Abstract
We prove bulk scaling limits and fluctuation scaling limits for a two-parameter class ALE\(({\alpha },\eta )\) of continuum planar aggregation models. The class includes regularized versions of the Hastings–Levitov family HL\(({\alpha })\) and continuum versions of the family of dielectric-breakdown models, where the local attachment intensity for new particles is specified as a negative power \(-\eta \) of the density of arc length with respect to harmonic measure. The limit dynamics follow solutions of a certain Loewner–Kufarev equation, where the driving measure is made to depend on the solution and on the parameter \({\zeta }={\alpha }+\eta \). Our results are subject to a subcriticality condition \({\zeta }\leqslant 1\): this includes HL\(({\alpha })\) for \({\alpha }\leqslant 1\) and also the case \({\alpha }=2,\eta =-1\) corresponding to a continuum Eden model. Hastings and Levitov predicted a change in behaviour for HL\(({\alpha })\) at \({\alpha }=1\), consistent with our results. In the regularized regime considered, the fluctuations around the scaling limit are shown to be Gaussian, with independent Ornstein–Uhlenbeck processes driving each Fourier mode, which are seen to be stable if and only if \({\zeta }\leqslant 1\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
1.1 Hastings–Levitov aggregation
In many physical contexts there appear clusters whose shape is complex, formed apparently by some mechanism of random growth. It has long been a challenge to account for the observed variety of complex cluster shapes, starting from plausible physical principles governing the aggregation of individual microscopic particles. For clusters which are essentially two-dimensional, there is an approach introduced by Carleson and Makarov [4] and Hastings and Levitov [10], in which clusters are encoded as a composition of conformal maps, one for each particle. In this approach, a growing cluster is modelled by an increasing sequence of compact sets \(K_n\subseteq \mathbb {C}\) which are assumed to be simply connected. We will take the initial set \(K_0\) to be the closed unit disk \(\{|z|\leqslant 1\}\). The increments \(K_n\setminus K_{n-1}\) are then thought of as a sequence of particles added to the cluster. The idea is to study the clusters \(K_n\) via the conformal isomorphisms
where \(D_n\) is the complementary domain \(\mathbb {C}\setminus K_n\) and \(\Phi _n\) is normalized by \(\Phi _n(\infty )=\infty \) and \(\Phi _n'(\infty )>0\). Then \(\Phi _0(z)=z\) for all z and \(K_n\) has logarithmic capacity \(\Phi _n'(\infty )>1\) for all \(n\geqslant 1\). This formulation is convenient because the harmonic measure from \(\infty \) on the boundary \(\partial D_n\), which provides a natural way to choose the location of the next particle, is then simply the image under \(\Phi _n\) of the uniform distribution on \(\partial D_0=\{|z|=1\}\). Having chosen a random angle \({\Theta }_{n+1}\) to locate the next particle, and a model particle \(P_{n+1}\) attached to \(K_0\) at \(e^{i{\Theta }_{n+1}}\), for example a small disk tangent to \(K_0\), the cluster map is updated to
where \(F_{n+1}\) is the conformal isomorphism \(D_0\rightarrow D_0\setminus P_{n+1}\), normalized similarly to \(\Phi _n\). Then \(\Phi _{n+1}\) encodes the cluster
Thus, once we specify distributions for the angles \({\Theta }_n\) and model particles \(P_n\), we have specified a mechanism to grow a random cluster.
We will write
and we will refer to \({{\text {cap}}}(K_n)\) as the capacityFootnote 1 of \(K_n\) and \(c_n\) as the capacity of \(P_n\). Then
We will be looking for scaling limits where the particle capacities \(c_n\) and the associated particles \(P_n\) become small, but where n is chosen sufficiently large that the cluster capacities \({{\text {cap}}}(K_n)\) grow macroscopically.
A simple case is to choose \({\Theta }_{n+1}\) uniformly distributed on the unit circle and to take \(P_{n+1}=e^{i{\Theta }_{n+1}}P\), where P is a small disk tangent to the unit disk at 1, of radius r(c), chosen so that P has capacity c. Then in fact \(r(c)/\sqrt{c}\) has a positive limit as \(c\rightarrow 0\). The location of the new particle \(\Phi _n(P_{n+1})\) is then distributed according to harmonic measure on \(\partial K_n\). However, if we assume that \(\partial K_n\) is approximately linear on the scale of P, then we would have
so we would add an approximate disk of diameter proportional to \(\sqrt{c}|\Phi _n'(e^{i{\Theta }_{n+1}})|\).
In order to compensate for this distortion, Hastings and Levitov proposed the HL\(({\alpha })\) family of models where, once \({\Theta }_{n+1}\) is chosen, we choose \(P_{n+1}\) to be a particle of capacity
Then, in the case \({\alpha }=2\), the particles added to the cluster would be approximately of constant size. The approximation (2) is in fact misleading, at least on a microscopic level, because \(\partial K_n\) develops inhomogeneities on the scale of the particles. Nevertheless, HL(2) has been considered as a variant of diffusion-limited aggregation (DLA) [28], with some justification, see [10], derived from numerical experiments.
In general, the HL\(({\alpha })\) model offers a convenient mechanism for such experiments, and moves away from the lattice formulation of [28] which has been shown to lead to unphysical effects on large scales (see for example [8]). Moreover, it might be hoped that an evolving family of conformal maps would present a more tractable framework for the analysis of scaling limits than other growth models, while potentially sharing the same bulk scaling limit and fluctuation universality class. That is the direction explored in this paper.
Besides the mechanism of diffusive aggregation, based on harmonic measure, there is another one-parameter family of models, conceived originally in the lattice case, called dielectric breakdown models [20], which interpolates between DLA and the Eden model [7]. In the Eden model, each boundary site is chosen with equal probability. In the continuum setting, for an Eden-type model we would choose an attachment point on the boundary according to normalized arc length, which has density proportional to \(|\Phi _n'(e^{i{\theta }})|\) with respect to harmonic measure. We can widen our family of models to include a continuum analogue of dielectric breakdown models (DBM) by choosing
The case \(\eta =-1\) then provides a continuum variant of the Eden model.
In a law-of-large-numbers regime, it might be guessed that bulk characteristics of the cluster for the model incorporating both the \({\alpha }\) and \(\eta \) modifications would depend only on their sum \({\zeta }={\alpha }+\eta \) since, once this is fixed, up to a global time-scaling, the growth rate of capacity due to particles attached near \(e^{i{\theta }}\) does not depend further on \({\alpha }\) or \(\eta \). We will show, in the regime which we can address, that this is indeed true.
In this paper we investigate the two-parameter family of models just described, but modified by the introduction of a regularization parameter \({\sigma }>0\), which controls the minimum length scale over which feedback occurs through \(c_{n+1}\) and \({\Theta }_{n+1}\). Specifically, we require
This model was introduced in [27] as the (discrete-time) aggregate Loewner evolution model. We will require throughout that \({\sigma }\gg \sqrt{c}\) (and sometimes more) and we will restrict attention to the subcritical regime \({\zeta }\leqslant 1\). This includes the Eden case (\({\alpha }=2\), \(\eta =-1\)) but excludes continuum DLA (\({\alpha }=2\), \(\eta =0\)). In the regularized models, we will show fluctuation behaviour which is universal over all choices of particle family. Our first main result shows that, in this regime, in the limit \(c\rightarrow 0\), disks are stable, that is, an initial disk cluster remains close to a disk as particles are added and its capacity becomes large. Our second main result is to prove convergence of the normalized fluctuations of the cluster around its deterministic limit, to an explicit Gaussian process. The constraint \({\zeta }\leqslant 1\) appears sharp for this behaviour: we see an explicit dependence of the fluctuations on \({\alpha }\) and \(\eta \) and, in particular, an exponential instability of rate \(({\zeta }-1)k\) in the kth Fourier mode if we formally take \({\zeta }>1\).
1.2 Statement of results
In this section, we define the continuous-time ALE\(({\alpha },\eta )\) model, which is our object of study, and we specify our standing assumptions for individual particles. We then state our main results.
Our model is constructed as a composition of univalent functions on the exterior unit disk \(D_0=\{|z|>1\}\). Each of these functions corresponds to a choice of attachment angle \({\theta }\in [0,2\pi )\) and a basic particle P. Recall that \(K_0=\{|z|\leqslant 1\}\). By a basic particle P we mean a non-empty subset of \(D_0\) such that \(K_0\cup P\) is compact and simply connected. Set \(D=D_0\setminus P\). By the Riemann mapping theorem, there is a \(c\in (0,\infty )\) and a conformal isomorphism \(F:D_0\rightarrow D\) with Laurent expansion of the form
Then F is uniquely determined by P, and P has capacity c. Our model depends on three parameters \({\alpha },\eta \in \mathbb {R}\) and \({\sigma }\in (0,\infty )\), along with the choice of a family of basic particles \((P^{(c)}:c\in (0,\infty ))\) with \(P^{(c)}\) of capacity c. The associated maps \(F_c:D_0\rightarrow D^{(c)}\) then have the form (4) with \(a_k=a_k(c)\) for all k. We assume throughout that \(F_c\) extends continuously to \(\{|z|\geqslant 1\}\). We require that our particle family is nested
and satisfies, for some \({\Lambda }\in [1,\infty )\),
where
In our results, only small values of c are of interest. For such c, the last condition (6) forces our particles \(P^{(c)}\) to concentrate near the point 1 while never becoming too flat against the unit circle.
The following are all examples of particle families satisfying both conditions (5) and (6):
and
where in each case \({\delta }\) is a suitable increasing homeomorphism of \((0,\infty )\).
It will be convenient to place our aggregation models from the outset in continuous time. By a (continuous-time) aggregate Loewner evolution of parameters \({\alpha },\eta \in \mathbb {R}\), or ALE\(({\alpha },\eta )\), we mean a finite-rate, continuous-time Markov chain \((\Phi _t)_{t\geqslant 0}\) taking values in the set of univalent functions \(D_0\rightarrow D_0\), starting from \(\Phi _0(z)=z\), which, when in state \(\phi \), jumps to \(\phi \circ F_{c({\theta },\phi ),{\theta }}\) at rate \({\lambda }({\theta },\phi )d{\theta }/(2\pi )\), where
Since \({\sigma }>0\), the rate \({\lambda }({\theta },\phi )\) is continuous in \({\theta }\), so the total jump rate is finite. The model may be thought of equivalently in term of the random process of compact sets \((K_t)_{t\geqslant 0}\) given by
The effect of the jump just described is then to add to the current cluster the set \(\phi (e^{i{\theta }}P^{(c({\theta },\phi ))})\) thereby increasing its capacity by \(c({\theta },\phi )\).
An explicit realisation of this Markov chain can be constructed as follows. Given a univalent function \(\phi : D_0\rightarrow D_0\), define the normalising constant
Starting from \(\Phi _0(z) = z\), suppose that a realisation of \((\Phi _s)_{0 \leqslant s \leqslant t}\) has been constructed up to some \(t \geqslant 0\), and that \(\Phi _t = \phi \). Sample independently a random time \(T \sim \textrm{Exp}(c^{-1} Z_\phi / (2 \pi ))\) and random angle \(\Theta \) with density function \(|\phi '(e^{\sigma +i\theta })|^{-\eta }/ Z_\phi \). Then set \(\Phi _s = \phi \) for \(t< s < t+T\), and \(\Phi _{t+T}=\phi \circ F_{c({\Theta },\phi ),{\Theta }}\). It is straightforward to verify that this construction gives a Markov chain with distribution corresponding to the specification above.
Denote the jump times of the Markov chain by \(T_k\), \(k=1, 2, \dots \). By the explicit construction,
where \(F_n=F_{C_n,{\Theta }_n}\), for capacity \(C_n\) and attachment angle \({\Theta }_n\) satisfying
Therefore, if \(T_n \leqslant t < T_{n+1}\), we have
as in Fig. 1. Moreover, the capacity \({{\mathcal {T}}}_t\) of the cluster \(K_t\) is then given by
For certain parameter values, the process \((\Phi _t)_{t\geqslant 0}\) may explode, that is, may take infinitely many jumps in a finite time interval. We will show in Proposition A.1 that explosion occurs if and only if both \(\eta <0\) and \({\zeta }={\alpha }+\eta <0\), and in this case we also have \({{\mathcal {T}}}_t\rightarrow \infty \) at the explosion time. This phenomenon is however irrelevant to our main results on scaling limits, since explosion is excluded by these results (with high probability) over the relevant time interval. Hence we will make no attempt to define \(\Phi _t\) beyond explosion.
By reference to (1) and (3), it is immediate that the jump-chain \((\Phi _{T_n})_{n\geqslant 0}\) is exactly the discrete-time aggregate Loewner evolution process \((\Phi _n)_{n\geqslant 0}\) in the introductory discussion. In particular, in the case \(\eta =\sigma =0\), \((\Phi _t)_{t\geqslant 0}\) is the original Hastings–Levitov process embedded in continuous time as a Poisson process with jumps of rate \(c^{-1}\). For clarity, from now on we denote the discrete-time process by \((\Phi ^{{\text {disc}}}_n)_{n\geqslant 0}\). Prior work on ALE models [21, 27] was framed in terms of this discrete-time process. The continuous-time framework allows a more local specification of the dynamics, without the need to normalise the distribution of attachment angles. It further allows us to organise the computation of martingales in terms of a standard calculus for Poisson random measures.
We can now state our first main result. Define
and for \(t<t_{\zeta }\) set
Note that \({\tau }_t\rightarrow \infty \) as \(t\rightarrow t_{\zeta }\) for all \({\zeta }\).
The result identifies the small-particle scaling limit of \(K_t\) in the case \({\zeta }\leqslant 1\) as a disk of radius \(e^{{\tau }_t}\), with quantified error estimates. It is proved in Proposition 5.7. The range of parameter values to which the result applies is indicated by the region shaded red in Fig. 2, with diagonal lines showing parameter pairs \(({\alpha },\eta )\) sharing a common bulk scaling limit. Recall that \({{\mathcal {T}}}_t=\log \Phi _t'(\infty )\), which is the capacity of \(K_t\), and set
Theorem 1.1
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), for all \({\varepsilon }\in (0,1/3]\) and \(\nu \in (0,{\varepsilon }/4]\), for all \(m\in \mathbb {N}\) and \(T\in [0,t_{\zeta })\), there is a constant \(C=C({\alpha },\eta ,{\Lambda },{\varepsilon },\nu ,m,T)<\infty \) with the following property. In the case \({\zeta }<1\), for all \(c\leqslant 1/C\) and all \({\sigma }\geqslant c^{1/2-{\varepsilon }}\), with probability exceeding \(1-c^m\), for all \(t\leqslant T\),
and, for all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
Moreover, in the case \({\zeta }=1\), for all \(c\leqslant 1/C\) and all \({\sigma }\geqslant c^{1/3-{\varepsilon }}\), with probability exceeding \(1-c^m\), for all \(t\leqslant T\),
and, for all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
We will show a similar result for the discrete-time process \((\Phi _n^{{\text {disc}}})_{n\geqslant 0}\). Set
Define
and for \(n<n_{\alpha }/c\) set
The following result is proved at the end of Sect. 5.2. The case \({\alpha }=0\) is Theorem 1.1 in [21] but with an improvement to the constraints on r and \(\sigma \), and the corresponding upper bound, in the \(\eta =1\) case.
