Before we begin, I want to fuss around with model theory again. Recall that if is a nonstandard complex number, then denotes the standard part of , if it exists. We previously defined what it meant for a nonstandard random variable to be infinitesimal in distribution. One can define something similar for any metrizable space with a notion of , where is infinitesimal provided that is. For example, a nonstandard random variable is infinitesimal in if for every compact set that can take values in, is infinitesimal, since is metrizable with

whenever is a compact exhaustion. If is nonstandard, is infinitesimal in some metrizable space, and is standard, then we call the standard part of in ; then the standard part is unique since metrizable spaces are Hausdorff.

If the metrizable space is compact, the case that we will mainly be interested in, then the standard part exists. This is a point that we will use again and again. Passing to the cheap perspective, this says that if is a compact metric space and is a sequence in , then there is a which is approximates infinitely often, but that’s just the Bolzano-Weierstrass theorem. Last time used Prokohov’s theorem to show that if is a nonstandard tight random variable, then has a standard part in distribution.

We now restate and prove Proposition 9 from the previous post.

Theorem 1 (distribution of random zeroes)Let be a nonstandard natural, a monic polynomial of degree with all zeroes in , and let be a zero of . Suppose that has no zeroes in . Let be a random zero of and a random zero of . Then:

- If (
case zero), then and are identically distributed and almost surely lie in the curveIn particular, in probability. Moreover, for every compact set ,

- If (
case one), then is uniformly distributed on the unit circle and is almost surely zero. Moreover,

**1. Moment-generating functions and balayage **

We first show that and have equal moment-generating functions in a suitable sense.

To do this, we first show that they have the same logarithmic potential. Let be a random variable such that almost surely (that is, is almost surely bounded). Then the logarithmic potential

is defined almost everywhere as we discussed last time, and is harmonic outside of the essential range of .

Lemma 2Let be a nonstandard, almost surely bounded, random complex number. Then the standard part of is according to the topology of under Lebesgue measure.

*Proof:* We pass to the cheap perspective. If we instead have a random sequence of and in distribution, then in , since up to a small error in we can replace with a test function ; one then has

where in the weak topology of measures, is the distribution of , is the distribution of , and is a compact set equipped with Lebesgue measure.

Lemma 3For every , we have

In particular, .

*Proof:* By definition, , so . Now is a disc with diameter where is a rotation around the origin. Taking reciprocals preserves discs and preserves , so sits inside a disc with a diameter . Then is convex, so the expected value of is also . Therefore the Stieltjes transform

satisfies . In particular,

But we showed that

almost everywhere last time. This implies that for almost every ,

but all terms here are continuous so we can promote this to a statement that holds for every . In particular,

hence

Since while , is bounded from above and below by a constant times . Therefore the same holds of its logarithm , which is bounded from above and below by a constant times . This implies the first claim.

To derive the second claim from the first, we use the previous lemma, which implies that we must show that

in . But this follows since is integrable in two dimensions.

Lemma 4Let be an almost surely bounded random variable. Then

*Proof:* One has the Taylor series

Indeed, by rescaling and using , we may assume . The summands expand as

and the imaginary parts all cancel by symmetry about . Using the symmetry about again we get

This equals the left-hand side as long as . Taking expectations and commuting the expectation with the sum using Fubini’s theorem (since is almost surely bounded), we see the claim.

Lemma 5For all , one has

In particular, and have identical moments.

*Proof:* If we take then we conclude that

The left-hand side is a Fourier series, and by uniqueness of Fourier series it holds that for every ,

This gives a bound on the difference of moments

which is only possible if the moments of and are identical. The left-hand side doesn’t depend on , but if , , then and so the claim holds. On the other hand, if then this claim still holds, since we showed last time that

and obviously .

Here I was puzzled for a bit. Surely if two random variables have the same moment-generating function then they are identically distributed! But, while we can define the moment-generating function of a random variable as a formal power series , it is not true that has to have a positive radius of convergence, in which case the inverse Laplace transform of is ill-defined. Worse, the circle is not simply connected, and in case one, we have to look at a uniform distribution on the circle, whose moments therefore aren’t going to points on the circle, so the moment-generating function doesn’t tell us much.

**2. Balayage **

We recall the definition of the Poisson kernel :

whenever is a radius. Convolving the Poisson kernel against a continuous function on solves the Dirichlet problem of with boundary data .

Definition 6Let be a random variable. Thebalayageof is

Balayage is a puzzling notion. First, the name refers to a hair-care technique, which is kind of unhelpful. According to Tao, we’re supposed to interpret balayage as follows.

