Tree-of-Stars in Graph Theory and Astrophysics

• Tree-of-Stars is a family of tree-based constructions featuring a central hub with pendant branches, linking graph theory with astrophysical and combinatorial applications.
• Spectral and algebraic analyses, including eigenvalue bounds and Salem number characterizations, provide precise structural insights for starlike trees.
• Applications range from MST-based mass segregation in star clusters to palindrome counting in edge-labeled trees and modeling explosive genealogies in probabilistic processes.

Searching arXiv for the cited works and related uses of “Tree-of-Stars.” In the cited arXiv literature, Tree-of-Stars is best understood as a family of related constructions rather than a single standardized term. The common motif is a hub-and-branches object represented directly as a tree, or a tree used to organize stars, star-like configurations, or star-shaped obstructions. In graph theory, the central object is the starlike tree, a tree with a unique high-degree vertex; in astrophysics, tree-based formalisms organize stellar positions, kinematics, or chemical abundances; in probability, an explosive genealogy may collapse onto a unique infinite-degree hub; and in infinite-graph theory and metric geometry, “stars” appear as separations or asymptotic incidence structures rather than celestial bodies (Shin, 2017, Torniamenti et al., 2021, Iyer et al., 2023, Bergen et al., 18 Mar 2026).

1. Core graph-theoretic object

A starlike tree is defined as a tree with exactly one vertex of degree greater than $2$. Equivalently, if one removes the central vertex $v_1$, the graph splits into disjoint paths, and the tree is written

$S(n_1,n_2,\dots,n_k),$

with

$$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$

where $P_n$ is a path on $n$ vertices. The integers $n_1,\dots,n_k$ are the branch lengths (Shin, 2017).

A particularly important family is

$S(n,k\cdot 1),$

which has one long arm of length $n$ and $k$ additional branches of length $v_1$0. In that case the tree has $v_1$1 arms in total and $v_1$2 vertices. This is the graph-theoretic setting in which the “Tree-of-Stars” label is used most literally: a central hub with one extended path and several pendant edges (Shin, 2017).

A second notation writes a starlike tree as

$v_1$3

where $v_1$4 is the number of vertices in the $v_1$5-th path attached to the center. If

$v_1$6

then $v_1$7 is a partition of $v_1$8, where $v_1$9 is the total number of vertices and

$S(n_1,n_2,\dots,n_k),$0

This identifies starlikes of fixed order with partitions, a viewpoint that is central to their spectral ordering (Oliveira et al., 2017).

2. Spectral and algebraic theory

For a starlike tree, the main spectral invariant is the largest eigenvalue $S(n_1,n_2,\dots,n_k),$1 of the adjacency matrix, i.e. the spectral radius. A general result cited from Lepović and Gutman states that for any starlike tree $S(n_1,n_2,\dots,n_k),$2,

$S(n_1,n_2,\dots,n_k),$3

For the special family $S(n_1,n_2,\dots,n_k),$4, the sharper estimate

$S(n_1,n_2,\dots,n_k),$5

holds for all $S(n_1,n_2,\dots,n_k),$6 and $S(n_1,n_2,\dots,n_k),$7. The lower bound comes from interlacing, since the graph contains a star on $S(n_1,n_2,\dots,n_k),$8 vertices as a subgraph, and the upper bound comes from the location of a root of an associated polynomial (Shin, 2017).

The distinctive feature of $S(n_1,n_2,\dots,n_k),$9 is a non-isomorphic spectral equivalence: $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$0 where $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$1 is the starlike tree with $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$2 equal branches, each of length $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$3 (equivalently described in the paper as having $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$4 branches of length $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$5 after a shift in convention). The proof rewrites the characteristic polynomial in terms of path polynomials $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$6, yielding

$$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$7

After substituting $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$8 and then $$S(n_1,n_2,\dots,n_k)-v_1=P_{n_1}\cup P_{n_2}\cup\cdots\cup P_{n_k},$$9, the spectral equation becomes

$P_n$0

This is exactly the polynomial previously obtained for the related balanced starlike tree (Shin, 2017).

The same paper derives algebraic consequences. For $P_n$1 and $P_n$2, $P_n$3 is hyperbolic, meaning it has exactly one eigenvalue greater than $P_n$4. If $P_n$5 is defined by

$P_n$6

then $P_n$7 is a Salem number. Writing

$P_n$8

its largest real root satisfies

$P_n$9

and the paper concludes that there is a sequence of Salem numbers converging to every integer greater than $n$0 (Shin, 2017).

