Lecture 4 – Hitting time theorem

I am aiming to write a short post about each lecture in my ongoing course on Random Graphs. Details and logistics for the course can be found here.

This lecture consisted of revision of the most relevant theory of Galton-Watson trees, with a focus on the case where the offspring distribution is Poisson, since, as we have seen in previous lectures, this is a strong candidate to approximate the structure of $G(n,\lambda/n)$. It makes sense to cover the theory of the trees before attempting to make rigorous the sense of approximation.

Given a Galton-Watson tree T, it is natural to label the vertices in a breadth-first order as $\varnothing=v_1,v_2,\ldots,v_{|T|}$. This is easiest if we have constructed the Galton-Watson tree as a subset of the infinite Ulam-Harris tree, where vertices have labels like (3,5,17,4), whose parent is (3,5,17). If this child vertex is part of the tree, then so are (3,5,17,1), (3,5,17,2), and (3,5,17,3). This means our breadth-first order is canonically well-defined, as we have a natural ordering of the children of each parent vertex.

Note: one advantage of using breadth-first order rather than depth-first order (which corresponds to the usual dictionary, or lexicographic ordering of the labels) is that if the tree is infinite, we don’t explore all of it during a depth-first search. (In the sense that there exist vertices which are never given a finite label.) For breadth-first search, a similar problem arises precisely when some vertex has infinitely many children. For a conventional Galton-Watson tree, the latter situation is much less of a problem than the infinite total population problem, which happens with positive probability whenever $\mu=\mathbb{E}[X]>1$.

Anyway, given the depth-first order, one can consider an exploration process $S_0,S_1,S_2,\ldots,S_{|T|}$ given by

$S_0=1,\quad S_i=S_{i-1}+(X_i-1),$

where $X_i$ is the number of children of $v_i$. In this way, we see that

$S_i=\big| \Gamma(v_1)\cup\ldots\cup\Gamma(v_i)\backslash \{v_1,\ldots,v_i\}\big|,\quad i\ge 1,$

records the number of vertices in some stack containing those which we have ‘seen but not explored’. Some authors prefer to start from 0, in which case one ends up with a similar but slightly different interpretation of the ‘stack’, but that’s fine since we aren’t going to define formally what ‘seen’ and ‘explored’ means in this post.

Essentially, we exhaust the vertices of the tree whenever $S_t=0$, and so the condition that $|T|=n$ requires

$S_n=0,\quad S_m\ge 1,\; m=0,1,\ldots,n-1.$

Conveniently, so long as we have avoiding ordering ambiguity, for example by insisting that trees live within the Ulam-Harris tree, we can reconstruct T uniquely from $(S_0,S_1,\ldots,S_{|T|})$.

Furthermore, if T is a Galton-Watson process, then the numbers of children $X_i$ are IID, and so in fact this exploration process is a random walk, and the size of the tree can be recovered as the hitting time of zero.

Note: making fully rigorous the argument that children in the GW tree are independent of the breadth-first walk fully rigorous is somewhat technical, and not to be dismissed lightly, though not of principle interest at the level of this topics course. See Proposition 1.5 in Section 1.2 of Le Gall’s notes or Section 1.2.2 of my doctoral thesis for further discussion and argument.

The hitting time theorem allows us to study the distribution of the hitting time of a random walk whose increments are bounded below by -1, in terms of the distribution of the value of the random walk.

Theorem: Let $(S_n,\, n\ge 0)$ be a random walk with $S_0=0$ and IID increments $(X_n,n\ge 1)$ satisfying $\mathbb{P}(X_n\ge -1)=1$. Let $H_{-k}=\inf \left\{n\,:\, S_n=-k\right\}$ be the hitting time of -k.

Then $\mathbb{P}\big( H_{-k}=n\big) = \frac{k}{n}\mathbb{P}\big(S_n=-k)$.

