Multitype L\'evy trees as scaling limits of multitype Bienaym\'e-Galton-Watson trees
Pith reviewed 2026-05-23 04:05 UTC · model grok-4.3
The pith
Multitype Bienaymé-Galton-Watson trees conditioned to have large same-type subtrees converge to multitype Lévy trees.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Under sufficient mild conditions, a sequence of multitype Bienaymé-Galton-Watson trees conditioned on the size of the maximal subtree of vertices of the same type joined by the root to be large converges to a multitype Lévy tree obtained by gluing single-type Lévy trees together in a method determined by the limiting spectrally positive additive Lévy field.
What carries the argument
The gluing of single-type Lévy trees via the limiting spectrally positive additive Lévy field, using an iterative operation on compact marked metric spaces with vector-valued measures.
If this is right
- Convergence occurs in the Gromov-Hausdorff-Prohorov topology after scaling.
- The multitype Lévy tree is constructed by gluing single-type Lévy trees via the additive Lévy field.
- The approach improves results on convergence of compact marked metric spaces with vector-valued measures.
- Techniques from the single-type case are extended to account for inter-type interactions in the multitype setting.
Where Pith is reading between the lines
- This approach could be used to model evolutionary processes with multiple species or genetic types under large population limits.
- Similar conditioning might apply to other multitype branching models to obtain Lévy tree limits.
- Extensions to infinite variance cases or other Lévy processes could be explored based on the mild conditions used.
Load-bearing premise
The specific conditioning on the size of the maximal same-type subtree being large, together with the mild conditions on the offspring distributions, is sufficient to produce the claimed convergence in the Gromov-Hausdorff-Prohorov topology after scaling.
What would settle it
A numerical or analytical counterexample in which the rescaled multitype Bienaymé-Galton-Watson tree fails to converge to the glued multitype Lévy tree in the Gromov-Hausdorff-Prohorov metric under the stated mild conditions.
Figures
read the original abstract
We establish sufficient mild conditions for a sequence of multitype Bienaym\'e-Galton-Watson trees, conditioned in some sense to be large, to converge to a limiting compact metric space which we call a \emph{multitype L\'{e}vy tree}. More precisely, we condition on the size of the maximal subtree of vertices of the same type joined by the root to be large. While we employ a different conditioning, our result can be seen as a generalization to the multitype setting of the continuum random trees defined by Aldous, Duquesne and Le Gall in [Ald91a,Ald91b,Ald93,DLG02]. Our main result is an invariance principle for the convergence of such trees, by gluing single-type L\'{e}vy trees together in a method determined by the limiting spectrally positive additive L\'{e}vy field, as constructed by Chaumont and Marolleau [CM21]. Our approach is an improvement of a result about the convergence in the Gromov-Hausdorff-Prohorov topology, of compact marked metric spaces equipped with vector-valued measures, which are then glued via an iterative operation. To analyze the gluing operation, we extend the techniques developed by S\'enizergues [Sen19,Sen22] to the multitype setting. While the single-type case exhibits a more homogeneous structure with simpler dependency patterns, the multitype case introduces interactions between different types, leading to a more intricate dependency structure where functionals must account for type-specific behaviors and inter-type relationships.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims that, under sufficient mild conditions on the offspring distributions, a sequence of multitype Bienaymé-Galton-Watson trees conditioned on the size of the maximal same-type subtree (joined by the root) being large converges in the Gromov-Hausdorff-Prohorov topology to a multitype Lévy tree. The limit is constructed by gluing single-type Lévy trees according to the law of a spectrally positive additive Lévy field (following Chaumont-Marolleau), via an iterative gluing operation on compact marked metric spaces with vector-valued measures that extends techniques from Sénizergues. The result is positioned as a multitype generalization of Aldous-Duquesne-Le Gall continuum random trees.
Significance. If the convergence holds, the result supplies a scaling limit for a natural class of conditioned multitype branching trees and introduces a gluing construction that handles inter-type dependencies. This could serve as a technical tool for studying functionals of multitype Lévy trees. The manuscript credits the external constructions it relies upon and notes the increased complexity of the multitype dependency structure.
major comments (3)
- [Abstract, §1] Abstract and §1 (Introduction): the central invariance principle is stated only under 'sufficient mild conditions' on the offspring distributions, with no explicit list or formulation of those conditions (e.g., moment assumptions, irreducibility, or domain-of-attraction requirements). This is load-bearing for the claim that the chosen conditioning produces the glued multitype Lévy tree.
- [Gluing construction section] Section on the gluing construction (likely §4 or the extension of Sénizergues): the manuscript asserts that the iterative gluing operation on marked metric spaces with vector-valued measures preserves the metric properties needed for GHP convergence and tightness, but supplies no explicit verification or lemma showing that the multitype interactions do not violate the required completeness or separability after gluing.
