Large-time statistics of diffusion-limited aggregation: turbulent deterministic growth versus fluctuations, and the controlled derivation of DLA scaling exponents

In 1985 Parisi and Zhang formulated continuum field theories for the Eden model and for diffusion-limited aggregation (DLA), the second an intrinsically nonlocal reggeon-type field theory, and posed the open questions of their large-time behavior. The Eden and propagating-front part of that agenda was later absorbed by adjacent literatures: interface fluctuations became the Kardar-Paris-Zhang problem, and the front diffusion constant whose collective-coordinate (zero-mode) computation the paper left undone was derived for noisy reaction-diffusion fronts in 2001. The DLA core survives open four decades later: it is not known whether the asymptotic statistics of stochastic Laplacian growth are governed by deterministic chaotic (turbulent) classical solutions, with microscopic noise only regularizing finite-time singularities at small scales, or by a fluctuation-dominated statistical measure on growth trajectories that changes averaged observables; and the DLA scaling exponents, including the planar fractal dimension (numerically about 1.71) and the harmonic-measure spectrum, have no controlled first-principles derivation, the spherically symmetric mean-field solution being unstable.

Background

Diffusion-limited aggregation (DLA), introduced by Witten and Sander in 1981, is a minimal stochastic growth rule: random walkers diffuse in the ambient space and irreversibly stick to a growing cluster at first contact. The probability that the next particle attaches within a boundary element is proportional to the flux of the Laplace equation's solution there; this flux measure is called the harmonic (growth) measure, and DLA is the discrete archetype of stochastic Laplacian growth. The Eden model grows instead by adding particles uniformly at random to the cluster boundary, producing asymptotically compact clusters with mildly roughened interfaces. Despite the simplicity of its rule, two-dimensional DLA is a tenuous fractal: its mass grows with linear size as a power law whose exponent, the fractal dimension, is numerically close to $1.71$; its boundary measure is strongly intermittent; and no controlled analytic theory of its asymptotic form exists.

Parisi and Zhang ('Field theories and growth models', 1985) gave both models a continuum field-theoretic formulation. For Eden they wrote a hierarchy of evolution equations for the joint occupation probabilities $\rho_n(x_1,\ldots,x_n;t)$ of cluster sites and showed that it is reproduced by correlation functions of the reggeon field theory (RFT), a field theory originating in high-energy scattering whose local-interaction form also describes directed percolation, i.e. branching and annihilating random walks. Under the mean-field closure $\rho_2\approx\rho_1^2$ the lowest equation becomes a nonlinear reaction-diffusion equation of Fisher-KPP type (diffusion with linear growth and quadratic saturation),

$$\dot{\rho}_1=\alpha\Delta\rho_1+\mu^2\rho_1-g^2\rho_1^2,$$

whose traveling fronts are selected by the marginal-stability criterion of Dee and Langer (1983): the operator obtained by linearizing about the moving kink has a zero (marginal) eigenvalue. Fluctuations make the front position wander diffusively, and this same zero mode makes naive perturbation theory for the two-point function singular, so a collective-coordinate treatment, which promotes the front position to an independent dynamical variable, is required to compute the front diffusion constant. For DLA, the capture of diffusing walkers makes the continuum theory intrinsically nonlocal: the evolution involves inverse Laplacians (Green's functions of the walker field), giving an RFT with nonlocal spatial interactions. Companion works (Parisi and Zhang 1984; Peliti and Zhang 1985; Shapir and Zhang 1985) developed the same program, with Peliti and Zhang explicitly noting the difficulty of applying renormalization-group (RG) methods to aggregation processes.

The paper then identified what the formalism could not do: for DLA the spherically symmetric mean-field solution is unstable (the standard Laplacian-growth instability), so mean-field predictions fail in low dimensions where clusters are fractal; the role of fluctuations in the asymptotic regime was posed as a dichotomy between deterministic nonspherical 'turbulent' classical solutions and a fluctuation-dominated statistics that weights classical trajectories differently; no RG computation of the DLA scaling exponents existed; and the authors suggested importing turbulence-theory ideas (Kolmogorov-law analogues, that is, exact scaling relations among averaged observables analogous to Kolmogorov's inertial-range laws for turbulence) into DLA.