Theorem 1.2
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), for all \({\varepsilon }\in (0,1/3]\) and \(\nu \in (0,{\varepsilon }/4]\), for all \(m\in \mathbb {N}\) and \(N\in [0,n_{\alpha })\), not necessarily an integer, there is a constant \(C=C({\alpha },\eta ,{\Lambda },{\varepsilon },\nu ,m,N)<\infty \) with the following property. In the case \({\zeta }<1\), for all \(c\leqslant 1/C\) and all \({\sigma }\geqslant c^{1/2-{\varepsilon }}\), with probability exceeding \(1-c^m\), for all \(n\leqslant N/c\),
and, for all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
Moreover, in the case \({\zeta }=1\), for all \(c\leqslant 1/C\) and all \({\sigma }\geqslant c^{1/3-{\varepsilon }}\), with probability exceeding \(1-c^m\), for all \(n\leqslant N/c\),
and, for all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
We turn to our second main result, which describes the limiting fluctuations of ALE\(({\alpha },\eta )\). Denote by \({{\mathcal {H}}}\) the set of all holomorphic functions on \(D_0=\{|z|>1\}\) which are bounded at \(\infty \). We equip \({{\mathcal {H}}}\) with the metric
Then \({{\mathcal {H}}}\) is a complete separable metric space. Define for \(t<t_{\zeta }\)
Let \((B_t)_{t\geqslant 0}\) be a (real) Brownian motion. Let \((B_t(k))_{t\geqslant 0}\) for \(k\geqslant 0\) be a sequence of independent complex Brownian motions, independent of \((B_t)_{t\geqslant 0}\). We can define continuous Gaussian processes \(({\Gamma }_t(k))_{t<t_{\zeta }}\) and \(({\Gamma }^{{\text {cap}}}_t)_{t<t_{\zeta }}\) by the following Ornstein–Uhlenbeck-type stochastic differential equations
We show in Sect. 6.2 that the series
converges in \({{\mathcal {H}}}\), uniformly on compacts in \([0,t_{\zeta })\), almost surely. In fact \(({\hat{{\Gamma }}}_t)_{t<t_{\zeta }}\) satisfies the following stochastic differential equation in \({{\mathcal {H}}}\)
where \(Q_0f(z)=-(1-{\zeta })zf'(z)\) and
The following two results are proved in Sect. 6.
Theorem 1.3
Assume that \({\zeta }={\alpha }+\eta \in (-\infty ,1]\). Fix \(T\in [0,t_{\zeta })\) and \({\varepsilon }>0\) and consider the limit \(c\rightarrow 0\) with \({\sigma }\rightarrow 0\) subject to the constraint
Then
weakly in the Skorokhod space \(D([0,T],{{\mathcal {H}}}\times \mathbb {R})\).
As in the bulk scaling limit, we can deduce an analogous discrete-time fluctuation theorem. The case \({\alpha }=0\) recovers Theorem 1.2 in [21]. Define for \(t\geqslant 0\)
We have seen already in Theorem 1.2, for \(N<n_{\alpha }\), that \(({{\mathcal {T}}}^{{\text {disc}}}_n-{\tau }^{{\text {disc}}}_n)_{n\leqslant N/c}\) does not fluctuate at scale \(\sqrt{c}\). We can define a continuous Gaussian process \(({\hat{{\Gamma }}}^{{\text {disc}}}_t)_{t<n_{\alpha }}\) in \({{\mathcal {H}}}\) by
Theorem 1.4
Assume that \({\zeta }={\alpha }+\eta \in (-\infty ,1]\). Fix \(N\in [0,n_{\alpha })\), not necessarily an integer, and fix \({\varepsilon }>0\). In the limit \(c\rightarrow 0\) with \({\sigma }\rightarrow 0\) considered in Theorem 1.3, we have
weakly in \(D([0,N],{{\mathcal {H}}})\).
1.3 Commentary and review of related work
Hastings and Levitov [10] introduced the family of planar aggregation models HL(\({\alpha }\)), which are the cases \(\eta ={\sigma }=0\) of our ALE\(({\alpha },\eta )\) model. They discovered by numerical experiments that, for small particles, the models underwent a transition at \({\alpha }=1\): for \({\alpha }\leqslant 1\) the cluster grows like a disk, while for \({\alpha }>1\) it exhibits fractal properties. There are two natural scaling-limit regimes under which mathematical results have been established: capacity rescaling and the small-particle limit. Under capacity-rescaling, the particle capacity parameter c is kept fixed, and the cluster is rescaled to have logarithmic capacity 1, before the limit is taken as the number of particles goes to infinity. This corresponds to studying the limit of the map \(\hat{\Phi }_n^{{{\text {disc}}}}(z)\) as \(n \rightarrow \infty \). Under the small-particle limit, the parameter \(c \rightarrow 0\), but the rate at which particles arrive is increased to ensure a non-trivial limit. This is the regime followed in the present paper, and corresponds to studying the limit of the process \((\Phi ^{{\text {disc}}}_{n(t)}(z))_{t \geqslant 0}\) as \(c \rightarrow 0\), where n(t) is a suitable embedding of arrival times into continuous time. In most results to date, the embedding \(n(t) = \lfloor t/c \rfloor \) has been used.
The HL(0) model is the most mathematically tractable model in the Hastings–Levitov family as in this case the particle maps, \(F_n\), are i.i.d. It has been investigated rigorously in a series of works [24] (existence of a bulk scaling limit under capacity rescaling), [22] (bulk small-particle scaling limit), [26] (fluctuation small-particle scaling limit). Several variants exist, for example [1, 2] (versions of HL(0) grown in the upper half-plane) and [14, 17] (anisotropic versions of HL(0)). The \({\sigma }\)-regularized variant of HL\(({\alpha })\) was proposed in [15], where it was shown for slit maps that, if \({\sigma }\gg (\log (1/c))^{-1/2}\), there is disk-like behaviour in the small-particle limit for all \({\alpha }\geqslant 0\): it appeared that the observed fractal properties of HL\(({\alpha })\) for \({\alpha }>1\) were suppressed by strong regularization. In contrast, for the weaker regularization used in the present paper, the conjectured phase transition at \({\alpha }=1\) (or \({\zeta }=1\)) becomes visible at the level of fluctuations. The method of [15] used a comparison with an HL(0)-type model which breaks down for smaller values of \({\sigma }\). Regularized versions of HL\((\alpha )\) under capacity-rescaling are considered in [24] (estimates for the dimension) and [16] (fluctuation limit when \(0< \alpha < 2\) and \(\sigma = \infty \)).
The regularized ALE\(({\alpha },\eta )\) model first appeared in [27] where it was shown that, for slit maps, if \(\alpha \geqslant 0\) and \(\eta > 1 \), \({\sigma }\)-regularized ALE\((\alpha ,\eta )\) converges to a growing slit in the small-particle limit, provided \(\sigma \rightarrow 0\) sufficiently fast as \(c\rightarrow 0\). This result is a consequence of the singularities of the derivative of the slit map on the cluster boundary, which causes the cluster growth to concentrate at the tips of particles. Similar degeneracies are exploited in two recent papers [12, 13]. In [12] it was shown that, when \(\eta < -2\), ALE\((0,\eta )\) converges to a \(\text {SLE}_4\) curve. It is conjectured that, by making appropriate choices of particle shape, one can get convergence to \(\text {SLE}_\kappa \) for any \(\kappa \geqslant 4\). In [13], it is shown that ALE\((0, \eta )\), initiated from a needle-like configuration, converges to a Laplacian-path model [5]. Another model that fits into this framework is Quantum Loewner Evolution (QLE) [19]. The paper [27] contains a comprehensive discussion of connections between these and related models, so we do not repeat this here.
A new approach was begun in [21], treating regularized ALE\((0,\eta )\) as a Markov chain in univalent functions. By martingale arguments, a bulk small-particle scaling limit and fluctuation scaling limit were shown, subject to the constraint \(\eta \leqslant 1\) and to restrictions on \({\sigma }\) as a fractional power of c. These limits (in contrast to those above) turn out not to depend on the details of individual particle shapes. In this paper, we extend the analysis of [21] to ALE\(({\alpha },\eta )\), subject now to the constraint \({\zeta }={\alpha }+\eta \leqslant 1\). Thus we now include regularized HL\(({\alpha })\) for \({\alpha }\leqslant 1\). Hastings and Levitov had argued that there should be a trade-off between \({\alpha }\) and \(\eta \), with only \({\zeta }\) affecting the bulk scaling limit, and on this basis proposed HL(1), that is ALE(1, 0), as a continuum variant of the Eden model. A more direct continuum analogue of the Eden model is ALE\((2,-1)\). Our results, in the regularized case, both justify the trade-off argument and show a disk scaling limit whenever \({\zeta }\leqslant 1\). On the other hand, we show that ALE(1, 0) and ALE\((2,-1)\) have different fluctuation behaviour. As in [21], the behaviour of fluctuations as a function of \({\zeta }\) is consistent with the conjectured transition in behaviour at \({\zeta }=1\). We emphasise that scaling limits for the conjectured supercritical regime \(\zeta >1\) lie outside the scope of the present paper.
Hastings and Levitov [10] identify a Loewner–Kufarev-type equation, which they propose as governing the small-particle limit of HL\(({\alpha })\), citing a discussion of Shraiman and Bensimon [25] for the Hele–Shaw flow, where \({\alpha }\) is taken to be 2. This is the LK\(({\alpha })\) equation, which is the subject of the next section. As noted by Sola in a contribution to [18], there is a lack of mathematical theory for the LK\(({\alpha })\) equation, except in the case \({\alpha }=2\) when some special techniques become available (we refer the reader to [11] and to the monograph [9] which contains an extensive list of references). In this paper, since our focus is on clusters initiated as a disk, we are able to use an explicit solution of the equation, along with its linearization around that solution, so we do not rely on a general theory. However, the particle interpretation established here offers some evidence that for \({\alpha }\leqslant 1\), the LK\(({\alpha })\) equation may have a suitable existence, uniqueness and stability theory, and that it may be possible to derive the equation as a limit of particle models.
Our results depend on constraints on the regularization parameter \({\sigma }\), though substantially weaker ones than those used in [15]. These constraints limit the interactions of individual particles and place us in the simplest case of Gaussian fluctuations. At a technical level, for Theorem 1.1, these constraints come from the need to have \({\bar{{\delta }}}(e^{\sigma })\leqslant c^{\varepsilon }\) in Proposition 5.3, while for Theorem 1.3 they are needed to show that the Poisson integral process \((\Pi _t)_{t\geqslant 0}\) is a good approximation to the fluctuations in Proposition 5.7. In the case \({\zeta }=1\), the regularizing operator Q obtained by linearization of the LK\(({\zeta })\) equation collapses from a fixed multiple of the Cauchy operator to \({\sigma }\) times the second derivative. In general, for scaling regimes where \({\sigma }\rightarrow 0\) faster than our fluctuation results allow, it remains possible that ALE\(({\alpha },\eta )\) has different universal fluctuation behaviour, such as KPZ, as has been conjectured for the lattice Eden model.
1.4 Structure of the paper
In the next section, we discuss the Loewner–Kufarev equation for the limit dynamics. Then, in Sect. 3, we derive an interpolation formula between ALE\(({\alpha },\eta )\) and solutions of the limit equation. The terms in this formula are estimated in Sect. 4. Equipped with these estimates, we show the bulk scaling limit in Sect. 5 and the fluctuation scaling limit in Sect. 6. We collect in Appendix A some further estimates needed in the course of the paper, including estimates on the conformal maps which encode single particles and particle families.
2 Loewner–Kufarev Equation
Let \({{\mathcal {S}}}\) denote the set of univalent holomorphic functions \(\phi \) on \(\{|z|>1\}\) with \(\phi (\infty )=\infty \) and \(\phi '(\infty )\in (0,\infty )\). Then each \(\phi \in {{\mathcal {S}}}\) has the form
for some \(c\in \mathbb {R}\) and some sequence \((a_k:k\geqslant 0)\) in \(\mathbb {C}\). Fix parameters \({\zeta }\in \mathbb {R}\) and \({\sigma }\geqslant 0\). Given \(\phi _0\in {{\mathcal {S}}}\), consider the following Cauchy problem for \((\phi _t)_{t\geqslant 0}\) in \({{\mathcal {S}}}\)
where
The case \({\sigma }=0\) of this equation is the equation proposed by Hastings and Levitov as scaling limit for HL\(({\zeta })\), which we will call the LK\(({\zeta })\) equation. When \({\zeta }=0\), the value of \({\sigma }\) is immaterial and there is a unique solution given by
When \({\zeta }=2\) and \({\sigma }=0\), (9) is the Loewner–Kufarev equation associated to the Hele–Shaw flow. For \({\sigma }>0\), we will refer to (9) as the \({\sigma }\)-regularized LK\(({\zeta })\) equation. We will be interested in the subcritical case \({\zeta }\in (-\infty ,1]\).
The general form of the Loewner–Kufarev equation is given by
with \((\mu _t:t\geqslant 0)\) a given family of measures on \([0,2\pi )\). Thus the \({\sigma }\)-regularized LK\(({\zeta })\) equation is obtained by requiring that the driving measures are given by
Note that, when \({\zeta }={\alpha }+\eta \), the density of these driving measures is the product of the density of the local attachment rate and the local particle capacity (7) for ALE\(({\alpha },\eta )\). By the Loewner–Kufarev theory, for any solution \((\phi _t)_{t\geqslant 0}\) of (9), the sets
form an increasing family of simply-connected compacts, with capacities given by
2.1 Linearization
We compute the linearization of (9) around a solution \((\phi _t)_{t\geqslant 0}\). For \(\psi \) holomorphic in \(\{|z|>1\}\), we have
where
and, setting \(\rho =\psi '/\phi '\),
Note that first-order variations in \({{\mathcal {S}}}\) have the form
The process of first-order variations \((\psi _t)_{t\geqslant 0}\) around a solution \((\phi _t)_{t\geqslant 0}\) can be expected to satisfy the linearized equation
2.2 Linear stability of disk solutions in the subcritical case
Fix \({\tau }_0\in (0,\infty )\). A trial solution \(\phi _t(z)=e^{{\tau }_t}z\) for (9) leads to the equation
We solve to obtain \(({\tau }_t)_{t<t_{\zeta }}\), with \({\tau }_t\rightarrow \infty \) as \(t\rightarrow t_{\zeta }\), given by
where
For the associated solutions \((\phi _t)_{t<t_{\zeta }}\), the the sets \(K_t\) form a growing family of disks. We call such a \((\phi _t)_{t<{\tau }_{\zeta }}\) a disk solution.
For disk solutions, we have \(\phi _t'(z)=e^{{\tau }_t}\) for all z, so we can evaluate the integral (10) to obtain
where
Here and below, we write \(D\psi (z)\) for the radial derivative \(z\psi '(z)\). Consider the action of Q on the set of holomorphic functions on \(\{|z|>1\}\) which are bounded at infinity. Then Q is a multiplier operator
where
It is straightforward to obtain the following lower bounds. We have
Define for \({\tau }\geqslant 0\)
At least formally, at a disk solution, the linearized equation \({\dot{\psi }}_t=\nabla a(\phi _t)\psi _t\) has solution given by
In the case \({\sigma }=0\), we have \(q(k)=(1-{\zeta })k\) so \(P({\tau })\psi (z)=\psi (e^{(1-{\zeta }){\tau }}z)\) for suitable \(\psi \). Thus, if \({\zeta }>1\), as for example in the Hele–Shaw case when \({\zeta }=2\), we see that \(\psi _t\) is holomorphic in \(\{|z|>1\}\) only if \(\psi _0\) extends to a holomorphic function in the larger domain \(\{|z|>e^{(1-{\zeta })({\tau }_t-{\tau }_0)}\}\). On the other hand, if \({\zeta }\leqslant 1\), then \(P({\tau }_t-{\tau }_0)\) preserves the set of holomorphic first-order variations, so the variation \(\psi _t\) as given by (13) remains holomorphic for all t. We will show that this stability property in fact also holds whenever \({\sigma }\geqslant 0\) and \({\zeta }\leqslant 1\).