- [Main theorem section] Main convergence theorem (presumably Theorem 3.1 or equivalent): the identification of the limit as the glued object determined by the additive Lévy field of Chaumont-Marolleau is asserted after extending Sénizergues' techniques, yet the argument provides no separate tightness criterion or identification step that isolates the effect of the maximal-subtree conditioning from the total-population conditioning used in the single-type case.
minor comments (2)
- Notation for the vector-valued measures and type-specific functionals is introduced without a consolidated table or list of symbols, making it difficult to track inter-type dependencies across sections.
- [Abstract] The abstract refers to 'an improvement of a result about the convergence... of compact marked metric spaces' but does not cite the precise prior theorem being improved.
Simulated Author's Rebuttal
We thank the referee for the careful reading and constructive comments on our manuscript. We address each major point below and indicate where revisions will be made to improve clarity and explicitness without altering the core results.
read point-by-point responses
-
Referee: [Abstract, §1] Abstract and §1 (Introduction): the central invariance principle is stated only under 'sufficient mild conditions' on the offspring distributions, with no explicit list or formulation of those conditions (e.g., moment assumptions, irreducibility, or domain-of-attraction requirements). This is load-bearing for the claim that the chosen conditioning produces the glued multitype Lévy tree.
Authors: The conditions are stated in detail in Section 2 (Assumptions 2.1--2.3), which include finite second-moment requirements on the offspring distributions, irreducibility and aperiodicity of the mean matrix, and the domain-of-attraction condition for convergence to a spectrally positive Lévy process. These are the 'sufficient mild conditions' referenced in the abstract and introduction. We agree that an explicit enumerated list at the beginning of the paper would improve readability and will add this in the revised version. revision: yes
-
Referee: [Gluing construction section] Section on the gluing construction (likely §4 or the extension of Sénizergues): the manuscript asserts that the iterative gluing operation on marked metric spaces with vector-valued measures preserves the metric properties needed for GHP convergence and tightness, but supplies no explicit verification or lemma showing that the multitype interactions do not violate the required completeness or separability after gluing.
Authors: The gluing construction in Section 4 extends the single-type arguments of Sénizergues by treating the vector-valued measures componentwise while preserving the metric structure via the iterative operation. Completeness and separability follow from the fact that each gluing step is a continuous operation on the space of compact marked metric spaces equipped with the GHP topology. We acknowledge that a dedicated lemma isolating these properties for the multitype case would strengthen the presentation and will insert such a lemma in the revision. revision: yes
-
Referee: [Main theorem section] Main convergence theorem (presumably Theorem 3.1 or equivalent): the identification of the limit as the glued object determined by the additive Lévy field of Chaumont-Marolleau is asserted after extending Sénizergues' techniques, yet the argument provides no separate tightness criterion or identification step that isolates the effect of the maximal-subtree conditioning from the total-population conditioning used in the single-type case.
Authors: Tightness is inherited from the single-type Lévy-tree convergences (via the marginal projections) together with the continuity of the gluing map with respect to the GHP topology; the maximal-subtree conditioning enters the limit through the law of the driving additive Lévy field, which encodes the type-specific sizes. The identification step therefore differs from the total-population conditioning in the single-type literature. We agree that separating the tightness argument from the identification argument more explicitly would clarify the role of the new conditioning and will restructure the proof of the main theorem accordingly. revision: partial
Circularity Check
No circularity: convergence theorem built from external constructions and new multitype extensions
full rationale
The paper proves an invariance principle for multitype Bienaymé-Galton-Watson trees converging to multitype Lévy trees via gluing of single-type Lévy trees, conditioned on maximal same-type subtree size. This rests on cited external results (Aldous, Duquesne-Le Gall, Chaumont-Marolleau, Sénizergues) for the single-type case and marked metric space gluing, with the authors extending Sénizergues' techniques to handle inter-type dependencies. No self-citations by these authors appear as load-bearing steps, no parameters are fitted inside the paper and renamed as predictions, and no definitions are circular. The central claim is a mathematical convergence statement in Gromov-Hausdorff-Prohorov topology whose proof chain is independent of the target result itself.
Axiom & Free-Parameter Ledger
axioms (2)
- standard math Single-type Lévy trees exist as scaling limits of Bienaymé-Galton-Watson trees under the cited conditions of Aldous, Duquesne and Le Gall.
- standard math The spectrally positive additive Lévy field constructed by Chaumont and Marolleau exists and determines the gluing map.