Much of the Eden and front side of this agenda was subsequently absorbed by other frameworks. The Kardar-Parisi-Zhang (KPZ) equation, a nonlinear stochastic partial differential equation for a growing interface height field (Kardar, Parisi and Zhang 1986), cites the source work and reframed the fluctuations of growing Eden-type interfaces; that problem is now understood in terms of KPZ universality. The general theory of fronts propagating into unstable states (van Saarloos 1988) systematized velocity selection, and the front diffusion coefficient of noisy reaction-diffusion fronts, the very quantity whose zero-mode computation Parisi and Zhang left undone, was derived by Rocco, Casademunt, Ebert and van Saarloos (2001) via an eigenmode (collective-coordinate) decomposition of the fluctuating front. Rigorous results also exist for Eden-type growth in the conformal-map regularization of Hastings and Levitov (Norris and Turner 2012; Berger, Procaccia and Turner 2022).

The DLA side did not close. The most direct continuation of the turbulence analogy is the Hastings-Levitov program ('Laplacian growth as one-dimensional turbulence', 1998): Laplacian growth is regularized by a random sequence of iterated conformal maps with a size parameter $\alpha$, exhibiting a sharp transition at $\alpha=1$ between stable growth and noise-driven 'turbulent' growth, with sharp boundary features cascading from macroscopic to microscopic scales in analogy with the inertial range of turbulence (the intermediate range of scales between the large scales where energy enters and the small viscous scales where it dissipates); at $\alpha=2$, where the attached area stays constant as in DLA, the authors find numerically that the two-dimensional bump model has properties identical to DLA clusters and argue the two are equivalent, while noting that the precise correspondence to lattice DLA needs further exploration. These authors conjectured that the large-scale DLA dynamics is equivalent to deterministic chaotic Laplacian contour dynamics (the zero-noise continuum limit), with the deterministic finite-time singularities resolved by noise only at the microscale, a concrete realization of the 'turbulent classical solutions' scenario. Analytical renormalization-group work in this framework remains uncontrolled at the DLA point: Hastings's renormalization group (1997) is a lowest-order resummed perturbation theory in the interaction and noise strengths, applied at the DLA value of the model's parameters and yielding $D=1.7$, while his later treatment (2001) is an expansion near the upper critical value $\eta_c=4$ of the dielectric-breakdown model, a family of Laplacian growth models in which the growth probability scales as a power $\eta$ of the local field (DLA is the case $\eta=1$; at $\eta_c=4$ clusters become one-dimensional). Kinetic and real-space RG schemes (Wang, Shapir and Rubinstein 1989; the fixed-scale transformation, a real-space scheme that builds up the growth probabilities of a cluster scale by scale, reviewed by Erzan, Pietronero and Vespignani 1995), small-cluster renormalization (Davidovitch, Levermann and Procaccia 2000), and coupled-field continuum models (Keblinski et al. 1994) yield approximate estimates of the DLA exponents only. A 2026 assessment (Halsey) states that 'in no case is there an exact solution, or even a controllable approximation that can practically yield high precision results', and exact structural relations in the Hastings-Levitov framework, such as the recently proven amplitude refinement of the $\tau(3)=D$ relation between the third moment of the multifractal harmonic-measure spectrum $\tau(q)$ (the scaling exponents of the $q$-th moment of the growth probability) and the cluster fractal dimension $D$, connect observables to $D$ but take $D$ as input rather than output.

Problem Statement

The problem is to establish a controlled analytic description of the asymptotic (large-time, large-mass) statistics of diffusion-limited aggregation, resolving the surviving core of the questions posed by Parisi and Zhang (1985) for DLA.

