Translation-Invariant Splitting Gibbs Measures

• Translation-Invariant Splitting Gibbs Measures (TISGMs) are defined on Cayley trees as homogeneous Gibbs measures with constant boundary laws and fixed-point recursions.
• They reduce the complex infinite-volume DLR consistency to finite-dimensional fixed-point equations, as exemplified by the Blum–Kapel model.
• Their analysis via tree-indexed Markov chains reveals insights into phase transition, symmetry breaking, and the extremality conditions of Gibbs measures.

Searching arXiv for the primary paper and closely related TISGM literature on Cayley trees. Search query: (Xatamov et al., 2017) Translation-invariant Gibbs measures Blum-Kapel Cayley tree Translation-invariant splitting Gibbs measures (TISGMs) are Gibbs measures on trees that are simultaneously homogeneous and Markovian in the tree sense: they arise from constant boundary laws, equivalently from translation-invariant tree-indexed Markov chains. On Cayley trees, they serve as the canonical homogeneous phases, because the DLR consistency problem reduces to nonlinear recursion on descendants, and translation invariance turns this recursion into a finite-dimensional fixed-point problem. In the literature, TISGMs are therefore the natural interface between phase transition theory, boundary-law methods, and reconstruction on trees; the Blum–Kapel model on a Cayley tree is a particularly explicit instance of this framework (Xatamov et al., 2017).

1. Formal framework on Cayley trees

A Cayley tree of order $k$ is an infinite tree in which each vertex has degree $k+1$. After fixing a root $x^0$, one uses the spheres $W_n=\{x:d(x,x^0)=n\}$, balls $V_n=\{x:d(x,x^0)\le n\}$, and successor sets $S(x)$ to express Gibbs specifications recursively along generations. This rooted description is not merely notational: it is the mechanism through which tree models admit boundary-law equations and splitting constructions.

For nearest-neighbor models on a Cayley tree, finite-volume Gibbs measures are defined on $V_n$ with boundary fields placed on the outer layer $W_n$. Consistency of these finite-volume measures under marginalization is equivalent to the Dobrushin–Lanford–Ruelle condition and yields an infinite-volume Gibbs measure by Kolmogorov’s extension theorem. In the tree literature, a splitting Gibbs measure is exactly a Gibbs measure generated by such a compatible boundary-law recursion; equivalently, it is a tree-indexed Markov chain whose subtrees are conditionally independent given the parent spin (Külske et al., 2013).

A TISGM is obtained when the boundary law is constant across the tree. In practice, one starts with a site-dependent recursion $h_x=\sum_{y\in S(x)}F(h_y)$, or an equivalent recursion for positive ratios $z_{i,x}$, and imposes $k+1$0 or $k+1$1. The infinite-dimensional consistency problem then collapses to a fixed-point equation in finitely many variables. This reduction is the central structural feature of TISGM theory.

2. Boundary laws, fixed points, and the Markov-chain viewpoint

The modern tree formulation of TISGMs is boundary-law based. For the ferromagnetic $k+1$2-state Potts model, compatibility of the finite-volume Gibbs distributions is equivalent to a recursion of the form

$k+1$3

where $k+1$4, $k+1$5, and $k+1$6 is an explicit nonlinear map. Under translation invariance this becomes

$k+1$7

or, after exponentiating coordinates,

$k+1$8

A central structural theorem states that every translation-invariant solution has a block form in which some subset of coordinates is equal to a common $k+1$9 and the rest are equal to $x^0$0; this reduces the classification to scalar equations $x^0$1, indexed by subset size $x^0$2 (Külske et al., 2013).

That reduction has several consequences. First, the number of TISGMs can be counted exactly in many Potts regimes; in particular, at sufficiently low temperatures the number is $x^0$3, and there are $x^0$4 critical temperatures at which the count changes (Külske et al., 2013). Second, every TISGM comes with an explicit transition matrix

$x^0$5

so extremality becomes a question about reconstruction for a concrete tree-indexed Markov chain. Third, boundary conditions can be tied directly to TISGMs: for the Potts model on the binary tree, explicit classes of boundary configurations were constructed that converge to each prescribed TISGM (Gandolfo et al., 2015).