Define for \(r>1\)
For a multiplier operator M, given by
let us write \(A=A(M)\) for the smallest constant such that
The Marcinkiewicz multiplier theorem is recalled in Sect. A.3. This implies in particular that, for all \(p\in (0,\infty )\), there is a constant \(C=C(p)<\infty \) such that, for all \(r>1\), we have
We use this criterion to obtain some estimates on the operators \(P({\tau })\) and \(DP({\tau })\) for \({\tau }\geqslant 0\). Note that, if \(0\leqslant m(k)\leqslant m^*\), then \(A(M)\leqslant N(M)m^*\) where N(M) is the number of maximal intervals of constant sign in the sequence of increments \((m(k+1)-m(k):k\geqslant 0)\).
Lemma 2.1
For all \(n\geqslant 0\) and all \(p\in (1,\infty )\), there is a constant \(C=C(n,p)<\infty \) such that, for all \({\sigma }\geqslant 0\), for all holomorphic functions \(\psi \) on \(\{|z|>1\}\) bounded at \(\infty \), all \({\tau }\geqslant 0\) and all \(r>1\), we have, for \({\zeta }<1\),
and for \({\zeta }=1\)
Proof
Consider first the case where \({\zeta }\leqslant 0\). We split
Then, with obvious notation, \(P({\tau })=P_1({\tau })P_2({\tau })\) so
The sequence of multipliers \(k^ne^{-{\tau }k}\) for \((-D)^nP_1({\tau })\) is bounded by \((n/{\tau })^n\) and its increments change sign at most once, so \(A(D^nP_1({\tau }))\leqslant 2(n/{\tau })^n\). The sequence of multipliers \(e^{-{\tau }q_1(k)}\) for \(P_2({\tau })\) is bounded by 1 and its increments change sign at most once, so \(A(P_2({\tau }))\leqslant 2\). Hence \(\Vert D^nP({\tau })\Vert _{p,r}\leqslant C/{\tau }^n\) as claimed.
Consider next the case \({\zeta }\in (0,1)\). We make another split
Then \(P({\tau })=P_1({\tau })P_2({\tau })\) again, where the notation now corresponds to the new split. We have \(A(D^nP_1({\tau }))\leqslant 2(n/((1-{\zeta }){\tau }))^n\) by the argument used for \(D^nP_1\) in the case \({\zeta }\leqslant 0\). The sequence of multipliers \(e^{-{\tau }q_1(k)}\) for \(P_2({\tau })\) is bounded by 1 and is decreasing, so \(A(P_2({\tau }))\leqslant 1\). Hence \(\Vert D^nP({\tau })\Vert _{p,r}\leqslant C/((1-{\zeta }){\tau })^n\) as claimed.
Consider finally the case \({\zeta }=1\). We now write
and write \({\hat{P}}({\tau })\) for the operator with multipliers \(e^{-{\tau }{\hat{q}}(k)}\) and so on. As already observed, the sequence of multipliers \(k^ne^{-{\tau }k}\) for \((-D)^nP_1({\tau })\) is bounded by \((n/{\tau })^n\) and its increments change sign at most once, so \(A(D^nP_1({\tau }))\leqslant 2(n/{\tau })^n\). The sequence of multipliers \(k^ne^{-{\tau }{\sigma }k^2}\) for \((-D)^nP_2({\tau })\) is bounded by \((n/({\sigma }{\tau }))^{n/2}\) and its increments also change sign at most once, so \(A(DP_2({\tau }))\leqslant 2\). We use the inequality
to deduce that \(A(D{\hat{P}}({\tau }))\leqslant Ce^{{\sigma }n}/({\tau }^n\wedge ({\sigma }{\tau })^{n/2})\). Finally, it is straightforward to check that the sequence of multipliers \(e^{-{\tau }q_3(k)}\) for \(P_3({\tau })\) is bounded by 1 and decreasing, so \(A(P_3({\tau }))\leqslant 1\). Hence \(\Vert D^nP({\tau })\Vert _{p,r}\leqslant \Vert D^n{\hat{P}}({\tau })\Vert _{p,r}\Vert P_3({\tau })\Vert _{p,r}\leqslant Ce^{{\sigma }n}/({\tau }^n\wedge ({\sigma }{\tau })^{n/2})\) as claimed. \(\square \)
2.3 Transformation to (Schlicht function, capacity) coordinates
Write \({{\mathcal {S}}}_1\) for the set of ‘Schlicht functions at \(\infty \)’ on \(\{|z|>1\}\), given by
It will be convenient to use coordinates \(({\hat{\phi }},{\tau })\) on \({{\mathcal {S}}}\), given by
Then \({\hat{\phi }}\in {{\mathcal {S}}}_1\) and \({\tau }\in \mathbb {R}\). It is straightforward to show that, for a solution \((\phi _t)_{t\geqslant 0}\) to (9), the transformed variables \(({\hat{\phi }}_t,{\tau }_t)_{t\geqslant 0}\) satisfy
where
On linearizing (14) around a solution \(({\hat{\phi }}_t,{\tau }_t)_{t\geqslant 0}\), we obtain equations for first-order variations \(({\hat{\psi }}_t,\psi ^{{\text {cap}}}_t)_{t\geqslant 0}\) in the new coordinates, where now \({\hat{\psi }}_t\) is bounded at \(\infty \) for all t, reflecting the normalization of \({\hat{\phi }}_t\). These are then related to the first-order variations \((\psi _t)_{t\geqslant 0}\) in the old coordinates by
For a disk solution \((\phi _t)_{t<t_{\zeta }}\), we have \({\hat{\phi }}_t(z)=z\) and \(b({\hat{\phi }}_t,{\tau })=(0,e^{-{\zeta }{\tau }})\). The equations for first-order variations are then given by
with solutions
3 Interpolation Formula for Markov Chain Fluid Limits
We use an interpolation formula between continuous-time Markov chains and differential equations, which we first review briefly in a general setting. This formula is then applied to an ALE(\({\alpha },\eta \)) aggregation process \((\Phi _t)_{t\geqslant 0}\) with capacity parameter c, regularization parameter \({\sigma }\) and particle family \((P^{(c)}:c\in (0,\infty ))\), taking as limit equation the \({\sigma }\)-regularized LK\(({\zeta })\) equation with \({\zeta }={\alpha }+\eta \). We use (Schlicht function, capacity) coordinates for both the process and the limit equation.
3.1 General form of the interpolation formula
Let \((X_t)_{t\geqslant 0}\) be a continuous-time Markov chain with state-space E and transition rate kernel q, starting from \(x_0\) say. Suppose for this general discussion that \(E=\mathbb {R}^d\). Let b be a vector field on E with continuous bounded derivative \(\nabla b\). Write \((\xi _t(x):t\geqslant 0,x\in E)\) for the flow of b. The compensated jump measure of \((X_t)_{t\geqslant 0}\) is the signed measure \({\tilde{\mu }}^X\) on \(E\times (0,\infty )\) given by
Set \(x_t=\xi _t(x_0)\) and define, for \(s\in [0,t]\),
Then \(Z_0=x_t\) and \(Z_t=X_t\) and, on computing the martingale decomposition of \((Z_s)_{s\leqslant t}\), we obtain the interpolation formula
where
and
where \({\beta }\) is the drift of \((X_t)_{t\geqslant 0}\), given by
The identification of martingales associated with finite-rate continuous-time Markov chains is standard. The particular pathwise formulation in terms of the jump measure used here is developed in detail in [6]. We will use this formula in a case where the state-space E is infinite-dimensional. Rather than justify its validity generally in such a context, in the next section, we will prove directly the special case of the formula which we require. Note that the integrands in \(M_t\) and \(A_t\) depend on t. Nevertheless, we will call \(M_t\) the martingale term and \(A_t\) the drift term.
3.2 Proof of the formula for ALE(\({\alpha },\eta \))
Let \((\Phi _t)_{t\geqslant 0}\) be an ALE(\({\alpha },\eta \)) aggregation process with capacity parameter c, regularization parameter \({\sigma }\) and particle family \((P^{(c)}:c\in (0,\infty ))\). See Sect. 1.2 and (7) for the specification of this process. We use (Schlicht function, capacity) coordinates, as in Sect. 2.3, to obtain a continuous-time Markov chain \((X_t)_{t\geqslant 0}=({\hat{\Phi }}_t,{{\mathcal {T}}}_t)_{t\geqslant 0}\) in \({{\mathcal {S}}}_1\times [0,\infty )\). When in state \(x=({\hat{\phi }},{\tau })\), for all \({\theta }\in [0,2\pi )\), this process makes a jump of size \((\Delta ({\theta },z,c({\theta }),{\hat{\phi }}),c({\theta }))\) at rate \({\lambda }({\theta })d{\theta }/(2\pi )\), where
and
We can and do assume that the process is constructed from a Poisson random measure \(\mu \) on \([0,2\pi )\times [0,\infty )\times (0,\infty )\) of intensity \((2\pi )^{-1}d{\theta }dvdt\) by the following stochastic differential equation:
where
and
We use the vector field \(b=({\hat{b}},b^{{\text {cap}}})\) of the \({\sigma }\)-regularized LK\(({\zeta })\) equation (14), written in (Schlicht function, capacity) coordinates. Consider the disk solution \((x_t)_{t\geqslant 0}=({\hat{\phi }}_t,{\tau }_t)_{t<t_{\zeta }}\) with initial capacity \({\tau }_0=0\), which is given by
We will compute the form of the interpolation formula in this case and then prove directly that it holds. Note that
and, for \(y=({\hat{\psi }},\psi ^{{\text {cap}}})\),
and the first-order variation at time t due to a variation y at time \(s\leqslant t\) is given by
Write \({\tilde{\mu }}\) for the compensated Poisson random measure
Fix \(t\geqslant 0\) and set \({\bar{{\tau }}}_s={\tau }_t-{\tau }_s\). We alert the reader to the concealed dependence of \({\bar{{\tau }}}_s\) on t. The martingale term \(M_t=({\hat{M}}_t,M^{{\text {cap}}}_t)\) in the interpolation formula may then be written
The drift \({\beta }=({\hat{{\beta }}},{\beta }^{{\text {cap}}})\) for \(({\hat{\Phi }},{{\mathcal {T}}})\) is given by
Write \({\hat{\Psi }}_s(z)={\hat{\Phi }}_s(z)-{\hat{\phi }}_s(z)={\hat{\Phi }}_s(z)-z\) and \(\Psi ^{{\text {cap}}}_s={{\mathcal {T}}}_s-{\tau }_s\). Then we have formally
and so
The following interpolation identities may then be obtained formally by splitting equation (15) into its Schlicht function and capacity components.
Proposition 3.1
For all \(t<t_{\zeta }\) and all \(|z|>1\), we have
where
Proof
Fix \(t<t_{\zeta }\). For \(x\in [0,t]\), recall that \({\bar{{\tau }}}_x={\tau }_t-{\tau }_x\) and define for \(|z|>1\)
Set
and
We will show that, for all \(x\in [0,t]\) and all \(|z|>1\),
The case \(x=t\) gives the claimed identities. In the case \(x=0\), all terms are 0. The left-hand and right-hand sides are piecewise continuously differentiable in x, except for finitely many jumps, at the jump times of \((\Phi _x)_{0\leqslant x\leqslant t}\), which occur when \(\mu \) has an atom at \(({\theta },v,x)\) with \(v\leqslant {\Lambda }_x({\theta })\). It will suffice to check that the jumps and derivatives agree. Now \({\hat{A}}_{x,t}(z)\) and \(A^{{\text {cap}}}_{x,t}\) are continuous in x and, at the jump times of \(\Phi _x\), the jumps in \({\hat{\Psi }}_{x,t}(z)\) and \(\Psi ^{{\text {cap}}}_{x,t}\) are given by
and
So it remains to check the derivatives. We will use a spectral calculation for the semigroup of multiplier operators \(P({\tau })=e^{-{\tau }Q}\), whose justification is straightforward. Recall that \({\dot{{\tau }}}_t=e^{-{\zeta }{\tau }_t}\). We have
and
So, between the jump times, we have
and
and
Hence, between the jump times,
as required. \(\square \)
4 Estimation of Terms in the Interpolation Formula
We obtain some estimates on the terms in the interpolation formula (17) for ALE(\({\alpha },\eta \)) when it is close to the disk solution (16) of the LK\(({\zeta })\) equation, with \({\zeta }={\alpha }+\eta \). For \({\delta }_0\in (0,1/2]\), define
We estimate first the martingale term and then the drift term.
4.1 Estimates for the martingale terms
Recall that the martingale term \(({\hat{M}}_t,M^{{\text {cap}}}_t)\) in the interpolation formula is given by
where \(E(t)=[0,2\pi )\times [0,\infty )\times (0,t]\) and
with
and
Consider the following approximations to \({\hat{M}}_t(z)\) and \(M_t^{{\text {cap}}}\), which are obtained by replacing \({\hat{\Phi }}_{s-}\) by \({\hat{\phi }}_s\), \({{\mathcal {T}}}_{s-}\) by \({\tau }_s\) and \(e^{-c}F_c({\theta },z)-z\) by \(2cz/(e^{-i{\theta }}z-1)\). (Under our assumptions on the particle family, the last approximation becomes good in the limit \(c\rightarrow 0\). See Sect. A.2 and in particular equation (110).) Define
where
and
Lemma 4.1
For all \({\alpha },\eta \in \mathbb {R}\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,p,T)<\infty \), such that, for all \(c\in (0,1]\), all \({\sigma }\geqslant 0\) and all \({\delta }_0\in (0,1/2]\),
and
Proof
We write \(T_0\) for \(T_0({\delta }_0)\) in the proofs. Consider the martingale \((M_t)_{t<t_{\zeta }}\) given by
By an inequality of Burkholder, for all \(p\geqslant 2\), there is a constant \(C(p)<\infty \) such that, for all \(t\geqslant 0\),
We write here \(M^*_t\) for \(\sup _{s\leqslant t}|M_s|\) and similarly for other processes. See [3, Theorem 21.1] for the discrete-time case. The continuous-time case follows by a standard limit argument. Now
and
For all \(t\leqslant T_0\wedge T\) and all \({\theta }\in [0,2\pi )\), we have
so \(\langle M \rangle _t\leqslant Cc\) and \((\Delta M)^*_t\leqslant Cc\). Here and below, we write C for a finite constant of the dependence allowed in the statement. The value of C may vary from one instance to the next. We remind the reader that \(C_t({\theta })\) and \({\Lambda }_t({\theta })\) are defined at (18). Hence
Since \(M^{{\text {cap}}}_t=e^{-{\zeta }{\tau }_t}M_t\) for all \(t\leqslant T_0\), the first claimed estimate follows.
For the second estimate, we use instead the martingale \((M_t)_{t\geqslant 0}\) given by
Then
For \(t\leqslant T_0\wedge T\) and \({\theta }\in [0,2\pi )\), we have
so
Then \(\langle M \rangle _t\leqslant Cc{\delta }_0\) and \((\Delta M)_t\leqslant Cc\). Hence, by Burkholder’s inequality,
Since \(M^{{\text {cap}}}_t-\Pi ^{{\text {cap}}}_t=e^{-{\zeta }{\tau }_t}M_t\) for all \(t\leqslant T_0\), the second claimed estimate follows. \(\square \)
Note that, since \({\hat{\Phi }}_t\) takes values in \({{\mathcal {S}}}_1\), the holomorphic function \({\hat{\Psi }}_t(z)={\hat{\Phi }}_t(z)-z\) is bounded at \(\infty \) and hence has a limiting value \({\hat{\Psi }}_t(\infty )\). The same is true for the terms \({\hat{M}}_t\) and \({\hat{A}}_t\) in the interpolation formula. Instead of estimating these terms directly, we estimate first their values at \(\infty \) and then their radial derivatives \(D{\hat{M}}_t\) and \(D{\hat{A}}_t\), since this gives the best control of the derivative of \({\hat{\Phi }}_t\) near the unit circle, which drives the dynamics of the process.
