Nominal Sets and Finite Support

• Nominal sets are sets equipped with a group action where every element depends on only finitely many atoms.
• They formalize name abstraction, freshness, and α-equivalence by utilizing finite support and orbit-finiteness to capture binding phenomena.
• Recent developments extend nominal sets to arbitrary data symmetries and categorical frameworks, impacting automata theory and probabilistic programming.

Nominal sets are sets equipped with a symmetry action of names and constrained by finite support: one fixes a countably infinite set of atoms and lets finite permutations act on elements, while requiring that each element depend on only finitely many atoms. In the classical Gabbay–Pitts setting this gives a mathematically precise account of names, freshness, α-equivalence, and binding; later work generalizes the same pattern from equality atoms to arbitrary data symmetries given by automorphism groups, and develops categorical, algebraic, constructive, algorithmic, and mechanized variants (Bojańczyk et al., 2014, Balco et al., 2020).

1. Core notion and symmetry structure

In the standard formulation, one fixes a countably infinite set of atoms $A$ and the group $\mathrm{Perm}(A)$ or $\mathrm{Sym}_f(A)$ of finite permutations, i.e. bijections $\pi : A \to A$ that move only finitely many atoms. A nominal set $X$ is a $\mathrm{Perm}(A)$-set with action $(\pi,x)\mapsto \pi\cdot x$ such that every $x\in X$ has a finite support: a finite $S\subseteq A$ satisfying

$$\forall \pi\in \mathrm{Perm}(A).\; \bigl(\forall a\in S.\; \pi(a)=a\bigr)\Rightarrow \pi\cdot x=x.$$

Equivariant maps are exactly those functions $\mathrm{Perm}(A)$0 commuting with the action, $\mathrm{Perm}(A)$1 (Pagano et al., 2023, Balco et al., 2020).

This structure isolates dependence on names from the identities of names themselves. Orbits under the permutation action capture equality-pattern classes, and orbit-finiteness replaces ordinary finiteness in many constructions. Over equality atoms, tuples $\mathrm{Perm}(A)$2 decompose by equality pattern; over total order atoms, orbit structure is determined by relative order type, with $\mathrm{Perm}(A)$3 splitting into the three orbits $\mathrm{Perm}(A)$4, $\mathrm{Perm}(A)$5, and $\mathrm{Perm}(A)$6 (Bolognesi et al., 2016, Venhoek et al., 2019).

A major generalization replaces the equality symmetry by an arbitrary data symmetry $\mathrm{Perm}(A)$7, where $\mathrm{Perm}(A)$8 is an automorphism group of the underlying structure. A nominal $\mathrm{Perm}(A)$9-set is then a $\mathrm{Sym}_f(A)$0-set in which every element has finite support relative to $\mathrm{Sym}_f(A)$1. This supports automata and logics over ordered data, graphs, partial orders, and other structured infinite alphabets, not only pure equality (Bojańczyk et al., 2014). A further graph-induced variant replaces finite permutations of atoms by the automorphism group of the countable Rado graph, producing “Rado-nominal sets” whose supports record finite vertex dependencies (Ackerman et al., 2023).

2. Support, freshness, abstraction, and α-equivalence

Support induces freshness. In the classical setting one defines

$\mathrm{Sym}_f(A)$2

and, when least supports exist, $\mathrm{Sym}_f(A)$3 is the intersection of all finite supports of $\mathrm{Sym}_f(A)$4. This yields the usual nominal reading of “$\mathrm{Sym}_f(A)$5 does not matter for $\mathrm{Sym}_f(A)$6” (Balco et al., 2020, Gabbay et al., 2011).