(1) Fluctuation regime. Determine whether the large-scale statistics of DLA are governed by deterministic chaotic ('turbulent') nonspherical solutions of the continuum Laplacian-growth dynamics, with the microscopic stochasticity irrelevant at large scales and its role confined to regularizing the finite-time singularities of the deterministic flow at the microscale, as conjectured by Hastings and Levitov, or whether stochastic fluctuations induce a genuinely different statistical measure on the space of classical growth trajectories and thereby change averaged observables, and how the two regimes differ in the scaling with cluster size of the fluctuation corrections to averaged observables such as the cluster radius, growth rate, and harmonic-measure moments.

(2) Exponents. Derive, in a controlled way, the critical scaling exponents of DLA, at minimum the cluster fractal dimension $D$ (numerically close to $1.71$ in two dimensions) and the multifractal spectrum $\tau(q)$ of the harmonic measure (the scaling exponents of the $q$-th moment of the growth probability), from the microscopic growth rule or from a continuum or field-theoretic formulation, replacing the unstable spherically symmetric mean-field solution. A controlled numerical determination of $D$ or of $\tau(q)$, however precise, does not by itself resolve this part. Exact structural relations (Makarov's theorem, which states that the harmonic measure of any simply connected planar domain is concentrated on a boundary set of Hausdorff dimension one; the $\tau(3)=D$ relation and its amplitude refinements) may be used but do not by themselves answer the question, since they take $D$ as input; the Kolmogorov-law analogues suggested in the source work (exact scaling relations among averaged observables, analogous to Kolmogorov's inertial-range laws for turbulence) do not by themselves answer the question either.

Related context: the validity and limits of the continuum approximations and truncations used for DLA bear directly on both parts of the problem, since (1) fixes the regime and (2) computes its universal numbers within whatever continuum description is justified.