If is an initial datum for Brownian motion , then is the probability density of the first location where passes through . Tao asserts this without proof, but conveniently, this was a problem in my PDE class last semester. The idea is to approximate by the lattice , which we view as a graph where each vertex has degree , with one edge to each of the vertices directly above, below, left, and right of it. Then the Laplacian on is approximated by the graph Laplacian on , and Brownian motion is approximated by the discrete-time stochastic process wherein a particle starts at the vertex that best approximates and at each stage has a chance of moving to each of the vertices adjacent to its current position.

So suppose that and are actually vertices of . The probability density is harmonic in with respect to the graph Laplacian since it is the mean of as ranges over the adjacent vertices to ; therefore it remains harmonic as we take . The boundary conditions follow similarly.

Now if is a random initial datum for Brownian motion which starts in , the balayage of is again a probability density on that records where one expects the Brownian motion to escape, but this time the initial datum is also random.

I guess the point is that balayage serves as a substitute for the moment-generating function in the event that the latter is just a formal power series. We want to be able to use analytic techniques on the moment-generating function, but we can’t, so we just use balayage instead.

Let be the balayage of . Since is bounded, we can use Fubini’s theorem to commute the expectation with the sum and see that

provided that . It will be convenient to rewrite this in the form

so is uniquely determined by the moment-generating function of . In particular, and have identical balayage, and one has a bound

We claim that

which implies the bound

To see this, we discard the term since , which implies that

Up to a constant factor we may assume that the logarithms are base in which case we get a bound

The constant is absolute since .

By the integral test, we get a bound

Using the bound

for any and the change of variable (thus ), we get a bound

since the error in the exponent can’t affect the exponential decay of the integral in . Since we certainly have

this is a suitable tail bound.

To complete the proof of the claim we need to bound the main term. To this end we bound

Here denotes exponentiation. Now if is small enough (say ), this supremum will be attained when , thus . Therefore

Luckily is easy to differentiate: its critical point is . This gives

so

which was the bound we needed, and proves the claim. Maybe there’s an easier way to do this, because Tao says the claim is a trivial consequence of dyadic decomposition.

Let’s interpret the bound that we just proved. Well, if the balayage of is supposed to describe the point on the circle at which a Brownian motion with random initial datum escapes, a bound on a difference of two balyages should describe how the trajectories diverge after escaping. In this case, the divergence is infinitesimal, but at different speeds depending on . As , our infinitesimal divergence gains a positive standard part, while if stays close to , the divergence remains infinitesimal. This makes sense, since if we take a bigger circle we forget more and more about the fact that are not the same random variable, since Brownian motion has more time to “forget more stuff” as it just wanders around aimlessly. So in the regime where is close to , it is reasonable to take standard parts and pass to and , while in the regime where is close to this costs us dearly.

**3. Case zero **

Suppose that is infinitesimal.

We showed last time that , so is infinitesimal. Therefore almost surely.

I think there’s a typo here, because Tao lets range over and considers points , which don’t exist since while every point in has . I think this can be fixed by taking closures, which is what I do in the next lemma.

Tao proves a “qualitative” claim and then says that by repeating the argument and looking out for constants you can get a “quantitative” version which is what he actually needs. I’m just going to prove the quantitative argument straight-up. The idea is that if is a compact set which misses and then a Brownian motion with initial datum will probably escape through an arc which is close to , but is not close to so a Brownian motion which starts at will probably not escape through . Therefore have very different balayage, even though the difference in their balayage was already shown to be infinitesimal.

I guess this shows the true power of balayage: even though the moment-generating function is “just” a formal power series, we know that the essential supports of must “look like each other” up to rescaling in radius. This still holds in case one, where one of them is a circle and the other is the center of the circle. Either way, you get the same balayage, since whether you start at some point on a circle or you start in the center of the circle, if you’re a Brownian motion you will exhibit the same long-term behavior.

In the following lemmata, let be a compact set. The set is compact since it is the preimage of a compact set, so contained a compact interval .

Lemma 7One has

*Proof:* Since is compact the minimum is attained. Let be the minimum. Since is a real-valued harmonic function in , thus

the maximum principle implies that the worst case is when meets and , say . Then

Of course this is just a formal power series and doesn’t make much sense. But if instead where is very small depending on a given , then, after discarding quadratic terms in ,

This follows since in general

Now

since the integrand is maximized when , in which case the integrand evaluates to the measure of , which is since and has positive measure. Therefore

On the other hand, for any one has

so this implies gives a lower bound on the integral over .

Lemma 8If then

*Proof:* Let in the previous lemma, conditioning on the event , to see that

where . Taking expectations and dividing by the probability that , we can use Fubini’s theorem to deduce

where . Applying the bound on from the section on balayage, we deduce

We already showed that . So in order to show

which was the bound that we wanted, it suffices to show that for every such that ,

Tao says that “one can show” this claim, but I wasn’t able to do it. I think the point is that under those cirumstances one has and even as , so we have some control on . In fact I was able to compute

which suggests that this is the right direction, but the bounds I got never seemed to go anywhere. Someone bug me in the comments if there’s an easy way to do this that I somehow missed.

