Multipermutation Solutions in Yang–Baxter Theory

• Multipermutation solutions are set‑theoretic solutions of the Yang–Baxter equation that reduce to a one-point solution via iterated retractions.
• They bridge retraction theory with brace and skew‑brace formulations, revealing connections with group orderability, nilpotency, and decomposability.
• The theory extends to non‑involutive and degenerate cases, offering explicit classifications in low-level and square-free regimes through combinatorial and algebraic methods.

Multipermutation solutions are set-theoretic solutions of the Yang–Baxter equation whose iterated retractions eventually collapse to the one-point solution. In the classical setting, one starts from a non-degenerate involutive solution $(X,r)$ and measures its complexity by the least $m$ such that $\operatorname{Ret}^m(X,r)$ has cardinality $1$; this integer is the multipermutation level. The notion has since been extended to non-involutive and, more recently, degenerate settings, and it now serves as a central organizing principle linking retraction theory, braces and skew braces, orderability and nilpotency of associated groups, decomposability phenomena, and explicit classification results in low level and square-free regimes (Cedo et al., 2012, Jedlicka et al., 2024, Jedlička et al., 25 Aug 2025).

1. Retraction, irretractability, and multipermutation level

For a set-theoretic solution written in the classical form

$r(x,y)=(\sigma_x(y),\tau_y(x)),$

the retraction is obtained by quotienting by equality of actions. In the finite non-degenerate involutive setting, the standard relation is

$x\sim y \iff \sigma_x=\sigma_y,$

while for general non-degenerate solutions $(X,o,T)$, with

$r(x,y)=(o_x(y),\,T_y(x)),$

the retraction relation is

$x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$

The $m$-fold retraction is defined recursively, and $m$0 is a multipermutation solution of level $m$1 if $m$2 is minimal with $m$3 consisting of one element (Bachiller et al., 2017, Lebed et al., 2017, Jedlicka et al., 2024).

This definition isolates the class of retractable solutions: those that become trivial after finitely many successive collapses of indistinguishable action profiles. The opposite extreme is irretractability, where the first retraction does not reduce the solution at all. In the involutive literature, irretractable examples play a structural role because they delimit the scope of the multipermutation theory and provide counterexamples to earlier conjectural expectations (Bachiller et al., 2015).

A refined invariant has also been introduced. For a multipermutation solution $m$4, one defines

$m$5

as the smallest $m$6 such that $m$7 is a trivial solution, possibly with more than one point. This invariant satisfies

$m$8

and for indecomposable solutions of finite multipermutation level one has

$m$9

This variation is specifically designed to align the solution-theoretic and brace-theoretic filtrations exactly (Castelli et al., 2023).

For degenerate solutions, the classical relation “same action” need not be a congruence, so the naive retraction may fail to exist. The corrected definition uses the largest congruence contained in the same-action relation, and the retract is defined as the quotient by that largest congruence. This extends the notion of multipermutation level beyond the non-degenerate setting and places it in a universal-algebraic framework (Jedlička et al., 25 Aug 2025).

2. Brace and skew-brace formulations

The modern theory of multipermutation solutions is inseparable from brace theory. A left brace $\operatorname{Ret}^m(X,r)$0 carries an abelian additive group and a multiplicative group, with the map

$\operatorname{Ret}^m(X,r)$1

encoding the interaction between the two operations. For a left brace, the socle is

$\operatorname{Ret}^m(X,r)$2

equivalently the set of elements satisfying $\operatorname{Ret}^m(X,r)$3 for all $\operatorname{Ret}^m(X,r)$4. For the solution associated to a left brace, retraction is identified with quotienting by the socle: $\operatorname{Ret}^m(X,r)$5 This makes iterative retraction a socle-filtration process on the brace side (Cedo et al., 2012).

The skew-brace viewpoint sharpens this correspondence. If $\operatorname{Ret}^m(X,r)$6 is the permutation skew brace attached to a solution, then its socle series

$\operatorname{Ret}^m(X,r)$7

satisfies

$\operatorname{Ret}^m(X,r)$8

The exact identity

$\operatorname{Ret}^m(X,r)$9

shows that the modified level $1$0 is precisely the brace multipermutation level of the permutation skew brace (Castelli et al., 2023).

Nilpotency conditions on braces provide another characterization. For skew braces of nilpotent type, the paper on annihilator nilpotency shows that

$1$1

Since multipermutation solutions are characterized by right nilpotency of the associated structure or permutation brace in nilpotent type, annihilator nilpotency becomes an equivalent brace-theoretic proxy in that setting (Jespers et al., 2022).

The algebra attached to retraction also admits an explicit ideal-theoretic filtration. For a symmetric group $1$2, the derived chain of ideals

$1$3

satisfies

$1$4

Thus the chain records the retraction process step by step, and finite multipermutation level is equivalent to stabilization of this chain at the whole group (Gateva-Ivanova, 2015).