Lemma 4.2
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },p,T)<\infty \), such that, for all \(c\in (0,1]\), all \({\sigma }\geqslant 0\), all \({\delta }_0\in (0,1/2]\) and all \(t\leqslant T\),
and
Proof
By considering the Laurent expansions of \(F_c\) and \({\hat{\phi }}\), we have
Consider the martingale \((M_t)_{t<t_{\zeta }}\) given by
Then \({\hat{M}}_t(\infty )=e^{-{\tau }_t}M_t\) for all \(t\leqslant T_0\). By Proposition A.5, \(|a_0(c)|\leqslant Cc\) for all c. Hence
and, for \(p\geqslant 2\), since
we have
The first claimed estimate then follows from Burkholder’s inequality (22).
For the second estimate, we consider instead the martingale \((M_t)_{t<t_{\zeta }}\) given by
Then \({\hat{M}}_t(\infty )-{\hat{\Pi }}_t(\infty )=e^{-{\tau }_t}M_t\) for all \(t\leqslant T_0\). By Proposition A.5, we have \(|a_0(c)-2c|\leqslant Cc^{3/2}\). We combine this with (23) and (24) to see that
The second estimate then follows by Burkholder’s inequality as above. \(\square \)
Recall that, for \(p\in [1,\infty )\) and \(r>1\), we set
For a measurable function \(\Psi \) on \({\Omega }\times \{|z|>1\}\), we set
Lemma 4.3
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \({\varepsilon }\in (0,1/2)\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\varepsilon },{\Lambda },p,T)<\infty \) such that, for all \(c\in (0,1]\), all \({\sigma }\geqslant 0\), all \({\delta }_0\in (0,1/2]\) and all \(t\leqslant T\), for all \(r\geqslant 1+c^{1/2-{\varepsilon }}\), for \(\rho =(1+r)/2\), we have, in the case \({\zeta }<1\),
and
while in the case \({\zeta }=1\) the same bounds hold with \(\left( \frac{r}{r-1}\right) \) replaced in the first term on the right-hand side by \(\left( \frac{r}{r-1}\right) +\frac{1}{\sqrt{\sigma }}\left( \frac{r}{r-1}\right) ^{1/2}\).
Proof
Recall that we write \(T_0\) for \(T_0({\delta }_0)\). Fix \(t\leqslant T<t_{\zeta }\). For \(s\in [0,t]\), we will write \({\bar{{\tau }}}_s\) for \({\bar{{\tau }}}_{t,s}={\tau }_t-{\tau }_s\). Consider for \(|z|>1\), the martingale \((M_x(z))_{0\leqslant x\leqslant t}\) given by
By Burkholder’s inequality, for \(p\geqslant 2\) and all \(|z|>1\),
On the event \(\{t\leqslant T_0\}\), we have \(D{\hat{M}}_t(z)=M_t(z)\) so, on taking the \(\Vert .\Vert _{p,r}\)-norm in (29), we obtain
Now
and
Also
so
We have \({\Lambda }_s({\theta })\leqslant C/c\) for all \(s\leqslant T_0\) and \({\theta }\in [0,2\pi )\). Hence
and
Similarly,
We will split the jump \(\Delta ({\theta },z,c,{\hat{\phi }})\) as the sum of several terms, and thereby split \(H_s({\theta },z)\) and hence \(M_t\) also as a sum of terms. For each of these terms, we will use one of the inequalities (32), (33) and one of (31), (34) to obtain a suitable upper bound for the right-side of (30). These bounds will combine to prove the first claimed estimate.
Recall that \({\hat{\phi }}(z)=z+{\hat{\psi }}(z)\), so
where
We further split the second term by expanding in Taylor series, using an interpolation from z to \(F_c({\theta },z)\). For \(u\in [0,1]\), define
Then \(F_{c,0}({\theta },z)=z\) and \(F_{c,1}({\theta },z)=F_c({\theta },z)\). Fix c, \({\theta }\) and z and set
then
Set \(m=\lceil 1/(8{\varepsilon })\rceil \) and recall that our constants C are allowed to depend on \({\varepsilon }\). Then
where, for \(k=1,\dots ,m\),
and
Let us write
and
and
We consider first the contribution of
We make the further split \(\Delta _0=\Delta _{0,0}+\Delta _{0,1}\), where
and
We will exploit the more explicit form of \(\Delta _{0,0}\), which is the main term as \(c\rightarrow 0\) under our particle assumptions (4), (5) and (6), to obtain better estimates. We have, with obvious notation,
so, for \({\tau }\geqslant 0\),
By Proposition A.5, \(|a_0(c)|\leqslant Cc\) for all c. So, for \(|z|=r\) and \({\tau }\geqslant 0\),
and
Hence we have
We used the facts that \((d/ds){\bar{{\tau }}}_s=-{\dot{{\tau }}}_s\) and \({\dot{{\tau }}}_s=e^{-{\zeta }{\tau }_s}\) and \(e^{{\zeta }{\tau }_s}\leqslant C\) to see that, for all \({\lambda }>0\),
We will use similar estimates for other integrals of \(({\bar{{\tau }}}_s)_{s\leqslant t}\) without further explanation. Now \(q(k)\geqslant (1-{\zeta }^+)k\) so we obtain, for \({\zeta }<1\),
On the other hand, for \({\zeta }=1\), we have \(q(k)\geqslant (({\sigma }k^2)\wedge k)/C\) so we obtain
We use (31) and (37) to obtain, for \(|z|=r>1\),
On substituting the estimates for \(\langle M^{0,0}(z) \rangle _t\) and \((\Delta M^{0,0}(z))^*_t\) into (30), we obtain for \(r\geqslant 1+\sqrt{c}\) and \(p\geqslant 2\), for \({\zeta }<1\),
while, for \({\zeta }=1\),
We turn to the contribution of \(\Delta _{0,1}\). For \(s\leqslant T_0\) and all \({\theta }\in [0,2\pi )\), we have
By Proposition A.6, there is a family of functions \((Q_u:u\in [0,1])\), each holomorphic on \(\{|z|>1\}\), such that
and such that
where \(Q_u({\theta },z)=e^{i{\theta }}Q_u(e^{-i{\theta }}z)\). We use the Laurent series
to write
Hence we obtain, for \(|z|=r>1\),
where we used
and
Now
so, using again (38),
Hence, using the same lower bounds for q(k) as above, we obtain, for \({\zeta }<1\),
and, for \({\zeta }=1\),
Hence, for \(|z|=r>1\) and \({\zeta }<1\), we have
while, for \({\zeta }=1\), similarly,
Also, for all \(s\leqslant T_0\) and \(|z|=r>1\), we have
Hence we obtain, for \(p\geqslant 2\) and \(r\geqslant 1+\sqrt{c}\), for \({\zeta }<1\),
and, for \({\zeta }=1\), similarly,
We consider next, for \(k=1,\dots ,m\), the contribution of
In order to avoid the appearance of a spurious log term in the case \({\zeta }=1\) we treat this contribution a little differently. We take an additional derivative, estimate the derivative and finally integrate that estimate. We have
where \(L_u({\theta },z)=e^{i{\theta }}L_u(e^{-i{\theta }}z)\) and \(L_u(z)\) is given by (117). Then
so
where
Hence, for \(s\leqslant T_0\),
so
where
By Proposition A.6, for \(|z|=r\geqslant 1+\sqrt{c}\) and \(u\leqslant Cc\),
so
and so by Proposition A.7, for \(j=1,\dots ,k\) and \(\rho =(r+1)/2\) and \(\rho '=(3r+1)/4\), for \({\zeta }<1\),
In estimating \(\Vert D^2P({\tau })\Vert _{p,\rho '\rightarrow r}\), we used the better of two estimates – either the case \(n=0\) of Lemma 2.1 in conjunction with (119) or the case \(n=2\) of Lemma 2.1. A similar but easier estimate holds for \(\Vert h_s\Vert _{p,r}\). Now
so, for \(r\geqslant 1+\sqrt{c}\) and \({\zeta }<1\),
and so
We used here the inequality
which holds for all \(a>0\) and all \(p>1\), to see that
For \(p\geqslant 2\) and \(r>1\), we have
and, from (46), for \(r\geqslant 1+\sqrt{c}\), estimating as above but now using the second estimate of Proposition A.7, we get
so, for \(r\geqslant 1+\sqrt{c}\),
On substituting the estimates for \(\langle DM^k(z) \rangle _t\) and \((\Delta DM^k(.))^*_t\) into (30), we obtain for \(r\geqslant 1+\sqrt{c}\) and \(p\geqslant 2\), for \({\zeta }<1\),
In the case \({\zeta }=1\), we have to modify the above estimation in using
We obtain in this case
Now, since the holomorphic functions \(M^k_t\) and \({\hat{\Psi }}_{s-}\) vanish at \(\infty \), we have
and, for \(a\geqslant 1\),
so, on integrating (49) and (50) we obtain, for \({\zeta }<1\),
while, for \({\zeta }=1\),
(We remark that, if a similar argument is used to estimate directly, then one obtains the same estimate (51) for \({\zeta }<1\) but one faces in the case \({\zeta }=1\) the integral
The \(p=1\) case of (47) then generates a log term, which our method avoids.)
We consider finally the contribution of
Then
By Proposition A.5, we have
Hence, for \(s\leqslant T_0\) and \({\tau }\geqslant 0\),
By Lemma 2.1, for \({\zeta }<1\),
We have
and, since \(|F_{c,u}({\theta },z)|\geqslant |z|\), we have
Hence, for \({\zeta }<1\), we have
so, using (33),
and so
Here we have used our choice of \(m\geqslant 1/(8{\varepsilon })\) and the assumption \(r\geqslant 1+c^{1/2-{\varepsilon }}\) to see that
The bound (48) remains valid with \(M^{m+1}\) in place of \(M^k\). Hence for \({\zeta }<1\)
For \({\zeta }=1\), given the weaker bound for \(\Vert DP({\tau })\Vert _{p,r}\) in Lemma 2.1, we adapt the argument as above to obtain
where \(\log \) term has been absorbed using our choice of m, and then
Now
and we have shown that all terms on the right-hand side can be bounded by the right-hand side in (27), so this first estimate is now proved.
It remains to show the second estimate. Fix \(t\geqslant 0\) and consider, for \(|z|>1\), the martingale \((\Pi _x(z))_{x\geqslant 0}\) given by
Set \({\tilde{M}}_x(z)=M^{0,0}_x(z)-\Pi _x(z)\). Then
and
For all but the first term on the right, the bounds (44), (45), (51), (52), (53), (54), are sufficient for (28). It remains to show a suitable bound on \({\tilde{M}}_t\). We use the estimate (25) to see that, for \({\zeta }<1\),
while for \({\zeta }=1\) we obtain similarly
Otherwise we can proceed as for \(M^{0,0}\) to arrive as the following estimates, which suffice for (28). For \({\zeta }<1\), we have
while for \({\zeta }=1\)
\(\square \)
4.2 Estimates for the drift terms
We turn to the drift terms, beginning with estimates for the drift \(({\hat{{\beta }}},{\beta }^{{\text {cap}}})\) of the ALE\(({\alpha },\eta )\) process. Recall that \(({{\mathcal {T}}}_t)_{t\geqslant 0}\) has drift given by
where
Lemma 4.4
For all \({\zeta }\in \mathbb {R}\) and all \(T<t_{\zeta }\), there is a constant \(C({\zeta },T)<\infty \) such that, for all \({\delta }_0\in (0,1/2]\), all \(t\leqslant T\), all \({\hat{\phi }}\in {{\mathcal {S}}}_1\) and all \({\tau }\geqslant 0\), we have
whenever \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\), where \(\psi ^{{\text {cap}}}_t={\tau }-{\tau }_t\) and \({\hat{\psi }}(z)={\hat{\phi }}(z)-z\).
Proof
We have
and, for \(|w|\leqslant 1/2\),
so
where
whenever \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\). For \({\hat{\phi }}\in {{\mathcal {S}}}_1\), \({\hat{\psi }}\) is holomorphic in \(\{|z|>1\}\) and bounded at \(\infty \), so
The claimed estimate follows on integrating (55) in \({\theta }\). \(\square \)
Recall that the drift of \(({\hat{\Phi }}_t)_{t\geqslant 0}\) is given by
where
It is convenient in the following statement to use the notation
for functions \(\phi \) holomorphic in \(\{|z|>1\}\) and bounded at \(\infty \).
Lemma 4.5
For all \({\alpha },\eta \in \mathbb {R}\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },T)<\infty \) with the following property. For all \(c\in (0,1/C]\), all \({\sigma }>0\), all \({\delta }_0\in (0,1/2]\), all \(t\leqslant T\), all \({\hat{\phi }}\in {{\mathcal {S}}}_1\) and all \({\tau }\geqslant 0\), we have
whenever \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\), where \(\psi ^{{\text {cap}}}_t={\tau }-{\tau }_t\) and \({\hat{\psi }}(z)={\hat{\phi }}(z)-z\).
Moreover, for all \({\alpha },\eta \in \mathbb {R}\), all \({\varepsilon }\in (0,1/2]\), all \(p\geqslant 2\) and all \(T<{\tau }_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },p,T)<\infty \) with the following property. For all \(c\in (0,1/C]\), all \({\sigma }>0\), all \({\delta }_0\in (0,1/2]\), all \(t\leqslant T\), all \({\hat{\phi }}\in {{\mathcal {S}}}_1\) and all \({\tau }\geqslant 0\), for all \(r\geqslant 1+c^{1/2-{\varepsilon }}\) and \(\rho =(3r+1)/4\), we have
whenever \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\).
Proof
We use the split (35) and the Taylor expansion (36) to write
where \(m=\lceil 1/(8{\varepsilon })\rceil \). We further split
and
Note that we now split \(\Delta _0\) slightly differently to the split \(\Delta _0=\Delta _{0,0}+\Delta _{0,1}\) used for the martingale term: where before we had \(a_0(c)\) we now approximate by 2c, putting an additional error into the remainder term \({\tilde{\Delta }}_0\). Set
and note that
We use equation (55) to write
Now \({\hat{\psi }}'(z)\rightarrow 0\) as \(z\rightarrow \infty \), so
Hence
and
We will estimate the terms on the right-hand sides of (62) and (63), assuming from now on that t, \({\hat{\phi }}\) and \({\tau }\) are chosen so that \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\).
From (55) and (56), we have \(|{\gamma }_t({\theta },{\hat{\phi }},{\tau })|\leqslant C{\delta }_0^2\) and
We use (26) to see that
Write \(c({\theta })\) for \(c({\theta },{\hat{\phi }},{\tau })\) and \({\lambda }({\theta })\) for \({\lambda }({\theta },{\hat{\phi }},{\tau })\). Then
and, by Proposition A.5, we have
We can now estimate in (62) to obtain (58).
