Quasigroup
In mathematics, a quasigroup is a set Q equipped with a binary operation \cdot such that, for all a, b \in Q, the equations a \cdot x = b and y \cdot a = b have unique solutions x, y \in Q.[1] This property ensures that left and right multiplications by any element are bijective permutations of Q, and the multiplication table of a finite quasigroup corresponds to a Latin square, where each symbol appears exactly once in every row and column.[2] Equivalently, a quasigroup can be defined using three binary operations—multiplication \cdot, left division \backslash, and right division /—satisfying identities such as x \cdot (x \backslash y) = y and (y / x) \cdot x = y.[3] Quasigroups generalize groups by relaxing the requirements of associativity and identity elements, forming a foundational structure in non-associative algebra that emerged in the late 19th century as part of early studies in algebra and combinatorics, with the term "quasigroup" coined by Ruth Moufang in 1935 during her work on non-Desarguesian projective planes.[4] Key developments in the 1930s and 1940s, including contributions from mathematicians like A. A. Sushkevich, Ruth Moufang, and R. H. Bruck, established connections to loops (quasigroups with an identity element) and medial quasigroups (those satisfying (xy)(uv) = (xu)(yv)), which are affine over abelian groups via the Toyoda-Murdoch-Bruck theorem.[5] Notable subclasses include Moufang quasigroups, which satisfy identities like x(y(xz)) = ((xy)z)x, and quandles, idempotent right-distributive quasigroups used in knot theory.[2] Examples of quasigroups abound: any group is an associative quasigroup, the integers under subtraction form a quasigroup without identity, and the operation on a finite set defined by a Latin square yields a quasigroup of that order.[1] The number of quasigroups of order n grows rapidly—12 for n=3, 576 for n=4—and every quasigroup is isotopic to at least one loop, meaning there exist bijections aligning the operations up to permutation of variables.[4] Quasigroups find applications across mathematics, including the construction of Steiner triple systems and projective planes via coordinatization, cryptography through quasigroup string transformations and stream ciphers, and knot invariants via associated quandles, as developed by David Joyce in 1982.[6][7] Their multiplication groups, permutations generated by left and right translations, provide a group-theoretic lens for studying structure and symmetry.[8]Definitions
Algebraic Definition
A quasigroup is an algebraic structure consisting of a set Q together with a binary operation \cdot: Q \times Q \to Q such that, for all a, b \in Q, the equations a \cdot x = b and y \cdot a = b possess unique solutions x, y \in Q. This divisibility condition implies that the left multiplication map L_a: Q \to Q defined by L_a(x) = a \cdot x and the right multiplication map R_a: Q \to Q defined by R_a(x) = x \cdot a are bijective for every a \in Q.[9][10] The unique solvability enables the definition of division operations within the quasigroup. The right division a / b is the unique element x \in Q satisfying a \cdot x = b, while the left division b \backslash a is the unique element y \in Q satisfying y \cdot a = b. These operations extend the binary structure into a partial magma with inverses in a generalized sense, distinguishing quasigroups from more restrictive structures like groups, which additionally require associativity and an identity element.[10][4] For a finite quasigroup of order n, the multiplication table—where rows and columns are indexed by elements of Q and entries are given by the operation—forms a Latin square of order n. In this square, each symbol from Q appears exactly once in every row and every column, reflecting the bijectivity of the multiplication maps. This connection underscores the combinatorial significance of quasigroups.[9] The concept of quasigroups emerged in the 1930s, with the term "quasigroup" coined by Ruth Moufang in her investigations of non-Desarguesian projective planes, where such structures arose naturally. This development built upon Leonhard Euler's foundational 18th-century work on Latin squares, including his studies of orthogonal arrays and the 36 officers problem.[9][11] A simple infinite example is the set of integers \mathbb{Z} under subtraction, denoted (\mathbb{Z}, -), where a \cdot b = a - b. For any a, b, c \in \mathbb{Z}, the equation a - x = b solves uniquely to x = a - b, and y - a = b solves to y = a + b, satisfying the quasigroup axioms. Loops, which are quasigroups equipped with a two-sided identity element, provide a natural extension of this structure.[10]Universal Algebra Definition