Commentary: there are local central limit theorem estimates and large deviation estimates that allow good control of the probability on the RHS for a rich class of contexts. So at a meta-level, the hitting time theorem allows us to reduce a complicated (though still classical) problem, to a real classical problem, which is particularly helpful when the LHS is a device for capturing relevant information about our random tree model.

Random Maps 2 – The Schaeffer Bijection

As indicated at the end of the previous post, our aim is to find a natural bijection between the set of pointed, rooted quadrangulations with n faces, and some set of objects based on decorating rooted plane trees with n edges in some fashion. Unlike our previous example, the construction of this bijection is definitely not trivial. It seems like a foolish ambition to explain this without several pictures, so I’m going to focus on some aspects of the analysis which I found challenging, rather than the construction itself.

Anyway, we don’t yet know what the extended set of trees should be. We need an extra factor of $3^n$, so it is natural to consider adding some sort of labelling of the tree, where for each non-root vertex in turn there are three options. So, given a rooted tree T, we label the vertices such that the root has label 0, and if a parent vertex has label k, any offspring has label k-1, k or k+1. Such a labelling is called admissable, and $\mathbb{T}_n$ is the set of rooted plane trees with n edges and an admissable labelling.

We now demonstrate how to construct an element of $\mathcal{Q}_n$ from an element of $\mathbb{T}_n$. Various authors had considered this problem to various extents, and so what follows is known as the Cori-Vauquelin-Schaeffer bijection, at least in this course.

Consider a contour exploration of the tree. That is, start out at the root and at all times take first-edge you encounter going clockwise from your current direction. When you arrive at a leaf, you will indeed therefore immediately retrace your most recent step. The key property is that you traverse each edge exactly twice, and so we may think of the tree as having 2n oriented edges. It is more useful to think about corners. A corner is the directed arc (WLOG clockwise) between adjacent edges at a vertex. There is a natural bijection between corners and directed edges, by looking anti-clockwise from the tail of the edge. So the contour process explored the directed edges in some order, and hence explores the corners of the tree. One thing I found confusing initially was switching between considering vertices and corners. I feel in retrospect that the only reason we need the vertices themselves is to induce the labelling onto the corners. These are the only thing we will use in the construction.

As we trace out the contour process, naturally we see different labels. We define the successor of a corner with label k to be the next corner seen in the contour process (taken modulo 2n if necessary) with label k-1. Note that any corner on a vertex with minimal label will not have a successor. To counter this, we add a new vertex, suggestively called $v_*$, with a single corner (ie no edges yet) and denote this corner to be the successor of the corners in the original tree with minimal label.

To construct our quadrangulation, we simply join up every corner with its successor corner. Note that if you are thinking of the successor of a corner as a vertex (rather than as a corner) you will get in trouble here, as it might be several ways to draw this arc.

The red arcs and vertex v* are added to form the quadrangulation. Note the blue angles indicate the three corners around the vertex labelled -1.

It is not obvious that it is possible to do this so that the arcs do not overlap. However, by considering the label process as you explore via the contour process, it becomes clear that you can discount the possibility of any overlaps one by one. This applies equally to pairs of new arcs overlapping, as well as new arcs overlapping with edges of the original tree. In any case, we remove the edges of the original tree to obtain the quadrangulation.

Note that when you move from any corner of a vertex with label k to its successor, then to the successor of its successor and so on, the labels are decreasing, so eventually you must end up at a corner with minimal label, and hence at $v_*$. We conclude that the graph of arcs is connected. It remains to show that it is a quadrangulation.

This is rather fiddly to do without a diagram. Note first that whenever we have a directed edge in the tree going from label k to label k-1, then this edge essentially becomes an arc of the quadrangulation. We show that the edge oriented in the other direction, called say e, induces three further arcs of a quadrangle. So e goes from label k-1 to k. Consider the corners before and following e in the contour exploration, which is a corner around the vertex with label k. The successor of the corner after e is a corner with label k-1, and this has a successor with label k-2. By construction, this must also be the successor of the corner before e. Why? Well as we traverse the contour beyond e, the first appearance of label k-1 must happen before the first appearance of label k-2, as the increments can only be in {-1,0,1}. This gives us the three further arcs. Note also that the 2-colouring of the quadrangulation is given by the parity of the tree-labelling.