It remains to prove (59). For \(|z|=r>1\), we have
Since \({\hat{\psi }}\) is bounded at \(\infty \), by Marcinkiewicz’s multiplier theorem, \(\Vert {\hat{\psi }}\Vert _{p,r,0}\leqslant C\Vert D{\hat{\psi }}\Vert _{p,r}\) for all \(p>1\) and \(r>1\). Hence
It remains to deal with the final term in (63). We first estimate the function obtained on replacing \(c({\theta },{\hat{\phi }},{\tau })\) and \({\lambda }({\theta },{\hat{\phi }},{\tau })\) in that term by \(c_t=ce^{-{\alpha }{\tau }_t}\) and \({\lambda }_t=c^{-1}e^{-\eta {\tau }_t}\). Note that, in the case \(F_c(z)=e^cz\) and \(m=1\), the Taylor expansion (36) has the form
On the other hand, by Cauchy’s theorem,
Hence, on integrating in \({\theta }\) in (61), we see that
so, for \(r>1\) and \(\rho =(3r+1)/4\),
It remains to deal with the error made in replacing \(c({\theta },{\hat{\phi }},{\tau })\) and \({\lambda }({\theta },{\hat{\phi }},{\tau })\) by \(c_t\) and \({\lambda }_t\). We make a further split
where
Thus \({\tilde{\Delta }}_{0,0}=\Delta _{0,1}\), as considered in estimating the martingale terms, and \({\tilde{\Delta }}_{0,1}\) is the additional error introduced by the new split (60). We first estimate the \({\tilde{\Delta }}_{0,1}\) term. Since \(|\psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\psi }}'(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\), we have
Hence, by Proposition A.5, for \(c\leqslant 1/C\),
and, estimating as for (64), we obtain
By Proposition A.5, for \(c\leqslant 1/C\),
and, for \(c_1,c_2\in (0,c]\) and \(|z|\geqslant 1+\sqrt{c}\),
so
so, for \(|z|=r\geqslant 1+\sqrt{c}\),
We have
so, by Proposition A.5, for \(c\leqslant 1/C\),
and, for \(c_1,c_2\in (0,c]\) and \(|z|\geqslant 1+\sqrt{c}\),
so
so, for \(|z|=r\geqslant 1+\sqrt{c}\),
For \(k=2,\dots ,m\), we have
where \(f_c({\theta },z)=\log (F_c({\theta },z)/z)\). By Proposition A.5, for \(c\leqslant 1/C\) and \(|z|=r>1\),
and, for \(c_1,c_2\in (0,c]\) and \(|z|=r\geqslant 1+\sqrt{c}\),
so, for \(j=0,1,\dots ,k\),
so
and
so
so
and so, for \(r\geqslant 1+2\sqrt{c}\),
where we used the inequality \(\Vert {\hat{\psi }}\Vert _{p,r,0}\leqslant C\Vert D{\hat{\psi }}\Vert _{p,r}\) in the \(j=0\) term.
In the final step, we use our assumption that \(r\geqslant 1+c^{1/2-{\varepsilon }}\) and our choice of \(m=\lceil 1/(8{\varepsilon })\rceil \) to see that
Recall that
and, for \(|z|=r>1\), since \(|F_{c,u}({\theta },z)|\geqslant r\), by (118), we find, for \(\rho '=(7r+1)/8\),
So, for \(|z|=r>1\),
so
and so
The claimed estimate is obtained by combining (64), (65), (66), (67), (68), (69), (70) and (71). \(\square \)
Recall that the drift term \(({\hat{A}}_t,A^{{\text {cap}}}_t)\) in the interpolation formula (17) is given by
where \(\Psi ^{{\text {cap}}}_s={{\mathcal {T}}}_s-{\tau }_s\) and \({\hat{\Psi }}_s(z)={\hat{\Phi }}_s(z)-z\). Recall also that
Lemma 4.6
For all \({\zeta }\in \mathbb {R}\) and all \(T<t_{\zeta }\), there is a constant \(C({\zeta },T)<\infty \) such that, for all \({\sigma }>0\), all \({\delta }_0\in (0,1/2]\) and all \(t\leqslant T_0({\delta }_0)\wedge T\), we have
Proof
For all \(t\leqslant T_0({\delta }_0)\wedge t_{\zeta }\) and all \({\theta }\), we have \(|\Psi ^{{\text {cap}}}_t|\leqslant {\delta }_0\) and \(|{\hat{\Psi }}'_t(e^{{\sigma }+i{\theta }})|\leqslant {\delta }_0\) for all \({\theta }\). Hence, by Lemma 4.4, for \(t\leqslant T_0({\delta }_0)\wedge T\),
\(\square \)
Lemma 4.7
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },T)<\infty \) with the following property. For all \(c\in (0,1/C]\), all \({\sigma }>0\), all \({\delta }_0\in (0,1/2]\) and all \(t\leqslant T\),
Moreover, for all such \({\alpha },\eta \) and T, for all \({\varepsilon }\in (0,1/2]\) and all \(p\geqslant 2\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },p,T)<\infty \) with the following property. For all \(c\in (0,1/C]\), all \({\sigma }\geqslant 0\), all \({\delta }_0\in (0,1/2]\) and all \(t\leqslant T\), for all \(r\geqslant 1+c^{1/2-{\varepsilon }}\), for \(\rho =(1+r)/2\), we have in the case \({\zeta }<1\)
while for \({\zeta }=1\) the estimate (73) holds with the first factor of \(1+\log (\tfrac{r}{r-1})\) replaced by \(1+\log (\tfrac{r}{r-1})+\tfrac{1}{\sqrt{\sigma }}\) in each term on the right.
We remark that some of the log terms in (73) can be avoided when \({\zeta }<1\) by the same strategy used for (51). However, this does not work in the case \({\zeta }=1\) because that strategy also replaces the term \(\frac{1}{\sqrt{\sigma }}\) by \(\frac{1}{{\sigma }}\) which, for our main results, leads to a weaker conclusion. The \(\frac{1}{\sqrt{\sigma }}\) in (51) arises in a different way. Since spurious log terms for \({\zeta }<1\) do not affect the main results, and to economise the argument, we will not present the slightly stronger estimates than (73) that are available for \({\zeta }<1\).
Proof
The estimate (72) follows immediately from (58). Set \(\rho '=(3r+1)/4\). For \({\zeta }<1\), by Lemma 2.1, we have
By Lemma 4.5, for \(s\leqslant T_0\),
so, for \(s\leqslant t\),
Since
these estimates combine to prove (73). In the case \({\zeta }=1\), the estimate of Lemma 2.1 leads to a different integral on the right in (74), for which we have the following bound
Hence we obtain the modified form of (73) claimed for \({\zeta }=1\). \(\square \)
5 Bulk Scaling Limit for ALE\(({\alpha },\eta )\)
Recall that we write our ALE(\({\alpha },\eta \)) process \((\Phi _t)_{t\geqslant 0}\) in (Schlicht function, capacity) coordinates \(({\hat{\Phi }}_t,{{\mathcal {T}}}_t)\), and that we set
where \(({\hat{\phi }}_t,{\tau }_t)_{t<t_{\zeta }}\) is the disk solution to the LK\(({\zeta })\) equation with initial capacity \({\tau }_0=0\). We obtained the following interpolation formula (17)
and have estimated the terms on the right-hand sides in the preceding section. We now put these estimates together to obtain first \(L^p\)-estimates and then pointwise high-probability estimates which allow us to prove Theorems 1.1 and 1.2.
5.1 \(L^p\)-estimates
Recall that
Proposition 5.1
For all \({\alpha },\eta \in \mathbb {R}\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,p,T)<\infty \) such that, for all \(c\in (0,1]\) and all \({\delta }_0\in (0,1/2]\),
and
and
and
Proof
The first two estimates follow immediately from Lemmas 4.1 and 4.6. From Lemmas 4.2 and 4.7, we obtain, for all \(t\leqslant T\),
from which the third estimate follows by Gronwall’s lemma. The fourth estimate follows from the third, together with Lemmas 4.2 and 4.7. \(\square \)
Fix \({\sigma }>0\) and set
Define
The next estimates follow immediately from Lemmas 4.3 and 4.7.
Proposition 5.2
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \({\varepsilon }\in (0,1/2]\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },p,T)<\infty \) with the following property. For all \(c\in (0,1]\), all \({\delta }_0\in (0,1/2]\), all \(r,e^{\sigma }\geqslant 1+c^{1/2-{\varepsilon }}\) and all \(t\leqslant T\), setting \(\rho =(1+r)/2\), we have, for \({\zeta }<1\),
while, for \({\zeta }=1\),
The preceding estimate may be improved by an iterative argument to obtain the following result.
Proposition 5.3
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \({\varepsilon }\in (0,1/2]\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },p,T)<\infty \) with the following property. In the case \({\zeta }<1\), for all \(c\in (0,1]\), all \(r,e^{\sigma }\geqslant 1+c^{1/2-{\varepsilon }}\) and all \(t\leqslant T\), for all \(\nu \in (0,{\varepsilon }/2]\), setting \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\), we have
Moreover, in the case \({\zeta }=1\) and \({\varepsilon }\leqslant 1/3\), for all \(c\in (0,1]\), all \(r\geqslant 1+c^{1/2-{\varepsilon }}\), all \(e^{\sigma }\geqslant 1+c^{1/3-{\varepsilon }}\) and all \(t\leqslant T\), for \(\nu \in (0,{\varepsilon }/2]\), setting \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\), we have
Proof
We begin with a crude estimate which allows us to restrict further consideration to small values of c. The function \({\hat{\Phi }}_t(z)\) is univalent on \(\{|z|>1\}\), with \({\hat{\Phi }}_t(z)\sim z\) as \(z\rightarrow \infty \). So, by a standard distortion estimate, for all \(|z|=r>1\),
and so
It is straightforward to check that this implies the claimed estimates in the case where \(c>1/C\), for any given constant C of the allowed dependence. Hence it will suffice to consider the case where \(c\leqslant 1/C\).
Consider first the case \({\zeta }<1\). On substituting the chosen value of \({\delta }_0\) in (75), we obtain
Note that, for \(\rho =(1+r)/2\), we have \(R(\rho )\leqslant 2R(r)\) and \(L(\rho )\leqslant 2L(r)\), so \({\bar{{\delta }}}(\rho )\leqslant 4{\bar{{\delta }}}(r)\) and \({\bar{{\delta }}}(\rho )\leqslant 4{\bar{{\delta }}}(r)\). Note also that, for \(r\geqslant 1+c^{1/2-{\varepsilon }/2}\) and \(e^{\sigma }\geqslant 1+c^{1/2-{\varepsilon }}\), for all sufficiently small c,
where \(C^*\) is the constant in Proposition 5.2. We restrict to such c. Set \(C_0=1\) and for \(k\geqslant 0\) define recursively \(C_{k+1}=2^{k+1}C_k+1\). We will show that, for all \(k\geqslant 0\), all \(r\geqslant 1+2^kc^{1/2-{\varepsilon }/2}\) and all \(t\leqslant T\),
The case \(k=0\) is implied by (80). Suppose inductively that (81) holds for k, for all \(r\geqslant 1+2^kc^{1/2-{\varepsilon }/2}\) and all \(t\leqslant T\). Take \(r\geqslant 1+2^{k+1}c^{1/2-{\varepsilon }/2}\) and \(t\leqslant T\). Then \(\rho =(r+1)/2\geqslant 1+2^k c^{1/2-{\varepsilon }/2}\) so, for all \(s\leqslant t\),
Since \(r\geqslant 1+c^{1/2-{\varepsilon }/2}\), we can use Proposition 5.2 with \({\varepsilon }\) replaced by \({\varepsilon }/2\) and substitute the last inequality into (77) to obtain
Hence (81) holds for \(k+1\) and the induction proceeds. Choose now \(k=\lceil 3/{\varepsilon }\rceil \). Then
For c sufficiently small, we have \(c^{{\varepsilon }/2}\leqslant 2^{-k/2}\). Then, for all \(r\geqslant 1+c^{1/2-{\varepsilon }}\), we have \(r\geqslant 1+2^kc^{1/2-{\varepsilon }/2}\), so we obtain
For c sufficiently small, we have
so this is a bound of the claimed form.
We turn to the case \({\zeta }=1\). On substituting the chosen value of \({\delta }_0\) in (76), we obtain
Note that, for \(\rho =(1+r)/2\), we have \(R_1(\rho )\leqslant 2R_1(r)\) and \(L_1(\rho )\leqslant 2L_1(r)\), so \({\bar{{\delta }}}_1(\rho )\leqslant 4{\bar{{\delta }}}_1(r)\) and \({\bar{{\delta }}}_1(\rho )\leqslant 4{\bar{{\delta }}}_1(r)\). Note also that, for \(r\geqslant 1+c^{1/2-{\varepsilon }/2}\) and \(e^{\sigma }\geqslant 1+c^{1/3-{\varepsilon }}\), for all sufficiently small c,
where \(C^*\) is the constant in Proposition 5.2. We restrict to such c. Set \(C_0=1\) and for \(k\geqslant 0\) define recursively \(C_{k+1}=2^{2k+1}C_k+1\). Then, by an analogous inductive argument, we obtain, for all \(k\geqslant 0\), all \(t\leqslant T\) and all \(r\leqslant 1+2^kc^{1/2}\),
Choose now \(k=\lceil 1/{\varepsilon }\rceil \) and assume that \(r\geqslant 1+c^{1/2-{\varepsilon }}\). Then
and, for c sufficiently small, we have \(c^{\varepsilon }\leqslant 2^{-k}\), so \(r\geqslant 1+2^kc^{1/2}\) and so
\(\square \)
We note also the following estimates, which are deduced from (28) and (73) using the estimates of Proposition 5.3
Proposition 5.4
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \({\varepsilon }\in (0,1/2]\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },p,T)<\infty \) with the following property. In the case \({\zeta }<1\), for all \(c\in (0,1]\), all \(r,e^{\sigma }\geqslant 1+c^{1/2-{\varepsilon }}\) and all \(t\leqslant T\), for all \(\nu \in (0,{\varepsilon }/2]\), setting \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\), we have
Moreover, in the case \({\zeta }=1\) and \({\varepsilon }\leqslant 1/3\), for all \(c\in (0,1]\), all \(r\geqslant 1+c^{1/2-{\varepsilon }}\), all \(e^{\sigma }\geqslant 1+c^{1/3-{\varepsilon }}\) and all \(t\leqslant T\), for \(\nu \in (0,{\varepsilon }]\), setting \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\), we have
We turn now to some estimates needed for the discrete-time results Theorems 1.2 and 1.4. Write \({{\mathcal {V}}}_t\) for the number of particles added by time t and define for \(t<t_{\zeta }\)
It is straightforward to see that, for all \({\alpha },\eta \in \mathbb {R}\), we have \(\nu _t\rightarrow n_{\alpha }\) as \(t\rightarrow t_{\zeta }\). Also
Proposition 5.5
For all \({\alpha },\eta \in \mathbb {R}\), all \(p\geqslant 2\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,p,T)<\infty \) such that, for all \(c\in (0,1]\) and all \({\delta }_0\in (0,1/2]\),
Proof
Recall from Sect. 3.2 that
where
If we substitute the explicit appearances of \({\alpha }\) in the preceding line by 0, then \(C_s({\theta })\) becomes c and \(e^{-{\zeta }{\tau }_s}\) becomes \(e^{-\eta {\tau }_s}\). Then, applying these substitutions in the line above, we recover the integral representations (83) of \(c{{\mathcal {V}}}_t\) and \(\nu _t\). The claimed estimate results from following through this modification in the calculations leading to Proposition 5.1. The details are left to the reader. \(\square \)
We can also improve on the estimate of \({{\mathcal {T}}}_t\) by \({\tau }_t\) in Proposition 5.1. Define, for \(c{{\mathcal {V}}}_t<n_{\alpha }\),
where \({\tau }_n^{{\text {disc}}}={\alpha }^{-1}\log (1+{\alpha }cn)\) as at (8). We leave any modifications needed for the case \({\alpha }=0\) to the reader. By allowing \({\tilde{{{\mathcal {T}}}}}_t\) to depend on the random time-scale of particle arrivals, we remove the main source of error when estimating \({{\mathcal {T}}}_t\) by \({\tau }_t\).
