] and the Cantor set {0,1} \{0, 1\}^\mathbb{N} are equivalent as measure spaces. (Here I’m giving the Cantor set the probability measure that corresponds to thinking of its elements as sequences of tosses of a fair coin, and of course I’m giving [0,1][0, 1] Lebesgue measure.) So, L p[0,1]L p({0,1} )L^p[0, 1] \cong L^p(\{0, 1\}^\mathbb{N}), which means that the previous result could equally well be thought of as a universal characterization of L p({0,1} )L^p(\{0, 1\}^\mathbb{N}).

On that basis, one might expect the initial object of 𝒜 \mathcal{A}^\infty to be L ({0,1} )L^\infty(\{0, 1\}^\mathbb{N}), the essentially bounded functions on the Cantor set. But it’s not! It’s the continuous functions.

Connecting up with actual function spaces   It’s all very well characterizing L p[0,1]L^p[0, 1] abstractly, but can we relate its elements to actual functions on [0,1][0, 1]?

The answer is yes, in two different senses.

First, we might want to start with a nice enough function on [0,1][0, 1] and produce from it an element of the abstractly-characterized L p[0,1]L^p[0, 1]. Let’s interpret “nice enough” as “continuous”, so that we’re looking to derive the inclusion map C[0,1]L p[0,1]C[0, 1] \to L^p[0, 1] from the universal properties above.

It can be done! The way it goes is this. The usual map {0,1} [0,1]\{0, 1\}^\mathbb{N} \to [0, 1] induces a map C[0,1]C({0,1} )C[0, 1] \to C(\{0, 1\}^\mathbb{N}). Also, the universal property of C({0,1} )C(\{0, 1\}^\mathbb{N}) induces a map from it to L p[0,1]L^p[0, 1]. Composing the two gives a map C[0,1]L p[0,1]C[0, 1] \to L^p[0, 1], and I show that in concrete terms, it’s the inclusion.

Second — and in the opposite direction — we might want to start with an element of the abstractly characterized L 1[0,1]L^1[0, 1] and produce from it an actual function on [0,1][0, 1]. Well: we might want to do this, but it’s not realistic: after all, elements of L 1[0,1]L^1[0, 1] are equivalence classes of integrable functions, up to equality almost everywhere. So that’s the best we can hope to get. There can be no hope of evaluating an element of L 1[0,1]L^1[0, 1] at an element of [0,1][0, 1].

What the universal property of L 1[0,1]L^1[0, 1] does get us is a canonical map L 1[0,1]C[0,1]L^1[0, 1] \to C[0, 1]. Concretely, this is what generations of undergraduates would write as fFf \mapsto F, where F(x)= 0 xfF(x) = \int_0^x f. (Thanks to Mark Meckes for pointing this out to me and letting me include it in the paper.)

I’m not saying quite how you get this map fFf \mapsto F; it’s Proposition 2.4 in the paper. But the point is that this map comes straight from the universal property of L 1[0,1]L^1[0, 1], and from FF we can extract an actual function representing ff. Indeed, it’s a version of the fundamental theorem of calculus that F=fF' = f almost everywhere. So that’s you extract a genuine function.

Conjugate pairing   A whole lot of stuff in analysis involves exponents pp and qq that are “conjugate”, meaning that 1/p+1/q=11/p + 1/q = 1. (And a whole lot of people have observed that in retrospect, it might have been better to index the L pL^p spaces using 1/p1/p rather than pp. Too late!) Most famously, when pp and qq are conjugate, L p[0,1]L^p[0, 1] and L q[0,1]L^q[0, 1] are dual, as long as pp and qq are not 11 or \infty.

We don’t seem to get this duality directly from the universal property, but we do get the pairing function

L p[0,1]×L q[0,1]𝔽. L^p[0, 1] \times L^q[0, 1] \to \mathbb{F}.

Concretely, this function is (f,g)fg(f, g) \mapsto \int f \cdot g. In other words, it’s the composite

L p[0,1]×L q[0,1]L 1[0,1]𝔽, L^p[0, 1] \times L^q[0, 1] \to L^1[0, 1] \to \mathbb{F},

where the first map is multiplication and the second is integration. We already derived the second map from the universal property of L 1[0,1]L^1[0, 1], and the paper shows how to derive the first map from the universal properties of L pL^p and L qL^q.