3. Classical characterizations via structure groups and orderability

For finite non-degenerate involutive solutions, multipermutation behavior is detected exactly by the structure group. The fundamental theorem states that for such a solution $1$5, the following are equivalent:

1. $1$6 is a multipermutation solution;
2. $1$7 is left orderable;
3. $1$8 is poly-$1$9 (Bachiller et al., 2017).

This equivalence transforms a retraction-theoretic property into a purely group-theoretic one. In the same involutive setting, the structure-group paper strengthens this picture by proving that multipermutation involutive solutions are the only involutive solutions with diffuse structure group, and that biorderability is much more restrictive: $r(x,y)=(\sigma_x(y),\tau_y(x)),$0 Thus left orderability captures the full multipermutation class, whereas biorderability captures only the trivial involutive case (Lebed et al., 2017).

Abelianness conditions on Yang–Baxter groups also force retractability. If a finite non-degenerate involutive solution has abelian associated involutive Yang–Baxter group, then it is a multipermutation solution. On the brace side, this is explained by the fact that an abelian multiplicative group forces the brace to be two-sided, and finite non-trivial two-sided braces have nonzero socle (Cedo et al., 2012).

Arithmetic restrictions on associated groups yield further criteria. If the permutation group $r(x,y)=(\sigma_x(y),\tau_y(x)),$1 of a finite non-degenerate involutive solution has cube-free cardinality, then the solution is a multipermutation solution. The proof proceeds through the brace structure on $r(x,y)=(\sigma_x(y),\tau_y(x)),$2, using annihilation relations between additive Sylow subgroups to force a nonzero socle and hence finite multipermutation level (Smoktunowicz, 2015).

These characterizations explain why multipermutation solutions are often regarded as the tractable part of the set-theoretic Yang–Baxter landscape: they are precisely the solutions whose associated groups admit strong decompositional or order-theoretic structure.

4. Non-involutive, distributive, and degenerate extensions

The non-involutive theory preserves the same retraction philosophy but requires additional structure. For a non-degenerate solution $r(x,y)=(\sigma_x(y),\tau_y(x)),$3, the diagonal maps

$r(x,y)=(\sigma_x(y),\tau_y(x)),$4

are mutually constrained: any non-degenerate solution is bijective, and the mappings $r(x,y)=(\sigma_x(y),\tau_y(x)),$5 and $r(x,y)=(\sigma_x(y),\tau_y(x)),$6 commute. These diagonal operators become part of the structural toolkit for analyzing retraction and decomposability (Jedlicka et al., 2024).

A major advance is the equational characterization of multipermutation solutions in the non-involutive setting. For a non-degenerate solution, Theorem 4.4 gives equivalence between “multipermutation level at most $r(x,y)=(\sigma_x(y),\tau_y(x)),$7” and explicit identities involving the iterated expressions $r(x,y)=(\sigma_x(y),\tau_y(x)),$8 and $r(x,y)=(\sigma_x(y),\tau_y(x)),$9. This leads to the notion of a $x\sim y \iff \sigma_x=\sigma_y,$0-permutational solution, and the paper states that a non-degenerate solution is of multipermutation level $x\sim y \iff \sigma_x=\sigma_y,$1 if and only if it is $x\sim y \iff \sigma_x=\sigma_y,$2-permutational and not $x\sim y \iff \sigma_x=\sigma_y,$3-permutational. It also shows that each $x\sim y \iff \sigma_x=\sigma_y,$4-permutational solution is $x\sim y \iff \sigma_x=\sigma_y,$5-reductive (Jedlicka et al., 2024).

The same paper extends classical decomposability results of Rump and Gateva-Ivanova. In particular, non-degenerate multipermutation square-free solutions are decomposable, and this holds without finiteness assumptions on the cardinality. This is the non-involutive analogue of the classical square-free involutive decomposability theorem (Jedlicka et al., 2024).

For distributive biracks, the theory becomes especially rigid. If $x\sim y \iff \sigma_x=\sigma_y,$6 is a distributive birack and $x\sim y \iff \sigma_x=\sigma_y,$7, then the following are equivalent:

1. $x\sim y \iff \sigma_x=\sigma_y,$8;
2. $x\sim y \iff \sigma_x=\sigma_y,$9 is $(X,o,T)$0-reductive;
3. $(X,o,T)$1 is $(X,o,T)$2-permutational;
4. $(X,o,T)$3 is nilpotent of class at most $(X,o,T)$4. This gives a non-involutive class in which multipermutation level is exactly equivalent to nilpotency of the multiplication group (Jedlička et al., 2019).

The degenerate case required a redefinition of retraction. There, the largest congruence below the same-action relation replaces the classical quotient, and the paper proves that the resulting notion of “multipermutation level at most $(X,o,T)$5” is again equivalent to a $(X,o,T)$6-permutational identity, now written in terms of the iterated expressions $(X,o,T)$7. This shows that the retraction hierarchy survives outside the non-degenerate framework, although in a more elaborate form (Jedlička et al., 25 Aug 2025).