Proposition 5.6
For all \({\alpha },\eta \in \mathbb {R}\), all \(p\geqslant 2\), all \(T<t_{\zeta }\) and all \(N<n_{\alpha }\), there is a constant \(C({\alpha },\eta ,p,T,N)<\infty \) such that, for all \(c\leqslant 1/C\) and all \({\delta }_0\in (0,1/2]\),
Proof
Set
Then
so
We have, for \(c{{\mathcal {V}}}_t\leqslant N\),
and, for \(t\leqslant T_0({\delta }_0)\), as in the proof of Lemma 4.4,
so
and, using (57),
and
and so
Set
Then, by Burkholder’s and Jensen’s inequalities, for \(p\geqslant 2\), and all \(t\leqslant T\),
and the claimed estimate follows by Gronwall’s lemma. \(\square \)
5.2 Spatially-uniform high-probability estimates
We now pass from the \(L^p\)-estimates of the preceding section to pointwise estimates which hold with high probability on the function \({\hat{\Psi }}_t(z)={\hat{\Phi }}_t(z)-z\), uniformly in \(t\in [0,T]\) and \(|z|\geqslant r(c)\) as \(c\rightarrow 0\), for a suitable function r(c), which is specified in the next result, and tends to 1 as \(c\rightarrow 0\). In order to show the desired uniformity, we combine the usual \(L^p\)-tail estimate with suitable dissections of [0, T] and \(\{|z|\geqslant r(c)\}\), choosing p large to deal with an increasing number of terms as \(c\rightarrow 0\). We see at the same time that the event \(\{T_0({\delta }_0)>T\}\), to which our previous estimates were restricted, is in fact an event of high probability as \(c\rightarrow 0\), thus closing the argument for convergence to a disk. The following result contains Theorem 1.1.
Proposition 5.7
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\), all \({\varepsilon }\in (0,1/2]\) and all \(\nu \in (0,{\varepsilon }/4]\), all \(m\in \mathbb {N}\) and all \(T<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,{\Lambda },{\varepsilon },\nu ,m,T)<\infty \) with the following property. In the case \({\zeta }<1\), for all \(c\leqslant 1/C\), for \(e^{\sigma }\geqslant 1+c^{1/2-{\varepsilon }}\) and \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\), there is an event \({{\Omega }_0\subseteq \{T_0({\delta }_0)>T\}}\) of probability exceeding \(1-c^m\) on which, for all \(t\leqslant T\) and all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
and
and
and
and
Moreover, in the case \({\zeta }=1\) with \({\varepsilon }\in (0,1/3]\), for all \(c\leqslant 1/C\), for \(e^{\sigma }\geqslant 1+c^{1/3-{\varepsilon }}\) and \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\) there is an event \({{\Omega }_0\subseteq \{T_0({\delta }_0)>T\}}\) of probability exceeding \(1-c^m\) on which, for all \(t\leqslant T\), the estimates (84) and (87) hold and, for all \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
and
and
Proof
We will give details for the case \({\zeta }\in [0,1)\). Some minor modifications are needed for the case \({\zeta }=1\) because of the weaker \(L^p\)-estimate (79) which applies in that case, and also for the case \({\zeta }<0\). These are left to the reader.
Fix \({\alpha },\eta ,{\varepsilon },\nu ,m\) and T as in the statement. By adjusting the value of \({\varepsilon }\), it will suffice to consider the case where \(e^{\sigma }\geqslant 1+2c^{1/2-{\varepsilon }}\), and to find an event \({{\Omega }_0\subseteq \{T_0({\delta }_0)>T\}}\), of probability exceeding \(1-c^m\), on which the claimed estimates holds whenever \({r\geqslant 1+2c^{1/2-{\varepsilon }}}\) and \(t\leqslant T\). There is a constant \(C<\infty \) of the desired dependence, such that \({\delta }_0\leqslant 1/2\) whenever \(c\leqslant 1/C\). We restrict to such c. Set
Recall that \({{\mathcal {V}}}_t\) denotes the number of particles added to the cluster by time t. Consider the event
Note that, on \({\Omega }_1\), for all \(t\leqslant T_0({\delta }_0)\wedge T\), there exists \(n\in \{1,\dots ,N_0\}\) such that \({\hat{\Psi }}_t={\hat{\Psi }}_{t(n)}\). Since \({\delta }_0\leqslant 1/2\), there is a constant \(C<\infty \) of the desired dependence such that the process \(({{\mathcal {V}}}_t)_{t\leqslant T_0({\delta }_0)}\) is a thinning of a Poisson process of rate C/c. Hence
and hence \(\mathbb {P}({\Omega }_1^c)\leqslant c^m/3\) for all \(c\leqslant 1/(3C)\). We restrict to such c.
Fix an integer \(p\geqslant 2\), to be chosen later, depending on m and \(\nu \). By Proposition 5.1, there is a constant \(C<\infty \) of the desired dependence such that, for \(\mu _0=C\left( \sqrt{c}+{\delta }_0^2\right) \), we have
Set \({\lambda }_0=(6c^{-m})^{1/p}\) and consider the event
Then \(\mathbb {P}({\Omega }_2^c)\leqslant 2{\lambda }_0^{-p}=c^m/3\). We choose \(p\geqslant m/\nu \). Then, since \(e^{\sigma }\geqslant 1+2c^{1/2-{\varepsilon }}\) and \(\nu \leqslant {\varepsilon }\), there is a constant \(C<\infty \) of the desired dependence such that, for \(c\leqslant 1/C\), on the event \({\Omega }_2\), for all \(t\leqslant T_0({\delta }_0)\wedge T\),
We restrict to such c. Set
Then \(K\leqslant \lfloor \log (1/c)\rfloor +1\). For \(k=1,\dots ,K\), set
Then \(\rho (k)\geqslant \rho (1)=1+c^{1/2-{\varepsilon }}\) for all k and \(r(K)\in [2,4]\). By Proposition 5.3, there is a constant \(C<\infty \) of the desired dependence such that, for \(k=1,\dots ,K\) and all \(t\leqslant T\),
where
Set \({\lambda }=\left( 3KTc^{-2m-3}\right) ^{1/p}\) and consider the event
Then
so
Fix \(r\geqslant 1+2c^{1/2-{\varepsilon }}\). Then \(r(k)\leqslant r<r(k+1)\) for some \(k\in \{1,\dots ,K\}\), where we set \(r(K+1)=\infty \). Note that \(zD{\hat{\Psi }}_t(z)\) is a bounded holomorphic function on \(\{|z|>\rho (1)\}\). We use the inequality (118) to see that, on the event \({\Omega }_3\), for \(n\leqslant N_0\),
so
We choose \(p\geqslant (2m+4)/\nu \). Then there is a constant \(C<\infty \) of the desired dependence such that, for \(c\leqslant 1/C\), on \({\Omega }_3\), for \(n=1,\dots ,N_0\) and all \(r\geqslant 1+2c^{1/2-{\varepsilon }}\), we have
and
We restrict to such c. Set
Then \(\mathbb {P}({\Omega }_0^c)\leqslant c^m\) and, on the event \({\Omega }_0\), for all \(t\leqslant T_0({\delta }_0)\wedge T\) and all \(r\geqslant 1+2c^{1/2-{\varepsilon }}\),
and
In conjunction with (90), this forces \(T_0({\delta }_0)>T\) on \({\Omega }_0\) and so concludes the proof of (84) and (86).
We deduce (85) using the identity
On the event \({\Omega }_2\), for all \(t\leqslant T_0({\delta }_0)\wedge T\),
On the other hand, \({\Omega }_0\subseteq {\Omega }_2\) and on \({\Omega }_0\) we have \(T_0({\delta }_0)>T\) and, using (86), for \(t\leqslant T\) and \(|z|=r\geqslant 1+c^{1/2-{\varepsilon }}\),
Since \(r\geqslant 1+c^{1/2}\), the \(\log \) factor can be absorbed in \(c^{1/2-\nu }\) by adjustment of \(\nu \). Then, on combining the last two estimates, we obtain (85).
The estimate (87) may now be deduced from Proposition 5.1 using standard \(L^p\) tail estimates. The details are left to the reader.
For the estimate (88), define
where C is the constant in Proposition 5.1, and define
where C is the constant of Proposition 5.4. Set \({\tilde{{\Omega }}}_0={\Omega }_1\cap {\tilde{{\Omega }}}_2\cap {\tilde{{\Omega }}}_3\), where
and
We follow a similar argument to above to see that \(\mathbb {P}({\tilde{{\Omega }}}_0^c)\leqslant 2c^m\) and on \({\tilde{{\Omega }}}_0\) we have \(T_0({\delta }_0)>T\) and for \(t\leqslant T\)
and for \(|z|=r\geqslant 1+2c^{1/2-{\varepsilon }}\),
Finally we can integrate as above to deduce (88). \(\square \)
Proof of Theorem 1.2
We will write the argument for the case \({\zeta }<1\), omitting the modifications needed for \({\zeta }=1\), which are left to the reader. Since \(N<n_{\alpha }\), we can choose \({\delta }={\delta }({\alpha },\eta ,N)>0\) and \(T<t_{\zeta }\) such that \(\nu _T=N+{\delta }\). Choose \({\delta }_0\) and \({\Omega }_0\) as in Proposition 5.7, with the choice of T just made. Write C for the constant appearing in Proposition 5.7 and set
Then, for all \(|z|\geqslant 1+c^{1/2-{\varepsilon }}\) and all \(t\leqslant T\), on the event \({\Omega }_0\), we have \(|{\hat{\Phi }}_t(z)-z|\leqslant \Delta \). Then, by Propositions 5.5 and 5.6, choosing \({\delta }_0\) as in Proposition 5.7 and using an \(L^p\)-tail estimate for suitably large p, there is an event \({\Omega }_1\subseteq {\Omega }_0\), of probability exceeding \(1-2c^m\), on which, for all \(t\leqslant T\), both \(|c{{\mathcal {V}}}_t-\nu _t|\leqslant \Delta \) and, provided \(c{{\mathcal {V}}}_t\leqslant N\), also
We can choose C so that, for \(c\leqslant 1/C\), we have \(\Delta \leqslant {\delta }\), so \(c{{\mathcal {V}}}_T\geqslant N+{\delta }-\Delta \geqslant N\) always on \({\Omega }_1\). Now, for all \(n\leqslant N/c\), we have \({{\mathcal {V}}}_t=n\) for some \(t\leqslant T\) with \(c{{\mathcal {V}}}_t\leqslant N\), so on \({\Omega }_1\), for all \(|z|\geqslant 1+c^{1/2-{\varepsilon }}\), we have
\(\square \)
6 Fluctuation Scaling Limit for ALE\(({\alpha },\eta )\)
Given an ALE\(({\alpha },\eta )\) process \((\Phi _t)_{t\geqslant 0}\), recall that
The fluctuations in these coordinates are given by
Recall that we write \({{\mathcal {H}}}\) for the set of holomorphic functions on \(\{|z|>1\}\) which are bounded at \(\infty \), and we use on \({{\mathcal {H}}}\) the topology of uniform convergence on \(\{|z|\geqslant r\}\) for all \(r>1\). In this section we prove Theorem 1.3 and then, at the end, we deduce Theorem 1.4.
6.1 Reduction to Poisson integrals
Our starting point is the interpolation formula (17)
As a first step, we study the approximations \({\hat{\Pi }}_t(z)\) and \(\Pi ^{{\text {cap}}}_t\) to \({\hat{M}}_t(z)\) and \(M^{{\text {cap}}}_t\) which have a simple form and which prove to be the dominant terms in the considered limit. Set
Recall the multiplier operator \(P({\tau })\) defined at (12). Then
Recall that \(c_t=ce^{-{\alpha }{\tau }_t}\) and \({\lambda }_t=c^{-1}e^{-\eta {\tau }_t}\), and that we define for \(|z|>1\)
The following result allows us to deduce the weak limit of the normalized fluctuations from that of the Poisson integrals \(({\hat{\Pi }}_t,\Pi ^{{\text {cap}}}_t)_{t\geqslant 0}\).
Proposition 6.1
For all \({\alpha },\eta \in \mathbb {R}\) with \({\zeta }={\alpha }+\eta \leqslant 1\),
in \({{\mathcal {H}}}\times \mathbb {R}\) uniformly on compacts in \([0,t_{\zeta })\), in probability, in the limit \(c\rightarrow 0\) and \({\sigma }\rightarrow 0\) considered in Theorem 1.3.
Proof
In Theorem 1.3, for \({\zeta }<1\), we restrict to \({\sigma }\geqslant c^{1/4-{\varepsilon }}\) and take \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1)\) with \(\nu \leqslant {\varepsilon }/4\). On the other hand, for \({\zeta }=1\), we restrict to \({\sigma }\geqslant c^{1/5-{\varepsilon }}\) and take \({\delta }_0=c^{1/2-\nu }e^{\sigma }/(e^{\sigma }-1))\). In both cases, the right-hand sides in (87), (88) and (89) are therefore small compared to \(\sqrt{c}\) in the considered limit. The claim thus follows from Proposition 5.7. \(\square \)
Since the integral (91) converges absolutely for all \({\omega }\), we can exchange limits to see that
where
Set \(q_0(k)=(1-{\zeta })k\) and define, for all \({\zeta }\in (-\infty ,1]\) and \(t<t_{\zeta }\),
Proposition 6.2
For all \({\alpha },\eta \in \mathbb {R}\) with \({\alpha }+\eta ={\zeta }\leqslant 1\), and all \(t<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,t)<\infty \) such that, for all \(k\geqslant 0\),
and
Moreover, C may be chosen so that, for all \(h\in [0,1]\) and all stopping times \(T\leqslant t-h\),
Proof
The estimates for \((\Pi ^{{\text {cap}}}_t)_{t<t_{\zeta }}\) are standard and are left to the reader. For \((\Pi _t(k))_{t<t_{\zeta }}\), we use time-dissection to obtain estimates with good dependence on k. Set \({\kappa }=1+q(k)\) and define
Set \(n=\lceil {\kappa }{\tau }_t\rceil \) and \(t(n)=t\). Set \(t(i)=i/{\kappa }\) for \(i=0,1,\dots ,n-1\). Then \(t(i+1)-t(i)\leqslant 1/{\kappa }\) for all i. We have
so, by Doob’s \(L^2\)-inequality,
Now, for \(t(i)\leqslant s\leqslant t(i+1)\),
so
and so
For the second estimate, set \({\kappa }_0=1+(1-{\zeta })k\) and note that
Restrict for now to the case \({\zeta }\geqslant 0\), when \({\kappa }\geqslant {\kappa }_0\), and define
and
Note that
so, by a similar argument,
Now
so
and so
For \(\Pi ^0_s(k)\), we used the estimate (93) with \({\kappa }\) replaced by \({\kappa }_0\), which is the special case \({\sigma }=0\). A similar argument holds in the case \({\zeta }<0\), with the roles of \({\kappa }\) and \({\kappa }_0\) interchanged, which leads to the same estimate. It remains to show the third estimate, which we will do for general \({\sigma }\geqslant 0\). We have
where we redefine
Now
so
On the other hand, since \(T\leqslant t\),
The claimed estimate follows. \(\square \)
6.2 Gaussian limit process
By Proposition 6.1, in order to compute the weak limit of \(c^{-1/2}({\hat{\Psi }}_t,\Psi ^{{\text {cap}}}_t)_{t<t_{\zeta }}\), it suffices to compute the weak limit of \(c^{-1/2}({\hat{\Pi }}_t,\Pi ^{{\text {cap}}}_t)_{t<t_{\zeta }}\). This process is a deterministic linear function of the compensated Poisson random measure \({\tilde{\mu }}\). We are guided to find the weak limit process by replacing \({\tilde{\mu }}\) in (91) and (92) by a Gaussian white noise on \([0,2\pi )\times [0,\infty )\times (0,\infty )\) of the same intensity. At the same time, we set \({\sigma }=0\) in the limit,Footnote 2 replacing the multiplier operator \(P({\tau })\) by the corresponding operator \(P_0({\tau })\) when \({\sigma }=0\). Then, using the scaling properties of white noise, we arrive at candidate limit processes \(({\hat{{\Gamma }}}_t(z))_{t<t_{\zeta }}\) and \(({\Gamma }^{{\text {cap}}}_t)_{t<t_{\zeta }}\) which are defined as follows. Let W be a Gaussian white noise on \([0,2\pi )\times (0,\infty )\) of intensity \((2\pi )^{-1}d{\theta }dt\). Define for each \(|z|>1\) and \(t\in [0,t_{\zeta })\)
where these Gaussian integrals are understood by the usual \(L^2\) isometry. Define for \(t\geqslant 0\) and \(k\geqslant 0\)
We can and do choose versions of \((B_t(k))_{t\geqslant 0}\) and \((B_t)_{t\geqslant 0}\) which are continuous in t. Then \((B_t(k))_{t\geqslant 0}\) is a complex Brownian motion for all k, \((B_t)_{t\geqslant 0}\) is a real Brownian motion, and all these processes are independent. Note that, almost surely, for all \(t<t_{\zeta }\),
Define for \(t\in [0,t_{\zeta })\) and \(k\geqslant 0\)
The following estimate may be obtained by (a simpler version of) the argument used for Proposition 6.2.