Sequence spaces   Everything so far has been about spaces of functions on [0,1][0, 1]. But we can follow a similar strategy to get universal characterizations of spaces of sequences. All I’ll do here is point to Proposition 2.10 of the paper, which states a universal property of the sequence space p\ell^p. That’s for p<p \lt \infty. When you put p=p = \infty, what pops out is not \ell^\infty but the space c 0c_0 of sequences converging to 00.

The L pL^p functors   The second half of the paper is about L pL^p as a functor, rather than L p[0,1]L^p[0, 1] or L p()= pL^p(\mathbb{N}) = \ell^p.

The L pL^p construction takes a measure space as input and spits out a Banach space. There are at least a couple of ways in which it’s functorial:

  • The one you hear about most is that it’s contravariant with respect to measure-preserving maps, via composition.

  • But it’s also covariant with respect to embeddings (i.e. inclusions of measure subspaces, up to isomorphism). If you’ve got an L pL^p function defined on a subspace YXY \subseteq X, you can simply extend it by 00 to get an L pL^p function on all of XX.

  • Combining these two types of functoriality, L pL^p is contravariant with respect to measure-preserving partial maps.

I’ll write Meas\mathbf{Meas} for the category of measure spaces and measure-preserving partial maps, so that L pL^p is a functor

L p:Meas opBan. L^p: \mathbf{Meas}^{op} \to \mathbf{Ban}.

By “measure space” I always mean a finite measure space — not one whose underlying set is necessarily finite, but where the total measure of the space is finite.

For each measure space XX, there’s a special element of L p(X)L^p(X): the function I XI_X with constant value 11. The functor L pL^p, together with the family

(I XL p(X)) XMeas, \bigl(I_X \in L^p(X)\bigr)_{X \in \mathbf{Meas}},

has some elementary properties that we could write down. For example, if s:XYs: X \to Y is a measure-preserving map then L p(s):L p(Y)L p(X)L^p(s): L^p(Y) \to L^p(X) maps I YI_Y to I XI_X.

The second main theorem is that the pair (L p,I)(L^p, I) is initial with these simple properties.

That is, I define a category p\mathcal{B}^p of pairs (F,v)(F, v) where F:Meas opBanF: \mathbf{Meas}^{op} \to \mathbf{Ban} and vv is a family (v XF(X)) XMeas(v_X \in F(X))_{X \in \mathbf{Meas}} satisfying those elementary properties, and the theorem is that (L p,I)(L^p, I) is the initial object of p\mathcal{B}^p.

Baby versions   Actually, this theorem is naturally the third of a trilogy of results. In the paper, I used the first two as stepping stones to get to the third.

The first is about a similar category of pairs (F,v)(F, v), where FF is now a functor from Meas op\mathbf{Meas}^{op} to the category of mere vector spaces (not Banach spaces). The initial object is the pair (𝒮,I)(\mathcal{S}, I), where 𝒮(X)\mathcal{S}(X) is the vector space of simple functions on XX.

The second is about a category of pairs (F,v)(F, v), where FF is a functor from Meas op\mathbf{Meas}^{op} to the category of normed vector spaces. The initial object is the pair (S p,I)(S^p, I), where S p(X)S^p(X) is the normed vector spaces of simple functions up to equality almost everywhere, with the pp-norm.

So, informally:

  • The universal vector space on a measure space consists of the simple functions.

  • The universal normed vector space consists of the a.e. equivalence classes of simple functions.

  • The universal Banach space consists of the a.e. equivalence classes of integrable functions.

Integration on an arbitrary measure space   The universal characterization of the functor L 1L^1 gives a unique characterization of integration.

Just as for [0,1][0, 1], this comes about by choosing a suitable object of 1\mathcal{B}^1 and applying the fact that (L 1,I)(L^1, I) is initial in 1\mathcal{B}^1. That “suitable object” is the constant functor 𝔽\mathbb{F} (the ground field) together with, for each measure space X=(X,μ X)X = (X, \mu_X), the total measure μ X(X)𝔽\mu_X(X) \in \mathbb{F}. The initiality of (L 1,I)(L^1, I) gives us a map L 1(X)𝔽L^1(X) \to \mathbb{F} for each measure space XX, and it won’t surprise you that it’s our old friend X\int_X.

The action of functions on measures   Whenever you have a measure space (X,μ)(X, \mu) and an integrable function ff on it, you get a new measure on XX that’s traditionally written as fdμf\,d\mu but which I prefer to write as fμf\mu. (It seems to me that the dds just make life messier.) It’s defined on measurable sets AXA \subseteq X by

(fμ)(A)= Afdμ. (f \mu)(A) = \int_A f \, d\mu.