The characteristic nominal construction for binding is atoms-abstraction. For a nominal set $\mathrm{Sym}_f(A)$7, the abstraction $\mathrm{Sym}_f(A)$8 quotients pairs $\mathrm{Sym}_f(A)$9 by α-renaming of the bound atom. Its action is

$\pi : A \to A$0

and, classically, abstraction removes the bound name from support: $\pi : A \to A$1 This makes α-equivalence semantic rather than merely syntactic. In lambda syntax with terms $\pi : A \to A$2, permutation acts pointwise on variables and binders, and support can be chosen as the free variables of the term. Swapping $\pi : A \to A$3 implements α-conversion; if $\pi : A \to A$4 is fresh for $\pi : A \to A$5, then $\pi : A \to A$6 and $\pi : A \to A$7 represent the same binding structure (Pagano et al., 2023, Paranhos et al., 30 Sep 2025, Balco et al., 2019).

A common classical formulation treats freshness and abstraction via least supports, but this point becomes subtle constructively. Brouwerian counterexamples show that least finite support need not exist in constructive set theory, and several seemingly natural principles about “the intersection of all supports” imply WLPO. This makes least support a genuine fault line between classical and constructive nominal mathematics rather than a merely technical choice (Swan, 2017). A constructive reformulation therefore works with “some support” and swap-based characterizations, rather than relying on least-support minimality (Paranhos et al., 30 Sep 2025, Pagano et al., 2023).

3. Constructive and mechanized perspectives

Constructive mechanizations have converged on a support-first presentation. In Agda, finite permutations are developed from bijections $\pi : A \to A$8 with finite support, equivalently bijections that are the identity outside a finite domain. The development represents finite permutations both syntactically as compositions of swaps and structurally as products of disjoint cycles, proves these representations equivalent, and defines a normalization

$\pi : A \to A$9

with correctness $X$0. Support is handled constructively via either pointwise fixing of a predicate $X$1 or the equivalent swap law “all swaps outside $X$2 fix $X$3” (Pagano et al., 2023).

The Rocq development adopts a different concrete representation of permutations, namely lists of swaps modulo setoid equivalence, with composition by list concatenation and inverse by reversal. Nominality is defined through an explicit support function and the swap-support law

$X$4

Freshness is given in existential and universal forms,

$X$5

and these are proved equivalent. The same development formalizes name abstraction, nominal α-equivalence, finitely supported functions $X$6 under conjugation, the Freshness theorem, and the Freshness Condition for Binders (Paranhos et al., 30 Sep 2025).

The constructive literature treats the nonexistence of least support as a foundational issue, not a deficiency of proof engineering. The counterexamples in constructive set theory imply that neither least finite support nor least subfinite support can be taken as globally available; even closure properties of support intersections can carry nonconstructive logical strength (Swan, 2017). A plausible implication is that constructive nominal metatheory is best organized around equivariance, some finite support, swap lemmas, and fresh-name existence, rather than around canonical least supports.

4. Categorical and algebraic formulations

The category $X$7 has nominal sets as objects and equivariant maps as morphisms. Besides the cartesian product, an important monoidal structure is the separated product

$X$8

which internalizes freshness side conditions. This structure is central in nominal string diagrams and nominal PROPs, where parallel composition is only defined on support-disjoint interfaces (Balco et al., 2020, Balco et al., 2019).

Internalizing string diagrams in $X$9 yields nominal PROPs and nominal monoidal theories. In this setting, names live on wires, permutation action renames entire diagrams, and internal names introduced by sequential composition become bound because they disappear from support. A central result is that ordinary PROPs and nominal PROPs are equivalent categories; completeness transfers between ordinary symmetric monoidal theories and their nominal counterparts (Balco et al., 2020).

Universal algebra over nominal sets can be formulated via monads and, after transport to the presheaf category $\mathrm{Perm}(A)$0, by algebraic theories over contexts. Two notions introduced for this purpose are finitary based monads and uniform monads. Uniform monads are those distributing over abstraction, hence stable under nominal binding. This framework recovers the logics of Gabbay–Mathijssen and Clouston–Pitts and yields HSP-style closure theorems adapted to abstraction (Kurz et al., 2010).

A newer synthesis gives nominal sets a substitution tensor analogous to Fiore-style substitution on presheaves. For nominal sets,

$\mathrm{Perm}(A)$1

with internal hom defined by finitely reducible equivariant maps out of fresh powers. This places nominal sets, renaming sets, and presheaf categories in a single action-based construction of closed monoidal substitution structures (Lenke et al., 12 Feb 2026).