5. Level $(X,o,T)$8: mediality, isotopy, and complete classifications

Multipermutation level $(X,o,T)$9 occupies a special position because it admits complete or nearly complete structural descriptions. In the involutive setting, the principal theorem is that a solution is of multipermutation level $r(x,y)=(o_x(y),\,T_y(x)),$0 if and only if it is medial. Within that class, distributive, $r(x,y)=(o_x(y),\,T_y(x)),$1-reductive, lri, and “first retraction trivial” are equivalent conditions (Jedlička et al., 2019).

Distributive level-$r(x,y)=(o_x(y),\,T_y(x)),$2 solutions are explicitly modeled by abelian groups and constants. If $r(x,y)=(o_x(y),\,T_y(x)),$3 is a disjoint union of abelian groups $r(x,y)=(o_x(y),\,T_y(x)),$4, then every distributive solution has the form

$r(x,y)=(o_x(y),\,T_y(x)),$5

for $r(x,y)=(o_x(y),\,T_y(x)),$6, $r(x,y)=(o_x(y),\,T_y(x)),$7, with each $r(x,y)=(o_x(y),\,T_y(x)),$8 generated by the constants in the $r(x,y)=(o_x(y),\,T_y(x)),$9-th column. Non-distributive level-$x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$0 solutions arise by isotopy from distributive ones (Jedlička et al., 2019).

For indecomposable involutive solutions of level $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$1, there is a complete characterization by quotients of a universal family $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$2. Every such solution is a homomorphic image of a previously constructed family, the congruences are described explicitly by triples $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$3, and the automorphism group of any indecomposable multipermutation-level-$x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$4 solution is regular (Jedlička et al., 2022).

The non-involutive classification follows the same isotopic philosophy. Every $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$5-permutational solution is uniquely determined, up to isomorphism, by a unique square-free $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$6-reductive solution and a unique pair of commuting automorphisms whose isotope recovers the original solution. This produces an explicit algorithm for all finite solutions of multipermutation level $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$7, and the classification was carried out up to size $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$8 (Hora et al., 2024).

A further refinement treats the indecomposable non-degenerate, not necessarily involutive, case. Every such solution of multipermutation level $x\sim y \iff o_x=o_y \text{ and } T_x=T_y.$9 is a homomorphic image of a universal solution, and there is also a concrete family on

$m$0

with $m$1 abelian, whose quotients are classified by a triple consisting of a subgroup $m$2, a subgroup $m$3, and a homomorphism

$m$4

Here again the automorphism group is regular (Jedlička et al., 25 Aug 2025).

6. Square-free structure, constructions, and limitations

Square-free solutions form one of the most intensively studied subclasses. A decisive theorem states that if $m$5 is an indecomposable involutive non-degenerate set-theoretic solution with

$m$6

where $m$7 are pairwise distinct primes, then $m$8 is a multipermutation solution. More sharply, if it is already multipermutation, then its level is at most $m$9. The same work proves that there is no simple solution of non-prime square-free cardinality, and that the bound is sharp by constructing indecomposable examples of cardinality $m$00 with multipermutation level exactly $m$01 (Cedó et al., 2022).

The proof is brace-theoretic. The permutation group $m$02 has only the primes $m$03 in its order, its Sylow $m$04-subgroups are elementary abelian, and the associated left brace admits an ordered product decomposition by additive Sylow subgroups that sit successively in the socle filtration. This rigid internal structure explains why square-free indecomposable solutions in this arithmetic regime are forced into the multipermutation class (Cedó et al., 2022).

Explicit constructions show that arbitrarily large level is possible even under strong constraints. For each positive integer $m$05, there exists a finite square-free multipermutation solution of multipermutation level $m$06 whose associated involutive Yang–Baxter group is an elementary abelian $m$07-group (Cedo et al., 2012). A different combinatorial construction via cycle matrices produces multipermutation solutions of every level $m$08 and gives an alternate proof that the class of permutation groups of solutions contains all finite abelian groups (Kanrar et al., 2023).

Recent work relates multipermutation behavior to the diagonal permutation. For finite involutive solutions whose diagonal is an $m$09-cycle, odd prime-power size forces finite multipermutation level. In the $m$10-power case, the outcome is dichotomic: either the solution has finite multipermutation level, or some iterated retraction is the unique irretractable solution of size $m$11, denoted $m$12 (Dietzel, 19 Apr 2025).

The boundaries of the theory are equally instructive. There exists a family of non-degenerate involutive solutions that are strong twisted unions of orbit subsolutions, each of multipermutation level at most $m$13, yet a large subfamily is square-free and irretractable. These examples show that assembling low-level multipermutation pieces does not force the global solution to be multipermutation, and in the finite case their structure groups are not poly-$m$14 (Bachiller et al., 2015).

Taken together, these results show that multipermutation solutions are neither accidental nor universal. They form a robust, highly structured class, detectable through retraction, braces, nilpotency, and orderability, sharply classifiable in several important regimes, yet separated from irretractable behavior by explicit and sometimes subtle counterexamples.