This boundary-law interpretation also clarifies terminology. Translation invariance concerns homogeneity of the boundary law or transition kernel; splitting refers to the Markov decomposition along forward subtrees; extremality concerns whether the resulting Gibbs measure is a pure state. These notions are related but not identical.

3. The Blum–Kapel model as a worked example

In the Blum–Kapel model studied on the Cayley tree, the spin space is

$x^0$6

and the Hamiltonian is

$x^0$7

Finite-volume Gibbs measures on $x^0$8 are defined with boundary fields $x^0$9 on $W_n=\{x:d(x,x^0)=n\}$0, and one introduces

$W_n=\{x:d(x,x^0)=n\}$1

The consistency theorem gives the boundary-law recursion

$W_n=\{x:d(x,x^0)=n\}$2

$W_n=\{x:d(x,x^0)=n\}$3

A TISGM is therefore a positive constant solution $W_n=\{x:d(x,x^0)=n\}$4 of

$W_n=\{x:d(x,x^0)=n\}$5

Each such solution defines a translation-invariant splitting Gibbs measure, and the measure is represented explicitly by a homogeneous transition matrix of a tree-indexed Markov chain (Xatamov et al., 2017).

The symmetric branch $W_n=\{x:d(x,x^0)=n\}$6 satisfies

$W_n=\{x:d(x,x^0)=n\}$7

For every $W_n=\{x:d(x,x^0)=n\}$8 and every $W_n=\{x:d(x,x^0)=n\}$9, this scalar equation has a unique positive solution. Hence there is always at least one symmetric TISGM. For $V_n=\{x:d(x,x^0)\le n\}$0, the symmetric equation becomes a cubic and admits an explicit Cardano-type solution $V_n=\{x:d(x,x^0)\le n\}$1 (Xatamov et al., 2017).

4. Multiplicity, phase transition, and symmetry breaking

The full classification in the Blum–Kapel paper is carried out explicitly for the binary tree $V_n=\{x:d(x,x^0)\le n\}$2. In that case there exists

$V_n=\{x:d(x,x^0)\le n\}$3

such that the translation-invariant phase diagram is exactly: $V_n=\{x:d(x,x^0)\le n\}$4

$V_n=\{x:d(x,x^0)\le n\}$5

Here $V_n=\{x:d(x,x^0)\le n\}$6 is the symmetric TISGM associated with $V_n=\{x:d(x,x^0)\le n\}$7, while $V_n=\{x:d(x,x^0)\le n\}$8 and $V_n=\{x:d(x,x^0)\le n\}$9 correspond to the asymmetric solutions $S(x)$0 and $S(x)$1 (Xatamov et al., 2017).

Using $S(x)$2, the corresponding critical temperature is

$S(x)$3

Thus the model has a unique TISGM for $S(x)$4 and exactly three TISGMs for $S(x)$5. The fixed-point picture is the standard symmetry-breaking scenario: a symmetric branch persists for all temperatures, and below $S(x)$6 two asymmetric branches bifurcate from it (Xatamov et al., 2017).

This is the archetypal role of TISGMs in tree models. The multiplicity of positive fixed points of the translation-invariant recursion is the multiplicity of homogeneous phases. In this sense, TISGMs convert phase transition questions into nonlinear fixed-point analysis.

5. Extremality, reconstruction, and what TISGMs do not imply

For the Blum–Kapel model, extremality was analyzed for the symmetric TISGM $S(x)$7 on the binary tree. The associated transition matrix has eigenvalues

$S(x)$8

with $S(x)$9, and the relevant Kesten–Stigum eigenvalue is $V_n$0. The sufficient condition for non-extremality is

$V_n$1

Numerically, this holds for

$V_n$2

Hence $V_n$3 is non-extreme in those regions (Xatamov et al., 2017).