5. Automata, computation, and algorithmics

Nominal sets have become a standard semantic base for automata over infinite alphabets. In the generalized $\mathrm{Perm}(A)$2-set perspective, deterministic and nondeterministic automata are defined exactly as in the finite case except that alphabets and state spaces are orbit-finite nominal sets and transitions are equivariant. The resulting Myhill–Nerode theory states that an equivariant language is recognized by a deterministic orbit-finite nominal automaton exactly when its equivariant syntactic quotient is orbit-finite (Bojańczyk et al., 2014).

Over ordered atoms $\mathrm{Perm}(A)$3, orbit structure simplifies because local symmetries of finite supports are trivial. This permits a direct orbit-level representation of ordered nominal sets by support dimension alone, together with explicit representations of equivariant maps by bit-strings and products by allocation strings. The ONS library uses this representation to derive concrete complexity bounds, such as $\mathrm{Perm}(A)$4 membership and $\mathrm{Perm}(A)$5 worst-case product construction (Venhoek et al., 2019).

Tree and word automata with nominal state spaces admit corresponding learning and coalgebraic theories. Deterministic bottom-up nominal tree automata satisfy a nominal Myhill–Nerode theorem and support an active learning algorithm for any data symmetry admitting least supports (Nakanishi et al., 2022). For nondeterministic orbit-finite automata and regular nondeterministic nominal automata with binding transitions, coalgebraic trace semantics and generalized determinization both recover the accepted language semantics; in the binding case the languages are bar languages modulo α-equivalence (Frank et al., 2022).

Several computational models exploit specifically nominal phenomena rather than merely orbit-finite representations. Separated nominal automata use the separated product to restrict transitions to support-disjoint inputs and can be exponentially smaller than classical nominal automata when the target language is closed under substitutions (Moerman et al., 2019). Nominal cellular automata use an infinite alphabet of pure names with only equality test, copying, and fresh-name generation; their local rules depend only on equality patterns, and they exhibit properly nominal particles and unbounded name creation not present in finite-alphabet cellular automata (Bolognesi et al., 2016).

6. Logical extensions and current directions

Nominal sets continue to serve as a semantic substrate for logics of binding, but recent work has also clarified their limits. Permissive-Nominal Logic takes semantics in permissive nominal sets or renaming sets, and a restricted subsystem translates soundly and completely to higher-order logic. What does not survive the translation is nominal equivariance: names and binding translate, but the symmetry principles native to nominal semantics do not (Dowek et al., 2023, Dowek et al., 2011).

Extensions of nominal algebra by permutation fixed-point constraints show a similar fault line. Fixed-point constraints $\mathrm{Perm}(A)$6, intended to express $\mathrm{Perm}(A)$7, are not sound in arbitrary nominal sets; counterexamples arise in $\mathrm{Perm}(A)$8 and in quotients by commutativity. Soundness can be recovered either by working in strong nominal sets, where support is characterized exactly by the permutations that fix an element, or by restricting the proof system (Caires-Santos et al., 2024).

Two recent foundational directions broaden the scope of nominal methods. Supported sets provide a symmetry-independent base category over which equality nominal sets, ordered nominal sets, and renaming sets arise monadically; this gives finite presentations of orbit-finite nominal objects and lifts a de Bruijn-style binding functor to classical nominal abstraction (Wißmann, 2022). In a different direction, graph-induced nominal sets based on automorphisms of the Rado graph support an internal probability monad and a semantics for probabilistic programming over random graphs, including Erdős–Rényi graphons (Ackerman et al., 2023).

Taken together, these developments suggest that “nominal sets” no longer names a single formalism but a family of closely related frameworks centered on symmetry action, finite dependence, and name abstraction. The classical theory over finite permutations remains the reference model, yet current research spans constructive metatheory, mechanized normalization, categorical algebra, coalgebraic semantics, substitution calculi, automata learning, and probabilistic models with structured atoms (Pagano et al., 2023, Paranhos et al., 30 Sep 2025, Lenke et al., 12 Feb 2026).