I was bothered about what happens if two vertices with label k-1 are in fact the same. This would happen if, for example, the vertex labelled k is a leaf. Then, at least two of the corners around the single vertex with label k-1 have the same corner as successor. A naïve attempt at drawing the resulting arcs did not give a quadrangle. The key observation is that you have to draw the arcs in the direction of the contour process. So in this case, the arc from the corner before edge e will loop all the way around the vertex with label k, so it contains the other two relevant arcs on its way to the vertex with label k-2, giving us the ‘pacman’ quadrangle discussed earlier.

The other case we have to check is when our base edge joins two vertices with label k. Then the other two vertices of the face will have label k-1. This is similar to the above, and slightly easier.

As a preliminary to checking that we can invert this construction, we observe that the vertices of the quadrangulation are the vertices of the original tree plus $v_*$, and furthermore, the labels in the tree are given by the graph distance from $v_*$ in the quadrangulation, with a constant added uniformly so that the root vertex has label 0.

At this point, we observe that in the construction, we didn’t specify how to choose the rooted edge of the quadrangulation. Canonically, we take it to be the arc between the first corner of the root in the contour process, and its successor. However, we can orient it in either direction, giving us the extra factor of 2 we were looking for.

Returning to the inverse, it is clear what to do when we see a quadrangle corresponding to the second case above – namely put an edge between the two vertices with label k. In the case where the face has labels {k,k-1,k-1,k-2} it is less obvious. Note though that by starting at the first corner of the root, which is identified by the rooted edge in the quadrangulation, we can recover the contour process from the arcs of the quadrangulation, and the labels. So when we see such a face, we can use this information to choose which of the (k-1)-labelled vertices to join to the vertex with label k.

Anyway, now we are convinced that this bijection works, the next stage is to apply it to gain extra information about a uniformly-chosen large quadrangulation. We can view the vertices as being those of a large uniform plane tree, and the labels as given by a random walk along this large tree. We might expect to see this labelling structure converge to something that looks like Brownian motion indexed by a Brownian continuum random tree, in a sense to be made more precise. And the labelling is not merely a decoration in the quadrangulation, since it specifies the distance to the identified point $v_*$. In particular, this gives a bound on the distance between any two vertices in the quadrangulation, eg two vertices chosen uniformly at random. In fact, by looking more carefully at the scaling limit of the uniform tree’s contour process, we can say rather more than that.

Long Paths and Expanders

I’m in Birmingham this week for the LMS-EPSRC summer school on Random Graphs, Geometry and Asymptotic Structure. The event consists of three five-hour mini-courses, a plenary lecture, leaving plenty of time for problem sheet and discussion. I thought it would be worth trying to say a couple of interesting things each day – I do not know whether this will succeed, but I might as well try.

Today, a few thoughts on the first two lectures of Michael Krivelevich’s course on Long Paths and Hamiltonicity in Random Graphs. The aim is to develop tools to investigate the threshold for the presence of a Hamiltonian cycle in G(n,p). In this first part of the course, we were mainly thinking about long paths.

One tool we used a lot was the Depth-First Search algorithm. This is very similar to the exploration process I’ve talked about before. Essentially, here we consider trying to explore the graph in a depth-first way, but instead of viewing all the edges incident to a vertex we have just arrived at, we only look to see whether there is an edge out of the new vertex. If there is, we explore it, then come back eventually to look for more. It really comes down to a difference in the information we are storing. In this DFS, we store the vertices which we haven’t finished exploring, which is the set of vertices on the explored path between the root and the current vertex. So the size of this set evolves like the contour process. In particular, we can read off the sizes of paths from this description. These dynamics are useful in particular because we know there are no edges between the set of vertices we have finished exploring, and the ones we have yet to explore. The stack of ‘processing’ vertices must glue everything else together.

