Minion Homomorphisms in Algebra and CSP

• Minion homomorphisms are arity-preserving maps that respect minor operations and height-1 identities, forming the backbone of algebraic CSP classifications.
• They provide a unifying framework for CSP/PCSP complexity analysis by enabling reductions, explicit classification via cores, and the establishment of structured homomorphism posets.
• Multisorted generalizations of minion homomorphisms connect abstract category theory with concrete polymorphism theory, offering insights into algorithmic tractability and core identification.

Searching arXiv for papers on minion homomorphisms and closely related classifications. Minion homomorphisms are arity-preserving maps between minions that commute with all minor operations, and therefore preserve precisely the height-1 identities encoded by permutation, identification, and introduction of dummy variables. In the function setting, if $f:A^n\to B$ and $\alpha:[n]\to[m]$, the minor $f^{(\alpha)}$ is given by

$$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$

and a minion homomorphism $\xi$ satisfies

$\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$

This notion appears under several names, including minor-preserving maps and height-1 clone homomorphisms, and serves as the central morphism in the algebraic theory of CSPs and PCSPs, in multisorted minion theory, and in recent classification programs for clones up to minor-equivalence (Barto et al., 2024).

1. Definition and basic formalism

A function minion on $(A,B)$ is a set of operations from $A$ to $B$ closed under taking minors. In the multisorted setting, for a $k$-sorted domain $\alpha:[n]\to[m]$0 and codomain $\alpha:[n]\to[m]$1, an $\alpha:[n]\to[m]$2-ary operation is a tuple $\alpha:[n]\to[m]$3 with $\alpha:[n]\to[m]$4, and minors are defined componentwise. An abstract minion $\alpha:[n]\to[m]$5 consists of nonempty sets $\alpha:[n]\to[m]$6 for each arity $\alpha:[n]\to[m]$7, together with maps $\alpha:[n]\to[m]$8 for every $\alpha:[n]\to[m]$9, satisfying identity and functoriality axioms. Categorically, this is exactly a finitary functor from the category of nonempty finite sets and functions to sets (Barto et al., 2024).

A minion homomorphism $f^{(\alpha)}$0 is a family of maps $f^{(\alpha)}$1 such that for all $f^{(\alpha)}$2 and all $f^{(\alpha)}$3,

$f^{(\alpha)}$4

Equivalently, $f^{(\alpha)}$5 preserves arities and all minor operations. In the multisorted formulation this becomes the declaration-preserving condition $f^{(\alpha)}$6, where declarations encode input sorts and output sort (Lehtonen et al., 2020).

The terminology “height-1 clone homomorphism” reflects the fact that the condition is exactly preservation of identities in which each side contains one function symbol and no nesting. Thus minion homomorphisms are strictly weaker than ordinary clone homomorphisms, which must preserve all equational identities, including nested term identities. This makes minion homomorphisms the appropriate morphisms when composition is not the invariant of interest, or when one works with minions rather than clones (Barto et al., 2024).

A standard preorder is induced by existence of minion homomorphisms: one writes $f^{(\alpha)}$7 if there exists a minion homomorphism $f^{(\alpha)}$8, and $f^{(\alpha)}$9 if both $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$0 and $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$1. The quotient by $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$2 yields the minion homomorphism poset. Two distinguished extremal objects recur in the literature: the largest minion, up to isomorphism, is the clone $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$3 of all operations on a singleton set, while the smallest is the projections-only clone $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$4 (Barto et al., 2024).

2. Height-1 identities and minor conditions

Minion homomorphisms can be characterized entirely in terms of minor conditions. A minor identity has the form

$$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$5

where only variables are substituted into operation symbols and no composition is allowed. A minor condition is a finite set of such identities. A key fact used repeatedly in recent work is that there is a minion homomorphism $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$6 if and only if every minor condition satisfied by $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$7 is also satisfied by $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$8 (Fioravanti et al., 9 Aug 2025).

This interpretation organizes the relation between specific algebraic conditions and incoming or outgoing minion homomorphisms. For example, the clone $$f^{(\alpha)}(a_1,\dots,a_m)=f(a_{\alpha(1)},\dots,a_{\alpha(n)}),$$9 of all idempotent operations on a two-element set is characterized by a basis consisting of a quasi Maltsev operation together with fully symmetric operations of all arities: if a clone over a finite domain has a quasi Maltsev operation and fully symmetric operations of all arities, then it satisfies all minor conditions satisfied by $\xi$0, equivalently there is a minion homomorphism

$\xi$1

This replaces an earlier basis involving generalized minority operations of all odd arities and totally symmetric operations of all arities (Meyer et al., 2024).