Reference graph
Works this paper leans on
-
[1]
[Abb17] Emmanuel Abbe, Community detection and stochastic block models: recent developments , J. Mach. Learn. Res. 18 (2017), Paper No. 177,
work page 2017
-
[2]
MR 3827065 [ABBG12] L. Addario-Berry, N. Broutin, and C. Goldschmidt, The continuum limit of critical random graphs , Probab. Theory Related Fields 152 (2012), no. 3-4, 367–406. MR 2892951 [ABBGM17] Louigi Addario-Berry, Nicolas Broutin, Christina Goldschmidt, and Gr´ egory Miermont, The scaling limit of the minimum spanning tree of the complete graph , A...
work page 2012
-
[3]
MR 373040 [Ang20] Osvaldo Angtuncio Hern´ andez, On Multitype Random Forests with a Given Degree Sequence, the Total Population of Branching Forests and Enumerations of Multitype Forests , arXiv e-prints (2020), arXiv:2003.03036. 56 Multitype L´ evy trees as scaling limits of MBGW trees [AP00] David Aldous and Jim Pitman, Inhomogeneous continuum random tr...
work page internal anchor Pith review Pith/arXiv arXiv 2020
-
[4]
MR 1406564 [Ber17] , Markovian growth-fragmentation processes , Bernoulli 23 (2017), no. 2, 1082–1101. MR 3606760 [BJR07] B´ ela Bollob´ as, Svante Janson, and Oliver Riordan, The phase transition in inhomogeneous random graphs, Random Structures Algorithms 31 (2007), no. 1, 3–122. MR 2337396 [BO18] Gabriel Hern´ an Berzunza Ojeda, On scaling limits of mu...
work page 2017
-
[5]
[CL16] Lo¨ ıc Chaumont and Rongli Liu, Coding multitype forests: application to the law of the total population of branching forests, Trans. Amer. Math. Soc. 368 (2016), no. 4, 2723–2747. MR 3449255 [Cla24a] David Clancy, Jr., Component sizes of rank-2 multiplicative random graphs , arXiv e-prints (2024), arXiv:2410.09224. [Cla24b] David Clancy, Jr., Near...
-
[6]
MR 4193902 [CM21] Lo¨ ıc Chaumont and Marine Marolleau, Extinction times of multitype, continuous-state branching pro- cesses, arXiv e-prints (2021), arXiv:2109.02912. [CPGUB13] M. Emilia Caballero, Jos´ e Luis P´ erez Garmendia, and Ger´ onimo Uribe Bravo,A Lamperti-type repre- sentation of continuous-state branching processes with immigration , Ann. Pro...
-
[7]
MR 3916322 [DLG02] Thomas Duquesne and Jean-Fran¸ cois Le Gall, Random trees, L´ evy processes and spatial branching processes, Ast´ erisque (2002), no. 281, vi+147. MR 1954248 [dR17] Lo¨ ıc de Raph´ elis, Scaling limit of multitype Galton-Watson trees with infinitely many types , Ann. Inst. Henri Poincar´ e Probab. Stat.53 (2017), no. 1, 200–225. MR 3606...
work page 2002
-
[8]
[HW25] Alastair Haig and Minmin Wang, Random bipartite graphs with i.i.d. weights and applications to inhomogeneous random intersection graphs, arXiv e-prints (2025), arXiv:2503.16688. [HS21] B´ en´ edicte Haas and Robin Stephenson, Scaling limits of multi-type Markov branching trees , Probab. Theory Related Fields 180 (2021), no. 3-4, 727–797. MR 4288331...
-
[9]
[Jag89] Peter Jagers, General branching processes as Markov fields, Stochastic Process. Appl. 32 (1989), no. 2, 183–212. MR 1014449 [KN11] Brian Karrer and Mark EJ Newman, Stochastic blockmodels and community structure in networks , Physical review E 83 (2011), no. 1, 016107. [Khe20] Ali Khezeli, Metrization of the Gromov-Hausdorff (-Prokhorov) topology f...
work page 1989
-
[10]
MR 4235481 [LG05] Jean-Fran¸ cois Le Gall, Random trees and applications, Probab. Surv. 2 (2005), 245–311. MR 2203728 [LGLJ98a] Jean-Fran¸ cois Le Gall and Yves Le Jan, Branching processes in L´ evy processes: Laplace functionals of snakes and superprocesses, The Annals of Probability 26 (1998), no. 4, 1407–1432. [LGLJ98b] Jean-Francois Le Gall and Yves L...
work page 2005
-
[11]
MR 1725357 [S´22] Delphin S´ enizergues, Growing random graphs with a preferential attachment structure , ALEA Lat. Am. J. Probab. Math. Stat. 19 (2022), no. 1, 259–309. MR 4359793 [Sat99] Ken-iti Sato, L´ evy processes and infinitely divisible distributions , Cambridge Studies in Advanced Mathematics, vol. 68, Cambridge University Press, Cambridge, 1999,...
work page 2022
-
[12]
[Whi80] Ward Whitt, Some useful functions for functional limit theorems , Mathematics of operations research 5 (1980), no. 1, 67–85. Osvaldo Angtuncio Hern´andez, Centro de Investigaci´on en Matem´aticas, Departamento de Prob- abilidad y Estad´ıstica, Guanajuato, M´exico Email address: osvaldo.angtuncio@cimat.mx David Clancy, Jr., Department of Mathematic...
work page 1980
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.