We can translate one of the arguments back into the language for the old exploration process. Recall the increments of the exploration process are $\mathrm{Bin}(\alpha n,\frac{c}{n}) -1$ once we have explored $\alpha n$ vertices. We don’t need to worry about the -1 bit for now. Observe that because we are exploring in a depth-first way, if a subsequence of the Binomial variables of length k are all positive, this corresponds to a path of length (k-1).

So to prove, for example, that the longest path in a subcritical random graph is O(log n), it suffices to prove that there are O(log n) consecutive positive entries in the sequence of n binomial entries. Since the distribution changes continuously, it is convenient to prove that there are O(log n) consecutive positive entries in the first $\epsilon n$ binomial entries. The probability that any of these entries is positive is bounded below by some p, so it suffices to consider instead a sequence of Bernoulli RVs with parameter p. So if we never have clog n consecutive, this gives control of the sequence of geometric random variables corresponding to the gaps between 0s in the sequence. Precisely, these are Geom(q), and we must have $\frac{\epsilon n}{c\log n}$ of them independently being less than clog n. We have to chase a few constants, and use the fact that if $f(n)\rightarrow\infty, \frac{g(n)}{f(n)}\rightarrow\infty$, then

$(1-\frac{1}{f(n)})^{g(n)}\rightarrow 0,$

by comparison with the standard asymptotic result for $e^{-x}$. In any case, we get that this probability tends to 0 if we choose c small enough, and so with high probability there is a path of length clog n.

This is interesting, because we knew already that the largest component in a subcritical random graph had size O(log n). But we also knew that all the components were trees, or ‘almost trees’, and were uniformly chosen from the set of trees (or trees + an edge or two) with appropriate size. And the largest path in a UST on n vertices is $O(n^{1/2})$ with high probability. So we learn that there are enough components of size $\geq c\log n$ that it is actually very probable that one of them will have the unlikely property of being much more path-like than a typical tree.

Krivelevich also showed a pleasant elementary proof of the result that a supercritical random graph has a path of length O(n), using a similar idea.

The other definition of major interest was an expander graph. Often when doing calculations about neighbourhoods of sets of vertices, we run into the problem that the neighbourhoods may overlap, and so we cannot get the total outer neighbourhood (or outer boundary) just by summing over the individual neighbourhood sizes. In an expander graph, we demand that all small sets of vertices have neighbourhood at least as large as some constant multiple of the set size, essentially giving us a bound on the above problem. Concretely, G is a $(k,\alpha)$-expander is for any set of vertices $|U|\leq k, |N(U)|\geq \alpha |U|$.

There’s a very nice argument using Posa’s lemma, where we consider all the possible ways to rearrange the vertices in some longest path into a different longest path, and then focus on the endpoints of all these paths. With this so-called rotation-extension technique, we can show that a (k,2)-expander has a path of length at least 3k-1.

There are structural similarities between expander graphs and regular graphs, so it seems natural that there will be some interesting spectral properties. I don’t know much about this, but perhaps it will come up later in the week. But, returning to the random graph long path problem, it now suffices to show subcritical G(n,p) is a (clog n,2)-expander for some c. Expander properties are in some sense the opposite of clustering properties, and independence of a RG inhibit most clustering properties (as discussed in much greater detail in some of the posts about network models). Unfortunately, this doesn’t actually work, as in a subcritical graph, the typical expansion coefficient, even of a small set will be c, for G(n,c/n), which is not large enough. However, if you chose the constants carefully, such an argument should work for c>2, so long as you chose k=an, with a small enough that the probability of a vertex elsewhere in the graph being joined to (at least) two of the k vertices in the set, was small compared with (c-2).

REFERENCES

The course notes are not available, though chapter 3 from these 2010 notes by the same lecturer are related and interesting.