A complementary spectral result shows that starlikes of fixed order are completely separated by their indices. For $n$1, any two non-isomorphic starlike trees on $n$2 vertices have different largest adjacency eigenvalues. More strongly, the starlike trees with $n$3 vertices can be totally ordered by index, and that order agrees with the lexicographic order of the corresponding partitions. The proof uses the Jacobs–Trevisan diagonalization algorithm together with the recurrence

$n$4

and the characterization

$n$5

when $n$6 (Oliveira et al., 2017).

The matrix-theoretic generalization replaces the adjacency viewpoint by a symmetric integral $n$7-matrix attached to a weighted star tree,

$n$8

whose arms are ordinary type-$n$9 chains and whose central diagonal entry is $n_1,\dots,n_k$0. Its determinant and inertia are controlled by the Schur complement scalar

$n_1,\dots,n_k$1

Accordingly, $n_1,\dots,n_k$2 is positive definite iff $n_1,\dots,n_k$3, positive semidefinite of corank one iff $n_1,\dots,n_k$4, and has exactly one negative eigenvalue iff $n_1,\dots,n_k$5. The affine condition

$n_1,\dots,n_k$6

reduces classification to an Egyptian-fraction problem. The classical affine diagrams $n_1,\dots,n_k$7 occur as small subfamilies, while higher-arm cases produce further positive-semidefinite weighted star matrices with explicit Coxeter labels (Torrente-Lujan, 21 May 2026).

3. Labeled starlikes and palindromic complexity

In combinatorics on words, the starlike tree becomes an edge-labeled object. A $n_1,\dots,n_k$8-starlike tree has $n_1,\dots,n_k$9 branches, each of length $S(n,k\cdot 1),$0, and each edge carries a letter from an alphabet $S(n,k\cdot 1),$1. Every simple path determines a word by reading edge labels along that path. A palindrome is a word $S(n,k\cdot 1),$2 such that

$S(n,k\cdot 1),$3

where $S(n,k\cdot 1),$4 is the reverse of $S(n,k\cdot 1),$5. The extremal quantity of interest is $S(n,k\cdot 1),$6, the maximum number of distinct non-empty palindromes that can appear in a $S(n,k\cdot 1),$7-starlike tree (Glen et al., 2018).

The main counting tool is the Droubay–Justin–Pirillo lemma: a word of length $S(n,k\cdot 1),$8 contains at most $S(n,k\cdot 1),$9 distinct non-empty palindromes. Using that lemma, the paper proves that if the branches $n$0 are ordered by

$n$1

then the tree contains at most

$n$2

distinct non-empty palindromes. For equal branch lengths this yields

$n$3

When $n$4, the bound becomes

$n$5

Over a binary alphabet, the bound sharpens to

$n$6

The argument labels the branches $n$7 and uses the fact that two of the three last letters must coincide. The paper further conjectures that for all $n$8,

$n$9

and gives explicit examples, including branch labels

$k$0

that attain $k$1. The noteworthy point is that a straightforward global upper bound is available, while the exact extremal value appears harder to determine (Glen et al., 2018).

4. Stellar-cluster and Galactic applications

In observational and computational astrophysics, the phrase “Tree-of-Stars” is applied to tree-based descriptions of spatial, kinematic, or chemical structure. One line of work uses the minimum spanning tree (MST) to detect and quantify mass segregation in star clusters. An MST is the shortest network connecting a set of points with no closed loops, and the method compares the MST length of the $k$2 most massive stars with the MST lengths of many random sets of the same size. The mass segregation ratio $k$3 is the ratio of the mean random MST length to the massive-star MST length, with uncertainty derived from the dispersion of the random samples. The method is explicitly motivated as model independent, does not require defining a cluster center, and can reveal multiple segregation levels as $k$4 varies. Applied to the Orion Nebula Cluster, it detects the Trapezium at $k$5 with

$k$6

finds broader segregation for stars above about $k$7 at roughly

$k$8

and finds no evidence for further mass segregation below $k$9 (0901.2047).