This construction, too, can be derived from the universal property of L 1L^1. This time I won’t say how! It’s not at all hard, but I’m running out of steam, so I’ll just refer you to Proposition 3.10 of the paper. And I’ll note that being able to integrate ff on arbitrary, “small”, regions of XX is as close as we can truthfully come to realizing fL p(X)f \in L^p(X) as an actual function, given that in reality ff is only an equivalence class of functions.

Universal characterization of L 2L^2   We all know that there’s something special about p=2p = 2, and there’s a theorem that does a little bit to express that special nature. It says that (L 2,I)(L^2, I) is the initial object of a category of pairs (F,v)(F, v), where FF is now a functor from Meas op\mathbf{Meas}^{op} to Hilbert spaces. And the axioms that these pairs are required to satisfy are slightly simpler and more intuitive than for general pp.

What’s the point?   I’ve written about as much as I want to for one post, but I didn’t want to finish without mentioning this general question.

A certain kind of no-nonsense mathematician would find all these results very odd. What do they tell us that we didn’t already know? I did a lot of thinking about how to respond, and I gave five answers in the introduction (page 2). I think my favourite one is this:

Fourth, the second main theorem, characterizing the L pL^p functors, provides a guide for the discovery of new theories of integration. A researcher seeking the right notion of integration in some new context (perhaps some new kind of function on a new kind of space) could follow the same template: decide what kind of spaces the integrable functions should form (perhaps not Banach spaces) and what kind of functoriality should hold, formulate a universal property analogous to the one used below, and find the functor satisfying it.

I leave you to discuss this in the comments!

Posted at November 11, 2020 10:50 PM UTC

TrackBack URL for this Entry:   https://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/3266

9 Comments & 0 Trackbacks

Re: The Categorical Origins of Lebesgue Integration, Revisited

Looks like a good read, Tom!

We had a small burst of functional analysis meets category theory earlier this year with talk of Smith spaces (aka Waelbrock dual spaces) forming Ban opBan^{op}.

Smith spaces are more natural than Banach spaces from the condensed perspective because they are controlled by a nice compact subset instead of a nice open subset (the unit ball) like in the Banach setting. Also, Banach spaces are filtered colimits of Smith spaces while Smith spaces are filtered limits of Banach spaces; filtered colimits are nicer in their algebraic and homological properties so it makes sense to take Smith spaces as fundamental.

Presumably, your universality results could be dualized to Smith spaces.

Posted by: David Corfield on November 12, 2020 8:47 AM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

Thanks, David. I missed that part of that conversation.

It sounds like ultimately, liquid modules might be the way to go.

Posted by: Tom Leinster on November 12, 2020 10:03 AM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

Given recent events, and all the other striking neologisms coined by Clausen and Scholze, it’s a pity they didn’t have some algebraic objects which they could abbreviate to “SWORDs”…

Posted by: Yemon Choi on November 12, 2020 7:01 PM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

I’m missing the joke. Why SWORDs, Yemon?

Posted by: David Corfield on November 13, 2020 10:07 AM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

Courtesy of the Wayback Machine: a 2014 meme that coincidentally became rather topical on November 7th 2020.

(I also happen to quite like the album, but YMMV.)

Posted by: Yemon Choi on November 15, 2020 1:00 AM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

There’s a typo in the first equation. There’s no closing outer parenthesis on the right hand side. Also I suspect a missing power of 1/p.

Posted by: carl feynman on November 12, 2020 3:21 PM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

Thanks, but I don’t get it. Do you mean this equation?

first equation

That’s from a screenshot of how I see it on my browser. What browser are you using?

Posted by: Tom Leinster on November 12, 2020 4:31 PM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

In the “Related work” section, I should have mentioned Martín Escardó and Alex Simpson’s paper A universal characterization of the closed Euclidean interval. I’ll fix that when I come to revise the paper. Thanks to Christopher Townsend for pointing that out by email.

Posted by: Tom Leinster on November 12, 2020 4:25 PM | Permalink | Reply to this

Re: The Categorical Origins of Lebesgue Integration, Revisited

Very interesting characterizations!

Posted by: Bruce Bartlett on November 24, 2020 7:20 AM | Permalink | Reply to this

Post a New Comment