Now we take to complete the proof.

**4. Case one **

Suppose that is infinitesimal. Let be the expected value of (hence also of ). Let be a standard real.

We first need to go on an excursion to a paper of DĂ©got, who proves the following theorem:

Lemma 9One has

Moreover,

I will omit the proof since it takes some complex analysis I’m pretty unfamiliar with. It seems to need Grace’s theorem, which I guess is a variant of one of the many theorems in complex analysis that says that the polynomial image of a disk is kind of like a disk. It also uses some theorem called the Walsh contraction principle that involves polynomials on the projective plane. Curious.

In what follows we will say that an event is *standard-possible* if the probability that happens has positive standard part.

Lemma 10For every , is standard-possible. Besides, .

*Proof:* Since almost surely and

but

we have

Combining this with the lemma we see that the standard part of is , so

On the other hand,

and since is nonstandard, is infinitesimal, so the constant in gets eaten. In particular,

which implies that

and hence

Since this is true for arbitrary standard , underspill implies that there is an infinitesimal such that

But almost surely, and we just showed

So the claim holds.

We now allow to take the value , thus .

Lemma 11One has

and

Moreover, if , so has no zeroes in that disk.

*Proof:* Since

one has

Now and .

Here I drew two unit circles in , one entered at the origin and one at (since is infinitesimal); is (up to infinitesimal error) in the first circle and out of the second. The rightmost points of intersection between the two circles are on a vertical line which by the Pythagorean theorem is to the left of the vertical line , which in turn is to the left of the perpendicular bisector . Thus , and if then the real part of is . In particular, if the standard real part of is then , so has positive standard part.

By the previous lemma, it is standard-possible that the standard real part of is , so the standard real part of is standard-possibly positive and is almost surely nonnegative. Plugging into the above we deduce the existence of a standard absolute constant such that

In particular,

Keeping in mind that is nonstandard, this doesn’t necessarily mean that has nonpositive standard part, but it does give a pretty tight bound. Taking a first-order Taylor approximation we get

But one has

from the DĂ©got lemma. Clearly this term dominates so we have

Since one has a lower bound this implies is controlled from below by an absolute constant.

We also claim . In fact, we showed last time that

we want to show that , so it suffices to show that , or in other words that

Since by assumption on , this is trivial. We deduce that

and hence

Now Tao claims that the proof that is similar, if . Since was a valid choice of we have . Since , if then where is an absolute constant. Applying the fact that is standard-possible and is almost surely nonnegative we get

so we indeed have the claim.

We now prove the desired bound

Actually,

as we proved last time, so the bound guarantees the claim.4

In particular

by Fatou’s lemma. So almost surely. Therefore is harmonic on , and we already showed that if was small enough, thus

if was small enough. That implies on an open set and hence everywhere. Since

we can plug in and conclude that all moments of except the zeroth moment are zero. So is uniformly distributed on the unit circle.

By overspill, I think one can intuit that if is a random polynomial of high degree which has a zero close to , all zeroes in , and no critical point close to , then sort of looks like

where is a primitive root of unity of the same degree as . Therefore looks like a cyclotomic polynomial, and therefore should have lots of zeroes close to the unit sphere, in particular close to , a contradiction. This isn’t rigorous but gives some hint as to why this case might be bad.

Now one has

and in particular by Fatou’s lemma

But it was almost surely true that , thus that . So this enforces almost surely. In particular, almost surely,

Since is a contractible curve, its complement is connected. We recall that near infinity, and since we already know the distribution of , we can use it to compute near infinity. Tao says the computation of is a straightforward application of the Newtonian shell theorem; he’s not wrong but I figured I should write out the details.

For one has

where the denotes that this is a line integral in rather than in . Translating we get

which is the integral of the fundamental solution of the Laplace equation over . If (reasonable since is close to infinity), this implies the integrand is harmonic, so by the mean-value formula one has

and so this holds for both and near infinity. But then is harmonic away from , so that implies that

Since the distribution of is the Laplacian of one has

Therefore almost surely. In particular, is infinitesimal almost surely. This completes the proof in case one.

By the way, I now wonder if when one first learns PDE it would be instructive to think of the fundamental solution of the Laplace equation and the mean-value formulae as essentially a consequence of the classical laws of gravity. Of course the arrow of causation actually points the other way, but we are humans living in a physical world and so have a pretty intuitive understanding of what gravity does, while stuff like convolution kernels seem quite abstract.

Next time we’ll prove a contradiction for case zero, and maybe start on the proof for case one. The proof for case one looks really goddamn long, so I’ll probably skip or blackbox some of it, maybe some of the earlier lemmata, in the interest of my own time.