A second line of work turns a stellar cluster itself into a learned tree. A hierarchical generative model is built from sink particles in $v_1$00 SPH simulations of molecular clouds from Ballone et al. (2020), with cloud masses $v_1$01, initial temperature $v_1$02 K, initial density $v_1$03, a Burgers-spectrum turbulent field, and initial virial ratio

$v_1$04

Agglomerative clustering in phase space produces a dendrogram $v_1$05 whose internal nodes store a relative separation vector $v_1$06, a relative velocity vector $v_1$07, and a mass ratio

$v_1$08

Ward linkage performs best. New clusters are then generated recursively, with optional grafting of upper-level nodes from another tree. For the m1e4 cluster, $v_1$09 is identified as a good compromise: large-scale organization changes while small-scale structure remains broadly consistent. The generated systems reproduce the original pairwise-distance distribution $v_1$10, mass spectrum $v_1$11, velocity distribution $v_1$12, and multiscale fractal behavior, and direct $v_1$13-body integrations with NBODY6++GPU over $v_1$14 Myr give qualitatively similar evolution (Torniamenti et al., 2021).

A third astronomical use is Galactic phylogenetics, where elemental abundances play the role of inherited markers. Jofré et al. treat the interstellar medium as the shared ancestral environment and stellar abundances as a proxy for DNA, extending the logic of chemical tagging. Using solar twins from Nissen (2015) with accurate abundances for $v_1$15 elements, together with ages and kinematics, they build an evolutionary tree in chemical-abundance space. The resulting tree reveals three main branches, interpreted as the thick disk, thin disk, and an intermediate population, and branch lengths are used to estimate total chemical enrichment rates, with the thick disk branch showing a faster enrichment / star formation rate than the thin disk branch. The proceedings paper also stresses the method’s assumptions: chemical tagging must retain birth-environment information, the Galaxy may not be strictly hierarchical, and the sample was very small, so the result is presented as a proof of concept (Jofre et al., 2017).

5. Stars, separations, and asymptotic geometry

In infinite graph theory, a star attached to $v_1$16 is a subdivided infinite star whose leaves all lie in a specified vertex set $v_1$17, while a comb attached to $v_1$18 consists of a ray together with infinitely many pairwise disjoint finite teeth ending in $v_1$19. The classical star-comb lemma says that if $v_1$20 is infinite in a connected graph $v_1$21, then $v_1$22 contains either a comb attached to $v_1$23 or a star attached to $v_1$24. The duality theory developed from this starting point characterizes the absence of such objects. For combs, the complementary structures include a rayless normal tree containing $v_1$25, a rayless tree-decomposition into parts each containing at most finitely many vertices of $v_1$26, critical vertex sets, and a recursive $v_1$27-rank. For stars, the complementary structure is a locally finite normal tree containing $v_1$28 whose rays are undominated, or a locally finite tree-decomposition with finite and pairwise disjoint adhesion sets such that each part contains at most finitely many vertices of $v_1$29 (Bürger et al., 2020).

The third paper in the same series isolates the undominated comb. If $v_1$30 is normally spanned in $v_1$31, then the following are complementary: $v_1$32 contains an undominated comb attached to $v_1$33, or there exists a rayless tree $v_1$34 containing $v_1$35. In connected graphs, the same obstruction is equivalent to the existence of a star-decomposition with finite adhesion sets in which $v_1$36 lies in the central part and all undominated ends lie in leaf parts. A further corollary states that if a graph has a normal spanning tree, then it has a rayless spanning tree if and only if every ray of the graph is dominated (Bürger et al., 2020).

The vocabulary of stars of separations arises in tangle theory. There, a star is a set of non-degenerate oriented separations that point towards each other. An $v_1$37-tangle structure tree is a rooted separation tree whose leaves encode tangles or forbidden obstructions, and whose internal nodes certify that no $v_1$38-tangle extends through them. This object unifies a tree of tangles with a certificate of non-existence of tangles. When $v_1$39 consists of stars, as in classical tangle-tree duality, the paper shows how an irreducible $v_1$40-tree can be converted into a classical $v_1$41-tree over $v_1$42, recovering the usual decomposition-theoretic certificate (Bergen et al., 18 Mar 2026).