A sufficient condition for extremality comes from the Martinelli–Sinclair–Weitz criterion $V_n$4. In this model it reduces to the same threshold interval, yielding

$V_n$5

For the physically relevant ferromagnetic parametrization $V_n$6, this means that the symmetric TISGM is extreme at intermediate and high temperatures but becomes non-extreme at sufficiently low temperature, namely for $V_n$7 (Xatamov et al., 2017).

Two points follow. First, multiplicity of TISGMs and extremality are distinct issues: counting translation-invariant fixed points does not by itself identify pure phases. Second, even when a TISGM is uniquely specified inside the translation-invariant class, that does not automatically settle the structure of the full Gibbs simplex. In the Blum–Kapel paper, it is further shown that for $V_n$8 there are at least two extreme Gibbs measures, while the individual extremality of the asymmetric TISGMs $V_n$9 remains open (Xatamov et al., 2017).

6. Position within the broader TISGM literature

The Potts model is the benchmark case. For the ferromagnetic Potts model on a Cayley tree, all TISGMs are obtained from the scalar equations $W_n$0, and at sufficiently low temperature their number is $W_n$1 (Külske et al., 2013). Their extremality can be studied by coarse-graining: fuzzy transformations map Potts TISGMs to effective two-state models, and in a temperature interval there are at least $W_n$2 extremal TISGMs on the binary tree (Kuelske et al., 2014). The problem of which boundary conditions select which TISGM was solved explicitly for the order-two tree, again in terms of the tree recursion (Gandolfo et al., 2015).

The relation between translation invariance and periodicity is model dependent. In the zero-field Potts model, some periodic Gibbs measures collapse to translation-invariant ones: for the $W_n$3-state antiferromagnetic Potts model on the Cayley tree of order two and for the $W_n$4-state ferromagnetic Potts model on the Cayley tree of order $W_n$5, all periodic Gibbs measures are translation-invariant (Khakimov et al., 2018). This shows that TISGMs may exhaust the periodic sector in certain regimes, but not universally.

Mixed-spin and coupled models broaden the TISGM landscape substantially. The mixed spin-$W_n$6 Ising model on the second-order Cayley tree has three TISGMs in both the ferromagnetic and anti-ferromagnetic regimes, and the disordered TISGM has an exact extremality characterization via tree-indexed Markov chains (Akin et al., 2022). A periodic triple mixed-spin Ising model with spins $W_n$7 reduces to a scalar fixed-point equation and has at least three TISGMs when a derivative instability criterion is positive; on the binary tree, the plus and minus limits are themselves TISGMs (Mukhamedov et al., 12 Feb 2026). Coupled Ising–Potts models can be much richer still: one paper proves at least three TISGMs in general and at least eight for the $W_n$8-model on the binary tree (Rahmatullaev et al., 3 Nov 2025), while another finds at least $W_n$9 TISGMs at sufficiently low temperature and an exact count of $h_x=\sum_{y\in S(x)}F(h_y)$0 TISGMs for $h_x=\sum_{y\in S(x)}F(h_y)$1, $h_x=\sum_{y\in S(x)}F(h_y)$2 (Haydarov et al., 17 Feb 2025). In the HC–Blume–Capel model with wand graph, there is again a sharp one-versus-three TISGM dichotomy, together with a complete extremality analysis for one symmetric measure (Khatamov et al., 30 Mar 2026).

Taken together, these results establish the encyclopedic significance of TISGMs on trees. They are the principal homogeneous Gibbs phases, they are classified by boundary-law fixed points, they encode phase transitions through multiplicity changes, and their extremality is governed by reconstruction criteria for the associated tree-indexed Markov chains. The Blum–Kapel model provides a particularly transparent example, but the same boundary-law architecture persists across Potts, mixed-spin, hard-core, and coupled-spin systems (Xatamov et al., 2017).