Proposition 6.3
For all \({\alpha },\eta \in \mathbb {R}\) with \({\alpha }+\eta ={\zeta }\leqslant 1\), and all \(t<t_{\zeta }\), there is a constant \(C({\alpha },\eta ,t)<\infty \) such that, for all \(k\geqslant 0\),
The following identity holds in \(L^2\) for all \(|z|>1\) and \(t<t_{\zeta }\)
By Proposition 6.3, almost surely, the right-hand side in (94) converges uniformly on compacts in \([0,t_{\zeta })\), uniformly on \(\{|z|\geqslant r\}\), for all \(r>1\). So we can and do use (94) to choose a version of \({\hat{{\Gamma }}}_t(z)\) for each \(t<t_{\zeta }\) and \(|z|>1\) such that \(({\hat{{\Gamma }}}_t)_{t<t_{\zeta }}\) is a continuous process in \({{\mathcal {H}}}\) and (94) holds for all \({\omega }\).
The processes \(({\Gamma }_t(k))_{t<t_{\zeta }}\) and \(({\Gamma }^{{\text {cap}}}_t)_{t<t_{\zeta }}\) are also characterized by the following Ornstein–Uhlenbeck-type stochastic differential equations
These equations can be put in a simpler form by switching to the time-scale
which arises as the limit as \(c\rightarrow 0\) of a time-scale where particles arrive at a constant rate. Write \(\nu \mapsto t(\nu ):[0,n_{\alpha })\rightarrow [0,t_{\zeta })\) for the inverse map and set
and
Then \(e^{-{\alpha }{\tilde{{\tau }}}_\nu }=(1+{\alpha }\nu )^{-1}\). Also \(({\tilde{B}}_\nu (k))_{\nu <n_{\alpha }}\) is a complex Brownian motion for all k, \(({\tilde{B}}_\nu )_{\nu <n_{\alpha }}\) is a real Brownian motion, and these processes are independent. Then we have
We can define a Brownian motion \(({\tilde{B}}_\nu )_{\nu <n_{\alpha }}\) in \({{\mathcal {H}}}\) by
Set
On summing the equations (95), we see that \(({\tilde{{\Gamma }}}_\nu )_{\nu <n_{\alpha }}\) satisfies the following stochastic integral equation in \({{\mathcal {H}}}\)
6.3 Convergence
Given Proposition 6.1, the following result will complete the proof of Theorem 1.3.
Proposition 6.4
For all \({\alpha },\eta \in \mathbb {R}\) with \({\alpha }+\eta ={\zeta }\leqslant 1\) and all \(T<t_{\zeta }\), we have
weakly in \({{D([0,T],{{\mathcal {H}}}\times \mathbb {R})}}\) as \(c\rightarrow 0\) and \({\sigma }\rightarrow 0\) as in Theorem 1.3.
Proof
By Proposition 6.2, it will suffice to show the claimed limit with \(({\hat{\Pi }}_t)_{t\leqslant T}\) replaced by \(({\hat{\Pi }}^0_t)_{t\leqslant T}\). We first show that
in the sense of finite-dimensional distributions. For all \(n\geqslant 1\), all \(k_1,\dots ,k_n\geqslant 0\) and all \(t_1,\dots ,t_n\leqslant T\), any real-linear function of \(c^{-1/2}(\Pi ^0_{t_j}(k_j),\Pi ^{{\text {cap}}}_{t_j}:j=1,\dots ,n)\) can be written in the form
where
and \(({\theta },t)\mapsto f_t({\theta }):[0,2\pi )\times (0,T]\rightarrow \mathbb {R}\) is bounded, measurable and independent of c. Set
The same linear function applied to \(({\Gamma }_{t_j}(k_j),{\Gamma }^{{\text {cap}}}_{t_j}:j=1,\dots ,n)\) gives the random variable
Then
and, using the Campbell–Hardy formula, as \(c\rightarrow 0\),
The claimed convergence of finite-dimensional distributions follows, by convergence of characteristic functions.
Now, Proposition 6.2 shows that the processes \((\Pi ^0_t(k))_{t\leqslant T}\) and \((\Pi ^{{\text {cap}}}_t)_{t\leqslant T}\) all satisfy Aldous’s tightness criterion in \(D([0,T],\mathbb {C})\). Hence
weakly in \(D([0,T],\mathbb {C}^{\mathbb {Z}^+}\times \mathbb {R})\) as \(c\rightarrow 0\). Hence, for all \(K\geqslant 0\),
weakly in \(D([0,T],{{\mathcal {H}}}\times \mathbb {R})\) as \(c\rightarrow 0\), where, for \(f(z)=\sum _{k=0}^\infty a_kz^{-k}\),
For \(|z|=r\), we have
Hence, it will suffice to show, for \(r>1\) and all \({\varepsilon }>0\), that
But, since \({\alpha }+\eta ={\zeta }\leqslant 1\), by Proposition 6.2, there is a constant \(C({\alpha },\eta ,T)<\infty \) such that, for all \(c>0\) and all \(r>1\),
The desired limit follows. \(\square \)
Proof of Theorem 1.4
We will argue via the Skorokhod representation theorem. It will suffice to show the claimed convergence for all sequences \(c_k\rightarrow 0\) and \({\sigma }_k\rightarrow 0\) subject to the constraint assumed in Theorem 1.3. Given \(N<n_{\alpha }\), choose \({\delta }>0\) and \(T<t_{\zeta }\) such that \(\nu _T=N+{\delta }\), as in the proof of Theorem 1.2. By Theorem 1.3 and Propositions 5.5 and 5.7, and since \(D([0,T],{{\mathcal {H}}})\) is a complete separable metric space, there is a probability space on which are defined a sequence of ALE\(({\alpha },\eta )\) processes \((\Phi _t^{(k)})_{t\geqslant 0}\), with common particle family \((P^{(c)}:c\in (0,\infty ))\), and a Gaussian process \(({\hat{{\Gamma }}}_t)_{t<t_{\zeta }}\) with the following properties:
-
(a)
\((\Phi _t^{(k)})_{t\geqslant 0}\) has capacity parameter \(c_k\) and regularization parameter \({\sigma }_k\),
-
(b)
\(({\hat{{\Gamma }}}_t)_{t<t_{\zeta }}\) has the distribution of the limit Gaussian process in Theorem 1.3,
-
(c)
almost surely, as \(k\rightarrow \infty \),
$$\begin{aligned} \sup _{t\leqslant T}|c{{\mathcal {V}}}^{(k)}_t-\nu _t|\rightarrow 0 \end{aligned}$$and, for all \(r>1\),
$$\begin{aligned} \sup _{t\leqslant T}\sup _{|z|\geqslant r}\left| c^{-1/2}{\hat{\Psi }}^{(k)}_t(z)-{\hat{{\Gamma }}}_t(z)\right| \rightarrow 0. \end{aligned}$$
Here, \({{\mathcal {V}}}^{(k)}_t\) denotes the number of particles added in \((\Phi _t^{(k)})_{t\geqslant 0}\) by time t. Define for \(n\geqslant 0\) and \(\nu <n_{\alpha }\)
From (c), we deduce that, almost surely, as \(k\rightarrow \infty \),
and, for \(\nu \in [0,N]\) and \(n=\lfloor \nu /c\rfloor \), the following limit holds in \({{\mathcal {H}}}\)
But \(({\hat{{\Gamma }}}_{t(\nu )})_{\nu <n_{\alpha }}\) has the same distribution as \(({\hat{{\Gamma }}}^{{\text {disc}}}_\nu )_{\nu <n_{\alpha }}\). Hence
weakly in \(D([0,N],{{\mathcal {H}}})\). \(\square \)
Data Availability Statement
Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Notes
This is an abuse of terminology since it is then \(e^{{{\text {cap}}}(K_n)}\) which is the logarithmic capacity.
It is not necessary to pass to the limit \({\sigma }\rightarrow 0\). Indeed, the best Gaussian approximation for given \({\sigma }>0\) would be obtained using P instead of \(P_0\). The limit \(c\rightarrow 0\) with \({\sigma }\) fixed then holds uniformly in \({\sigma }\), subject to the restrictions stated in Theorem 1.3, and the limit processes for \({\sigma }\) fixed converge weakly to the case \({\sigma }=0\). We have stated only the joint limit, since this seems to us of main interest, and since the limit fluctuations have in this case a slightly simpler form.
References
Berestycki, N., Silvestri, V.: Explosive growth for a constrained Hastings–Levitov aggregation model. Ark. Mat. 61(1), 41–66 (2023)
Berger, N., Procaccia, E.B., Turner, A.: Growth of stationary Hastings–Levitov. Ann. Appl. Probab. 32(5), 3331–3360 (2022)
Burkholder, D.L.: Distribution function inequalities for martingales. Ann. Probab. 1, 19–42 (1973)
Carleson, L., Makarov, N.: Aggregation in the plane and Loewner’s equation. Commun. Math. Phys. 216(3), 583–607 (2001)
Carleson, L., Makarov, N.: Laplacian path models. J. Anal. Math. 87, 103–150 (2002). (Dedicated to the memory of Thomas H. Wolff)
Darling, R.W.R., Norris, J.R.: Differential equation approximations for Markov chains. Probab. Surv. 5, 37–79 (2008)
Eden, M.: A two-dimensional growth process. In: Proceedings of 4th Berkeley Symposium on Mathematical Statistics and Probability, vol. IV, pp. 223–239. University of California Press, Berkeley, CA (1961)
Grebenkov, D.S., Beliaev, D.: How anisotropy beats fractality in two-dimensional on-lattice diffusion-limited-aggregation growth. Phys. Rev. E 96(4), 042159 (2017)
Gustafsson, B., Teodorescu, R., Vasil’ev, A.: Classical and Stochastic Laplacian growth. Advances in Mathematical Fluid Mechanics, Birkhäuser/Springer, Cham (2014)
Hastings, M.B., Levitov, L.S.: Laplacian growth as one-dimensional turbulence. Physica D 116(1–2), 244 (1998)
Hastings, M.B.: Renormalization theory of stochastic growth. Phys. Rev. E (3) 55(1), 135–152 (1997)
Higgs, F.: SLE scaling limits for a Laplacian random growth model. Ann. Inst. Henri Poincaré Probab. Stat. 58(3), 1712–1739 (2022)
Higgs, F.: Tip growth in a strongly concentrated aggregation model follows local geodesics (2023). arXiv:2304.04417 [math.PR]
Johansson Viklund, F., Sola, A., Turner, A.: Scaling limits of anisotropic Hastings–Levitov clusters. Ann. Inst. Henri Poincaré Probab. Stat. 48(1), 235–257 (2012)
Johansson Viklund, F., Sola, A., Turner, A.: Small-particle limits in a regularized Laplacian random growth model. Commun. Math. Phys. 334(1), 331–366 (2015)
Liddle, G., Turner, A.: Scaling limits and fluctuations for random growth under capacity rescaling. Ann. Inst. Henri Poincaré Probab. Stat. 57(2), 980–1015 (2021)
Liddle, G., Turner, A.: Scaling limits of anisotropic growth on logarithmic time-scales. Electron. J. Probab. 28, 75 (2023)
Lundberg, E., Teodorescu, R.: Dima Khavinson’s 60th: a summary of open problems in analysis and mathematical physics. Anal. Math. Phys. 8(2), 309–324 (2018)
Miller, J., Sheffield, S.: Quantum Loewner evolution. Duke Math. J. 165(17), 3241–3378 ((2016)
Niemeyer, L., Pietronero, L., Wiesmann, H.J.: Fractal dimension of dielectric breakdown. Phys. Rev. Lett. 52, 1033–1036 (1984)
Norris, J., Silvestri, V., Turner, A.: Scaling limits for planar aggregation with subcritical fluctuations. Probab. Theory Related Fields 185(1–2), 185–250 (2023)
Norris, J., Turner, A.: Hastings–Levitov aggregation in the small-particle limit. Commun. Math. Phys. 316, 809–841 (2012)
Pommerenke, Ch.: Boundary Behaviour of Conformal Maps. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 299. Springer-Verlag, Berlin (1992)
Rohde, S., Zinsmeister, M.: Some remarks on Laplacian growth. Topol. Appl. 152(1–2), 26–43 (2005)
Shraiman, B., Bensimon, D.: Singularities in nonlocal interface dynamics. Phys. Rev. A (3) 30(5), 2840–2842 (1984)
Silvestri, V.: Fluctuation results for Hastings–Levitov planar growth. Probab. Theory Related Fields 167(1–2), 417–460 (2017)
Sola, A., Turner, A., Viklund, F.: One-dimensional scaling limits in a planar Laplacian random growth model. Commun. Math. Phys. 371(1), 285–329 (2019)
Witten, T.A., Sander, L.M.: Diffusion-limited aggregation, a kinetic critical phenomenon. Phys. Rev. Lett. 47(19), 1400–1403 (1981)
Zygmund, A.: Trigonometric Series. Vol. I, II. Cambridge Mathematical Library. Cambridge University Press, Cambridge (1988). (Reprint of the 1979 edition)
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by R. Seiringer
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Amanda Turner acknowledges partial support by EPSRC grant EP/T027940, Vittoria Silvestri acknowledges partial support from Indam – GNAMPA.
Miscellaneous Estimates
Miscellaneous Estimates
1.1 Explosion in continuous-time ALE(\({\alpha },\eta \))
In this paper we adopted a continuous-time formulation of ALE. When the cluster is in state \(\phi \), we add a particle with harmonic measure coordinate \({\theta }\in [0,2\pi )\) at rate \(c^{-1}|\phi (e^{{\sigma }+i{\theta }})|^{-\eta }d{\theta }/(2\pi )\). In previous works, this process had been considered in discrete time, that is, jump by jump. Besides being mathematically convenient, the continuous-time formulation has physical meaning since it considers the process in the natural physical time-scale. We now determine exactly for which parameter values \({\alpha }\) and \(\eta \) there is pathwise explosion for ALE. The definition and running assumptions (5) and (6) for ALE(\({\alpha },\eta \)) are given in Sect. 1.2.
Proposition A.1
Let \((\Phi _t)_{t\geqslant 0}\) be an ALE(\({\alpha },\eta \)) process. Denote by \(({{\mathcal {T}}}_t)_{t\geqslant 0}\) the associated process of capacities and by Z the explosion time of \((\Phi _t)_{t\geqslant 0}\). Then, on the event \(\{Z<\infty \}\), we have \({{\mathcal {T}}}_t\rightarrow \infty \) as \(t\rightarrow Z\). Moreover, if \(\eta \geqslant 0\) or \({\zeta }={\alpha }+\eta \geqslant 0\), then \(Z=\infty \) almost surely, while if \(\eta <0\) and \({\zeta }<0\) then \(Z<\infty \) almost surely.