The same paper isolates the role of failure of minor conditions in the pp-constructability poset. Proper lower covers of the class represented by $\xi$2 must fail either the quasi Maltsev condition or some fully symmetric condition $\xi$3; this dichotomy drives the classification of those lower covers into the transitive tournament $\xi$4 and a family indexed by finite simple groups (Meyer et al., 2024). This suggests a general pattern: minion homomorphism order is often controlled by a small set of height-1 obstructions.

In classification problems over fixed finite domains, separating non-equivalent classes also proceeds by exhibiting explicit minor conditions. In the classification of Mal’cev clones on a three-element set up to minion homomorphisms, distinctions between classes are witnessed by existence or non-existence of quasi majority, quasi minority, cyclic operations, or explicit height-1 conditions such as $\xi$5 and $\xi$6 (Fioravanti et al., 9 Aug 2025).

3. Multisorted generalizations and abstract–concrete viewpoints

The multisorted theory makes minor-preserving structure explicit at the level of sorts. For an $\xi$7-sorted set $\xi$8, an $\xi$9-sorted operation is a map $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$0 for a word $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$1, and minors are formed only along sort-compatible maps $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$2. A multisorted minion is any set of such operations closed under minors, and a multisorted minion homomorphism preserves both declarations and minors (Lehtonen et al., 2020).

This framework is especially important because many natural constructions in CSP and PCSP theory are inherently multisorted. The multisorted analogue of the Wonderland theorem states that for minions $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$3 and $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$4,

$\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$5

Here $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$6, $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$7, and $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$8 denote closure under Extensions, Reflections, and direct Powers. The surjective version replaces $\xi^{(m)}(f^{(\alpha)})=(\xi^{(n)}(f))^{(\alpha)}.$9 by $(A,B)$0 (Lehtonen et al., 2020).

The abstract viewpoint identifies minions with functors $(A,B)$1, where $(A,B)$2 is the category of non-zero finite ordinals and all functions, and minion homomorphisms become natural transformations. In this language the minor action $(A,B)$3 is simply functoriality with respect to $(A,B)$4, and the equation

$(A,B)$5

is the naturality condition for a transformation $(A,B)$6 (Juhrich, 17 Mar 2025).

The abstract–concrete connection has two notable consequences. First, every finitely generated minion is finitely representable, hence isomorphic to a polymorphism minion $(A,B)$7 for finite relational structures $(A,B)$8. Second, the homomorphism orders induced by minion homomorphisms admit strong categorical structure: products and coproducts induce meets and joins, and several of the resulting hom-orders are bounded distributive lattices and in fact bi-Heyting algebras (Juhrich, 17 Mar 2025). This places minion homomorphisms simultaneously in universal algebra, category theory, and computational complexity.

4. pp-constructability, reductions, and complexity

For finite relational structures, minion homomorphisms are the algebraic counterpart of primitive positive constructability. In one standard formulation, for finite structures $(A,B)$9,

$A$0

Equivalent reformulations state that every minor condition satisfied by $A$1 is also satisfied by $A$2 (Meyer et al., 2024).

The same correspondence appears in the PCSP setting. If there is a minion homomorphism

$A$3

then there is a log-space reduction

$A$4

Thus the homomorphism preorder on polymorphism minions organizes reductions in the opposite direction on promise problems. A related hardness criterion states that if $A$5 admits a minion homomorphism to a minion of bounded essential arity, then $A$6 is NP-hard (Juhrich, 17 Mar 2025).

Algorithmic hierarchies for PCSPs can also be phrased using minion homomorphisms. In the tensor-based framework of minion tests, a minion test $A$7 solves $A$8 if and only if

$A$9

This applies to minions capturing arc consistency, BLP, AIP, SDP, and the combined BLP+AIP relaxation. In particular, the paper introduces a new minion $B$0 characterizing standard SDP and proves that SDP solves $B$1 iff

$B$2

(Ciardo et al., 2022).

A plausible implication is that minion homomorphisms serve as a uniform algebraic interface between structural polymorphism theory and algorithm design: the same morphism notion classifies pp-constructability, transfers minor conditions, and characterizes solvability by several relaxation families. The cited papers make this role explicit rather than incidental.

5. Classification results and canonical representatives

One of the clearest uses of minion homomorphisms is to collapse large clone classes to a small number of canonical representatives. For clones determined by binary relations with coordinate projections of size at most $B$3, the class of single-sorted clones on finite sets coincides up to minion homomorphic equivalence with the class of multisorted Boolean clones determined by at most binary relations. More precisely, the two classes

1. all clones $B$4 on finite sets where $B$5 is a set of binary relations whose coordinate projections have size at most $B$6, and
2. all multisorted Boolean clones $B$7 where $B$8 is a set of at most binary relations on a finite multisorted Boolean set, coincide up to $B$9 (Barto et al., 2024).