A different generalization appears in Karlsson’s stars at infinity. For a based metric space $v_1$43, a subset $v_1$44, and $v_1$45, the halfspace is

$v_1$46

and the star of a boundary point $v_1$47 is

$v_1$48

In hyperbolic spaces, stars are singletons; in CAT(0) spaces they are closed angular balls of radius $v_1$49; and in bounded convex domains with the Hilbert metric the incidence relation is symmetric. The Diestel–Leader graph $v_1$50 provides a counterexample to symmetry: the paper constructs horofunction boundary points $v_1$51 and $v_1$52 with

$v_1$53

This shows that the boundary relation “$v_1$54” need not be symmetric in general (Jones et al., 2020).

6. Explosive genealogies and probabilistic tree-of-stars

In the theory of explosive Crump–Mode–Jagers branching processes, the genealogical tree at explosion time,

$v_1$55

is the tree of all individuals born before the explosion time

$v_1$56

A star in this setting means a node of infinite degree, and an infinite path means a ray $v_1$57. Kőnig’s Lemma implies that any infinite tree contains either a node of infinite degree or an infinite path, and the paper shows that explosive genealogies satisfy a strong version of this dichotomy (Iyer et al., 2023).

Under Assumption $v_1$58, almost surely the infinite tree $v_1$59 contains a node of infinite degree. Under Assumption $v_1$60, the tree $v_1$61 contains an infinite path almost surely on survival. Under Assumption $v_1$62, almost surely on survival, $v_1$63 contains exactly one of the following: a node of infinite degree or an infinite path. The paper also proves a non-$v_1$64-or-$v_1$65 phenomenon: the event that there is a star can have strictly positive probability less than $v_1$66 (Iyer et al., 2023).

When a star exists, the local structure around it is not arbitrary. For a finite rooted tree $v_1$67, Theorem $v_1$68 identifies conditions under which $v_1$69 appears infinitely often, or only finitely often, as a rooted subtree of children of the infinite-degree node. This is the formal probabilistic “Tree-of-Stars” phenomenon: a unique infinite hub surrounded by infinitely many child-subtrees, with sharp recurrence criteria (Iyer et al., 2023).

The framework is applied to explosive recursive trees with fitness, where

$v_1$70

For super-linear preferential attachment with fitness, the paper derives phase transitions. In the additive-weight case with power-law weights, there is a unique star if

$v_1$71

and a unique infinite path if

$v_1$72

In the mixed-weight, super-linear case with power-law weights, the threshold becomes

$v_1$73

and in the mixed-weight, log-stretched setting it is

$v_1$74

These statements formalize a winner-takes-all versus eternal-lineage dichotomy (Iyer et al., 2023).

7. Inverted world tree and circumpolar symbolism

A much older and symbolic usage appears in the study of the Varvarinsky I stone slab from Rostov Oblast. The slab was found about $v_1$75 m southeast of kurgan $v_1$76, at about $v_1$77 N, $v_1$78 E, and weighs about $v_1$79 kg. The authors argue that the petroglyph “tree” on the slab is simultaneously an astronomical sign, an orientation mark, and an image of the inverted World Tree. On their interpretation, the slab models the Earth’s surface, the tree marks the direction to the North, and the trunk corresponds to the astronomical world axis (Vodolazhskaya et al., 2016).

The astronomical component is tied to the geometry of analemmatic sundials. Using

$v_1$80

and

$v_1$81

the paper treats the branches as reference points for semi-minor semiaxes. For $v_1$82 cm, the values

$v_1$83

correspond to

$v_1$84

with

$v_1$85

The lowest branch is associated with the North Pole case, where

$v_1$86

so the ellipse becomes a circle (Vodolazhskaya et al., 2016).

The interpretive claim is explicitly mythological. Viewed from the position of the gnomon or human observer, the tree is said to be inverted: roots toward the edge of the slab and crown toward the center. The trunk is identified with the axis mundi, and the branches symbolize the visible daily path of the Sun and perhaps the night paths of stars across the sky at different latitudes. Comparison with Srubna vessels leads the authors to suggest that the mythical World Tree was likely a tree of the pine family, and that the inverted tree signified the North and/or the north pole of the world, while accompanying six-pointed and four-pointed stars referred to the polar star and nearby circumpolar asterisms. In this setting, “Tree-of-Stars” is not a formal mathematical object but a symbolic coupling of world-axis imagery, northward orientation, and circumpolar stellar motion (Vodolazhskaya et al., 2016).