Proof
The total jump rate \({\lambda }(\phi )\) at a state \(\phi \) is given by
so, by distortion estimates, there is a constant \(C(\eta ,{\sigma })<\infty \) such that
where \({\tau }=\phi '(\infty )\). Similarly, there is a constant \(C({\alpha },{\sigma })<\infty \) such that the next jump in capacity \(\Delta {\tau }\) satisfies
The upper bound in the first estimate implies that the jump rate is bounded if \(\eta \geqslant 0\), and is bounded on compacts in \({\tau }\) if \(\eta <0\). Hence \(Z=\infty \) almost surely if \(\eta \geqslant 0\), and if \(\eta <0\) then \(Z<\infty \) only if \({{\mathcal {T}}}_t\rightarrow \infty \) as \(t\rightarrow Z\). Moreover, using also the upper bound of the second estimate, we see that
is a local supermartingale up to Z. Hence \(Z=\infty \) almost surely if \({\zeta }\geqslant 0\).
It remains to show that \(Z<\infty \) almost surely in the case when \(\eta <0\) and \({\zeta }<0\). For this we use the lower bounds in the estimates above. Set
It will be convenient to choose \(C\geqslant {\alpha }c\) so that F is increasing on \([0,\infty )\). We know that \({{\mathcal {T}}}_t\) jumps up by at least \({\delta }({{\mathcal {T}}}_{t-})\) at rate at least \({\lambda }({{\mathcal {T}}}_{t-})\). Consider the Markov chain \((X_t)_{t\geqslant 0}\) starting from 0 which jumps up by \({\delta }(X_{t-})\) at rate \({\lambda }(X_{t-})\). Since \(\eta <0\), for as long as \(X_{t-}\leqslant {{\mathcal {T}}}_{t-}\), we may couple these processes so that \((X_t)_{t\geqslant 0}\) jumps whenever \(({{\mathcal {T}}}_t)_{t\geqslant 0}\) does. But \(X_0={{\mathcal {T}}}_0=0\) and at each jump of \(X_t\) we have \(X_t=F(X_{t-})\) and \({{\mathcal {T}}}_t\geqslant F({{\mathcal {T}}}_{t-})\). Since F is increasing, the inequality \(X_t\leqslant {{\mathcal {T}}}_t\) extends to all \(t<Z\). It will therefore suffice to show that \((X_t)_{t\geqslant 0}\) explodes. Now the sequence of values \((x(n):n\geqslant 0)\) taken by \((X_t)_{t\geqslant 0}\) is given by
and the holding times of \((X_t)_{t\geqslant 0}\) are independent exponential random variables of parameters \(({\lambda }(x(n)):n\geqslant 0)\). Hence \((X_t)_{t\geqslant 0}\) explodes if and only if
Note that, if \({\alpha }\leqslant 0\), then \(x(n)\geqslant cn/C\) for all n so
Assume then that \({\alpha }>0\). For \(x\geqslant 0\), define \((\psi _t(x):t\geqslant 0)\) by
We can solve to obtain
Since \({\alpha }>0\) and \(\psi _t(x)\) is increasing in t,
Since \(\psi _t(x)\) is increasing in x, it follows by induction that, for all n,
Hence
But \({\alpha }+\eta ={\zeta }<0\) so \(\eta /{\alpha }<-1\) and so
as required. \(\square \)
1.2 Estimates for single-particle maps
Let P be a basic particle and let
be the associated conformal map \(D_0\rightarrow D_0\setminus P\). We assume that F extends continuously to \(\{|z|\geqslant 1\}\). Set
We assume throughout that \({\delta }\leqslant 1\). We use the following well known estimates on the capacity c. There is an absolute constant \(C<\infty \) such that
The lower bound relies on Beurling’s projection theorem and a comparison with the case of a slit particle. The upper bound follows from a comparison with the case \(P_{\delta }=S_{\delta }\cap D_0\), where \(S_{\delta }\) is the closed disk whose boundary intersects the unit circle orthogonally at \(e^{\pm i{\theta }_{\delta }}\) with \({\theta }_{\delta }\in [0,\pi ]\) is determined by \(|e^{i{\theta }_{\delta }}-1|={\delta }\). See Pommerenke [23].
Write
where we understand the argument to be determined for each \(z\in D_0\) so that the left-hand side is holomorphic in \(D_0\) and such that \(v(z)\rightarrow 0\) as \(z\rightarrow \infty \). Then u and v are bounded and harmonic in \(D_0\), with continuous extensions to \(\{|z|\leqslant 1\}\), and \(u(z)\rightarrow c\) as \(z\rightarrow \infty \). Note also that
Lemma A.2
Assume that \(16{\delta }\leqslant \pi \). Then
and
Proof
Set
where B is a complex Brownian motion. Consider the conformal map f of \(D_0\) to the upper half-plane \(\mathbb {H}\) given by
Set \(b=f(e^{-i{\theta }_{\delta }})=\sin {\theta }_{\delta }/(1+\cos {\theta }_{\delta })\). Since \({\delta }\leqslant 1\), we have
and then \(b\leqslant 2\pi {\delta }/9\). By conformal invariance,
Hence
Now \(e^{i\pi }\) is not a limit point of P so \(e^{i\pi }=F(e^{i(\pi +{\alpha })})\) for some \({\alpha }\in \mathbb {R}\). Then \(u(e^{i(\pi +{\alpha })})=0\) and we can and do choose \({\alpha }\) so that \({\alpha }+v(e^{i(\pi +{\alpha })})=0\). Set
Then \({\theta }^-\leqslant {\theta }^+\). We will show that \(|{\theta }^\pm |\leqslant 16{\delta }\), which then implies (98). For \({\theta }\in [{\theta }^-,{\theta }^+]\), we have \(F(e^{i{\theta }})\in S_{\delta }\) so \(|{\theta }+v(e^{i{\theta }})|\leqslant {\theta }_{\delta }\). Set \(P^*=\{F(e^{i{\theta }}):{\theta }\in [{\theta }^-,{\theta }^+]\}\). Then \(P^*\subseteq S_{\delta }\) so, by conformal invariance,
On the other hand, for \({\theta },{\theta }'\in [{\theta }^+,{\theta }^-+2\pi ]\) with \({\theta }\leqslant {\theta }'\), by conformal invariance,
so v is non-decreasing on \([{\theta }^+,{\theta }^-+2\pi ]\), and so
Hence
and similarly
So we obtain, for all \({\theta }\in [{\theta }^-,{\theta }^+]\),
Since v is continuous and is non-decreasing on the complementary interval, this inequality then holds for all \({\theta }\). Now v is bounded and harmonic in \(D_0\) with limit 0 at \(\infty \), so
Hence
On combining this with (102), (103) and (104), we see that
But \(4{\theta }_{\delta }+4\pi p_{\delta }\leqslant 44\pi {\delta }/9\leqslant 16{\delta }\) by (100) and (101), so we have shown the claimed inequalities. \(\square \)
Proposition A.3
There is an absolute constant \(C<\infty \) with the following properties. In the case where \({\delta }={\delta }(P)\leqslant 1/C\), for all \(|z|>1\),
and, for all \(|z|>1\) with \(|z-1|\geqslant C{\delta }\),
and
and
Proof
Since \(z\log (F(z)/z)\) is bounded and holomorphic in \(\{|z|>1\}\), (105) follows from (97) and (99) by the maximum principle. The inequality (107) follows from (106) on letting \(z\rightarrow \infty \), since \(z(\log (F(z)/z)-c)\rightarrow a_0\). Moreover, since \((z-1)^2(\log (F(z)/z)-c)-a_0z\) is bounded and holomorphic on \(\{|z|>1\}\), (108) follows from (106) by the maximum principle, at the cost of replacing C by 6C, say. We will show (106) holds whenever \(|z-1|\geqslant 3a\), where \(a=16{\delta }\).
Since u is bounded and harmonic with \(u(z)\rightarrow c\) as \(z\rightarrow \infty \), we have
and, for all \(|z|>1\),
Let \({\alpha }\in (-\pi ,\pi ]\) and \(\rho >0\) be defined by
We use (98) to see that \(|{\alpha }|\leqslant a\) and \(\rho \in [\cos a,1)\). Now
For \(|z-1|\geqslant 2a\) and any \({\theta }\) such that \(u(e^{i{\theta }})>0\), we have
Hence, for \(|z|>1\) with \(|z-1|\geqslant 2a\),
and
and hence
We can extend F to a holomorphic function in \(\{|z-1|>a\}\) by setting \(F({\bar{z}}^{-1})=\overline{F(z)}^{-1}\). Then u and v also extend and it is straightforward to check that the estimate (109) remains valid for all \(|z-1|\geqslant 2a\). Since \(v(z)\rightarrow 0\) as \(z\rightarrow \infty \), a standard argument allows us to deduce from (109) that, for \(|z-1|\geqslant 3a\),
and hence
\(\square \)
We sometimes use exponentiated versions of the inequalities just proved, which are straightforward to deduce and are noted here for easy reference. There is an absolute constant \(C<\infty \) with the following properties. Suppose that \({\delta }\leqslant 1/C\). Then, for all \(|z|>1\),
and, in the case \(|z-1|\geqslant C{\delta }\),
and
Proposition A.4
There is an absolute constant \(C<\infty \) with the following properties. Let \(P_1,P_2\) be basic particles with \(P_1\subseteq P_2\). For \(i=1,2\), write \(F_i\) for the associated conformal map \(D_0\rightarrow D_0\setminus P_i\) and write \(c_i\) for the capacity of \(P_i\). Set \({\delta }_i={\delta }(P_i)\) and \(a_{0,i}=a_0(P_i)\) and set
Assume that \({\delta }_2\leqslant 1/C\). Then
and, for all \(|z|>1\) with \(|z-1|\geqslant C{\delta }_2\),
and
Proof
The inequalities (112) and (113) follow from (111) by the same argument used to deduce (107) and (108) from (106). Set \({\tilde{P}}=F_1^{-1}(P_2\setminus P_1)\). Write \({\tilde{F}}\) for the associated conformal map \(D_0\rightarrow D_0\setminus {\tilde{P}}\) and write \({\tilde{c}}\) for the capacity of \({\tilde{P}}\). Then
Note that, for \(z\in {\tilde{P}}\), we have \(F_1(z)\in P_2\), so \(|F_1(z)-1|\leqslant {\delta }_2\). But \(|e^{-c_1}F_1(z)-z|\leqslant C{\delta }_1\) for all \(|z|>1\) and \(c_1\leqslant C{\delta }_1^2\). Hence \(|z-1|\leqslant C{\delta }_2\) for all \(z\in {\tilde{P}}\) and so
Hence, for C sufficiently large and \({\delta }_2\leqslant 1/C\), for all \(|z|>1\) with \(|z-1|\geqslant C{\delta }_2\),
and in particular
Set \(z_t=z\exp (t\log ({\tilde{F}}(z)/z))\) and \(f(t)=\log (F_1(z_t)/F_1(z))\). Then
so
Now \(|\log ({\tilde{F}}(z)/z)|\leqslant C{\delta }_2\), so \(|z_t-z|\leqslant C{\delta }_2\) for all t. Hence, for C sufficiently large and \(|z-1|\geqslant C{\delta }_2\), we have \(|z_t-1|\geqslant C_0{\delta }_1\) for all t, where \(C_0\) is the constant from Proposition A.3. Then
where we used Cauchy’s integral formula for the second inequality, adjusting the value of C if necessary. On combining these estimates with (114) and (115), we see that
as claimed. \(\square \)
The following is a straightforward consequence of (96) and Propositions A.3 and A.4.
Proposition A.5
Let \((P^{(c)}:c\in (0,1])\) be a family of basic particles and suppose that the associated conformal maps \(F_c\) are given by
Fix \({\Lambda }\in [1,\infty )\) and assume that \({\delta }(c)\leqslant {\Lambda }r_0(c)\) for all c. Then there is a constant \(C({\Lambda })<\infty \) such that, for all \(c\leqslant 1/C\),
and, for all \(|z|>1\),
and
and
Moreover, if \((P^{(c)}:c\in (0,1])\) is nested, then C may be chosen so that, for all \(c_1,c_2\in (0,c]\),
and, for all \(|z-1|\geqslant C\sqrt{c}\),
and
and
For our final particle estimates, we use the following integral representation for the family of particle maps
for some measurable family of probability measures \((\mu _t:t\in (0,\infty ))\), with \(\mu _t\) supported on \(\{{\theta }:|e^{i{\theta }}-1|\leqslant {\delta }(t)\}\) for all t. This follows from our requirements that the particles \(P^{(c)}\) have capacity c, are contained in \(\{|z-1|\leqslant {\delta }(c)\}\) and are nested, by the Loewner–Kufarev theory. Our condition (6) and the inequality (96) then give a constant \(C({\Lambda })<\infty \) such that
Define holomorphic functions \(L_t\) and \(Q_t\) on \(\{|z|>1\}\) by
where
Note that \(l_t({\theta },z)\rightarrow 1\) and \(q_t({\theta },z)\rightarrow 0\) as \(z\rightarrow \infty \), uniformly in \({\theta }\). It is then straightforward to show the integral representations
Proposition A.6
There is a constant \(C({\Lambda })<\infty \) with the following property. For all \(t\leqslant 1/C\) and all \(|z|>1\),
Proof
We give the details for the second estimate, leaving the first which is similar but simpler to the reader. We split \(q_t({\theta },z)=g_t({\theta },z)+h_t({\theta },z)\), where
Now
so, on the support of \(\mu _t\), we have, for \(|z-1|\geqslant 2C\sqrt{t}\),
where C is the constant in (116). On the other hand, we showed above that, for all \(|z|>1\),
and \(F_t\) extends by reflection to a holomorphic function on \(\{|z-1|>C\sqrt{t}\}\) satisfying the same inequality. Hence, by Cauchy’s integral formula, for \(|z-1|\geqslant 2C\sqrt{t}\),
and so, for \({\theta }\) in the support of \(\mu _t\),
We have shown that, for all \(|z-1|\geqslant C\sqrt{t}\),
\(\square \)
1.3 Operator inequalities
Recall that, for a measurable function f on \(\{|z|>1\}\), for \(p\in [1,\infty )\) and \(r>1\), we set
Suppose that f is holomorphic and is bounded at \(\infty \). It is standard that, for \(\rho \in (1,r)\),
Moreover, there is an absolute constant \(C<\infty \) such that
where \(Df(z)=zf'(z)\). The function f has a Laurent expansion
Let M be an operator which acts as multiplication by \(m_k\) on the the kth Laurent coefficient. Thus
Assume that there exists a finite constant \(A>0\) such that, for all \(k\geqslant 0\),
and, for all integers \(K\geqslant 0\),
Then, by the Marcinkiewicz multiplier theorem [29, Vol. II, Theorem 4.14], for all \(p\in (1,\infty )\), there is a constant \(C=C(p)<\infty \) such that, for all \(r>1\),
We will use also the following estimate.
Proposition A.7
Let f and g be holomorphic in \(\{|z|>1\}\) and bounded at \(\infty \). Let M be a multiplier operator and let \(p\geqslant 2\). Set
and
Then, for all \(r,\rho >1\), we have
and
where
Proof
The second estimate is straightforward and is left to the reader. For the first, we can write
Then
so
where \(\tau _kg(z)=z^{-k}g(z)\). Hence
\(\square \)
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Norris, J., Silvestri, V. & Turner, A. Stability of Regularized Hastings–Levitov Aggregation in the Subcritical Regime. Commun. Math. Phys. 405, 74 (2024). https://doi.org/10.1007/s00220-024-04960-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00220-024-04960-5