In that setting, every relevant clone is either equivalent to $k$0 or can be described by a “description” $k$1 consisting of constraints built from the pointwise order $k$2, the relation $k$3, equality, and duality. After reduction using 2-SAT techniques, every such $k$4 collapses to exactly one among the concrete minions

$k$5

The induced order on these $k$6-classes is a countable planar lattice (Barto et al., 2024).

A parallel but domain-specific classification appears for Mal’cev clones on a three-element set. Every clone containing a quasi-Mal’cev operation is minor-equivalent to exactly one clone from the list

$k$7

These classes are separated by explicit minor conditions, and the resulting subposet of $k$8 is completely described (Fioravanti et al., 9 Aug 2025).

These examples show that the minion homomorphism preorder is often far smaller and more tractable than the inclusion order on clones. In both papers, large families of concrete clones collapse to a countable or finite set of core behaviors. This suggests that height-1 structure can be rigid enough for classification even when full clone theory remains unwieldy.

6. Minion cores and canonical forms

The notion of a minion core is the minion-theoretic analogue of the core of a relational structure. A minion $k$9 is a minion core if every endomorphism $\alpha:[n]\to[m]$00 is a minion automorphism. A minion core of $\alpha:[n]\to[m]$01 is a minion core $\alpha:[n]\to[m]$02 such that $\alpha:[n]\to[m]$03. If each $\alpha:[n]\to[m]$04 is finite, then $\alpha:[n]\to[m]$05 has a minion core, and this core is unique up to minion isomorphism (Barto et al., 2024).

This gives canonical representatives for equivalence classes under mutual minion homomorphisms. In the class of multisorted Boolean clones determined by at most binary relations, the canonical minions $\alpha:[n]\to[m]$06 and their one-sorted analogues are all minion cores. The proof uses the fact that in the Boolean case it suffices to control the binary part: if every endomorphism is bijective on arity $\alpha:[n]\to[m]$07, then the minion is a core (Barto et al., 2024).

A related core phenomenon appears in the broader abstract theory. In the hom-order of finitely representable minions, a coatom class can be represented by the idempotent minion $\alpha:[n]\to[m]$08, and an explicit core representative of that class is identified as

$\alpha:[n]\to[m]$09

The existence of core representatives is therefore not only an ad hoc device for a particular clone family but part of a wider structural pattern in minion homomorphism orders (Juhrich, 17 Mar 2025).

The significance of minion cores is methodological as much as classificatory. Once a $\alpha:[n]\to[m]$10-class is represented by a core, endomorphism redundancy has been eliminated, and the remaining object reflects height-1 behavior as canonically as possible. This is why recent classification papers often present results directly in terms of cores rather than arbitrary members of a minion-equivalence class.

7. Conceptual significance and current picture

Across the cited works, minion homomorphisms emerge as the morphisms that survive after composition is forgotten but minor structure is retained. They preserve exactly height-1 identities, correspond to pp-constructability in finite-domain CSP/PCSP theory, characterize solvability by several standard relaxation families, and induce rich but tractable homomorphism orders on minions (Barto et al., 2024).

The current literature also shows that these orders can be surprisingly structured. In restricted settings they collapse to explicit lattices of cores; in the abstract finitely generated setting they can embed $\alpha:[n]\to[m]$11, contain dense intervals and uncountable antichains, and support bi-Heyting operations (Juhrich, 17 Mar 2025). This coexistence of tame local classification and globally intricate order structure is one of the distinctive features of minion theory.

Recent work also indicates that basis theorems for specific minions may have strong consequences for surrounding regions of the pp-constructability poset. The characterization of the Boolean idempotent clone $\alpha:[n]\to[m]$12 by quasi Maltsev plus full symmetry leads directly to the classification of its lower covers by $\alpha:[n]\to[m]$13 and finite simple groups (Meyer et al., 2024). In a different direction, classifications up to minor-equivalence for three-element Mal’cev clones and for binary small-projection clones demonstrate that substantial relational classes can be organized entirely at the level of minion homomorphisms (Fioravanti et al., 9 Aug 2025).

This suggests a broader program already visible in the literature: pass from concrete polymorphism clones to abstract minions, quotient by mutual minion homomorphism, identify minion cores, and classify the resulting poset. Within CSP and PCSP theory, minion homomorphisms are therefore not merely auxiliary maps; they are the primary carriers of the height-1 algebraic information that underlies reductions, tractability criteria, and structural taxonomy.