References

1. Giorgio Parisi and Zhang Yi-Cheng, "Field theories and growth models", Journal of Statistical Physics 41, 1-16 (1985). DOI: 10.1007/BF01020601, https://doi.org/10.1007/BF01020601
2. Giorgio Parisi and Yi-Cheng Zhang, "Eden Model in Many Dimensions", Physical Review Letters 53, 1791-1794 (1984). DOI: 10.1103/PhysRevLett.53.1791, https://doi.org/10.1103/PhysRevLett.53.1791
3. Luca Peliti and Yi-Cheng Zhang, "Field theory approach to the Eden model and diffusion-limited aggregation", Journal de Physique Lettres 46, 1151-1157 (1985). DOI: 10.1051/jphyslet:0198500460240115100, https://doi.org/10.1051/jphyslet:0198500460240115100
4. Y. Shapir and Yi-Cheng Zhang, "Systematic Hamiltonian approach to Eden model of cluster growth", Journal de Physique Lettres 46, 529-533 (1985). DOI: 10.1051/jphyslet:019850046012052900, https://doi.org/10.1051/jphyslet:019850046012052900
5. Mehran Kardar, Giorgio Parisi, and Yi-Cheng Zhang, "Dynamic Scaling of Growing Interfaces", Physical Review Letters 56, 889-892 (1986). DOI: 10.1103/PhysRevLett.56.889, https://doi.org/10.1103/PhysRevLett.56.889
6. T. A. Witten and L. M. Sander, "Diffusion-Limited Aggregation, a Kinetic Critical Phenomenon", Physical Review Letters 47, 1400-1403 (1981). DOI: 10.1103/PhysRevLett.47.1400, https://doi.org/10.1103/PhysRevLett.47.1400
7. G. Dee and J. S. Langer, "Propagating Pattern Selection", Physical Review Letters 50, 383-386 (1983). DOI: 10.1103/PhysRevLett.50.383, https://doi.org/10.1103/PhysRevLett.50.383
8. Wim van Saarloos, "Front propagation into unstable states: Marginal stability as a dynamical mechanism for velocity selection", Physical Review A 37, 211-229 (1988). DOI: 10.1103/PhysRevA.37.211, https://doi.org/10.1103/PhysRevA.37.211
9. Andrea Rocco, Jaume Casademunt, Ute Ebert, and Wim van Saarloos, "Diffusion coefficient of propagating fronts with multiplicative noise", Physical Review E 65, 012102 (2001). DOI: 10.1103/PhysRevE.65.012102, https://doi.org/10.1103/PhysRevE.65.012102
10. X. R. Wang, Y. Shapir, and M. Rubinstein, "Improved kinetic renormalisation group approach to diffusion-limited aggregation", Journal of Physics A: Mathematical and General 22, L507-L512 (1989). DOI: 10.1088/0305-4470/22/11/010, https://doi.org/10.1088/0305-4470/22/11/010
11. A. Erzan, L. Pietronero, and A. Vespignani, "The fixed-scale transformation approach to fractal growth", Reviews of Modern Physics 67, 545-604 (1995). DOI: 10.1103/RevModPhys.67.545, https://doi.org/10.1103/RevModPhys.67.545
12. Pawel Keblinski, Amos Maritan, Flavio Toigo, and Jayanth R. Banavar, "Continuum approach to diffusion-limited-aggregation type of growth", Physical Review E 49, R4795-R4798 (1994). DOI: 10.1103/PhysRevE.49.R4795, https://doi.org/10.1103/PhysRevE.49.R4795
13. M. B. Hastings and L. S. Levitov, "Laplacian growth as one-dimensional turbulence", Physica D: Nonlinear Phenomena 116, 244-252 (1998). arXiv: cond-mat/9607021, DOI: 10.1016/S0167-2789(97)00244-3, https://arxiv.org/abs/cond-mat/9607021
14. Matthew B. Hastings, "Renormalization theory of stochastic growth", Physical Review E 55, 135-152 (1997). DOI: 10.1103/PhysRevE.55.135, https://doi.org/10.1103/PhysRevE.55.135
15. M. B. Hastings, "Growth exponents with 3.99 walkers", Physical Review E 64, 046104 (2001). DOI: 10.1103/PhysRevE.64.046104, https://doi.org/10.1103/PhysRevE.64.046104
16. Benny Davidovitch, Anders Levermann, and Itamar Procaccia, "Convergent calculation of the asymptotic dimension of diffusion limited aggregates: Scaling and renormalization of small clusters", Physical Review E 62, R5919-R5922 (2000). DOI: 10.1103/PhysRevE.62.R5919, https://doi.org/10.1103/PhysRevE.62.R5919
17. Eviatar B. Procaccia and Itamar Procaccia, "Dimension of diffusion-limited aggregates grown on a line", Physical Review E 103, L020101 (2021). DOI: 10.1103/PhysRevE.103.L020101, https://doi.org/10.1103/PhysRevE.103.L020101
18. James R. Norris and Amanda Turner, "Hastings-Levitov Aggregation in the Small-Particle Limit", Communications in Mathematical Physics 316, 809-841 (2012). DOI: 10.1007/s00220-012-1552-6, https://doi.org/10.1007/s00220-012-1552-6
19. Noam Berger, Eviatar B. Procaccia, and Amanda Turner, "Growth of stationary Hastings-Levitov", The Annals of Applied Probability 32, 3331 (2022). DOI: 10.1214/21-AAP1761, https://doi.org/10.1214/21-aap1761
20. Thomas C. Halsey, "Exact amplitude relations for diffusion-limited aggregation", arXiv:2607.02216 (2026). DOI: 10.48550/arXiv.2607.02216, https://arxiv.org/abs/2607.02216
21. Mauricio J. del Razo, Tommaso Lamma, and Wout Merbis, "Field theories and quantum methods for stochastic reaction-diffusion systems", Reviews of Modern Physics 98, 015001 (2026). arXiv: 2409.13377, DOI: 10.1103/9qlw-gyd7, https://arxiv.org/abs/2409.13377
22. N. G. Makarov, "On the distortion of boundary sets under conformal mappings", Proceedings of the London Mathematical Society (3) 51, 369-384 (1985). DOI: 10.1112/plms/s3-51.2.369, https://doi.org/10.1112/plms/s3-51.2.369