In universal algebra, a quasigroup is defined as the algebra (Q, \cdot, /, \backslash) consisting of a set Q equipped with three binary operations: multiplication \cdot, right division /, and left division \backslash, satisfying the identities y = x \cdot (x / y), \quad y = x / (x \cdot y), \quad y = (y \backslash x) \cdot x, \quad y = (y \cdot x) \backslash x for all x, y \in Q.[12] These identities ensure that the equations a \cdot y = b and x \cdot a = b are uniquely solvable for y and x, respectively, with the divisions providing the unique solutions.[13] This equational definition with three primitive operations is equivalent to the combinatorial definition using a single binary operation where left and right translations are bijective permutations of Q. Specifically, given a binary quasigroup (Q, \cdot), the division operations are uniquely determined by setting b \backslash a as the unique y such that y \cdot a = b and a / b as the unique x such that a \cdot x = b; conversely, the multiplication operation is uniquely determined by the division operations satisfying the defining identities.[13] The class of all quasigroups forms a variety in the sense of universal algebra, meaning it is defined by identities and thus closed under the formation of subalgebras, homomorphic images, and arbitrary direct products.[12] A derived ternary operation on a quasigroup can be introduced as [x, y, z] = (x \backslash y) \cdot z = x \cdot (y / z), where the equality holds by a derived identity of quasigroups ensuring consistency between left and right solvability; this ternary operation satisfies the projection identities [x, y, y] = x and [x, x, z] = z, and more generally the five-variable identities [[x,y,z],u,v] = [[x,u,v],y,z] = [x,[y,z,u],v] = [x,y,[u,v,z]].[12] The left and right multiplications are defined as the maps L_a: Q \to Q given by L_a(x) = a \cdot x and R_a: Q \to Q given by R_a(x) = x \cdot a for each a \in Q; the defining identities imply that each L_a and R_a is a bijection (permutation) on Q.[13]Basic Structures
Loops
A loop is a quasigroup (Q, \cdot) equipped with a two-sided identity element e \in Q satisfying e \cdot x = x \cdot e = x for all x \in Q.[14] This structure generalizes groups by relaxing the associativity axiom while preserving the ability to perform unique divisions. The term "loop" was coined in the early 1940s by A. A. Albert and his collaborators in Chicago, evoking the city's famous Loop district and rhyming with "group."[15] Every loop is a quasigroup, as the presence of the identity ensures that left and right multiplications remain bijective, allowing unique solutions to equations of the form a \cdot x = b and y \cdot a = b. However, the converse fails: not every quasigroup is a loop, since some lack an identity element. A standard example is the set of integers \mathbb{Z} under subtraction, where a * b = a - b; this operation yields a quasigroup, as divisions are uniquely solvable (e.g., solving a - x = b gives x = a - b), but no element e satisfies e - x = x for all x \in \mathbb{Z}.[16] In a loop, each element x \in [Q](/page/Q) admits a unique left inverse x^l such that x^l \cdot x = e and a unique right inverse x^r such that x \cdot x^r = e, with x^l and x^r not necessarily equal in non-associative cases. These inverses follow directly from the quasigroup divisions: x^l = x \backslash e and x^r = e / x. Groups provide the associative case of loops, where left and right inverses coincide and the operation satisfies (xy)z = x(yz) for all x, y, z \in [Q](/page/Q). A prominent non-associative example arises from the multiplication of nonzero octonions, which forms a loop (specifically, a Moufang loop) on the 8-dimensional real vector space excluding zero.[14]Latin Squares
A quasigroup on a finite set of order n is in one-to-one correspondence with an n \times n Latin square, where the Cayley table of the quasigroup—listing the products a \cdot b for elements a, b in the set—serves as the Latin square, with rows and columns indexed by the elements and entries being the symbols from the set.[17] This bijection arises because the quasigroup axioms ensure that left and right multiplications are bijective, guaranteeing that each symbol appears exactly once in every row and column of the table.[17] Conversely, any Latin square defines a quasigroup operation via its table, as the unique entry in each position satisfies the solvability conditions for the quasigroup equations.[18] If the Latin square is reduced—meaning its first row and first column are in natural order (1 to n)—then the corresponding quasigroup is a loop, possessing a two-sided identity element.[17] More generally, two quasigroups are isotopic if there exist bijections \alpha, \beta, \gamma on the underlying set such that \alpha(x) \cdot' y = \gamma(x \cdot \beta(y)) for all x, y, and this relation corresponds precisely to the isotopism of their associated Latin squares, where an isotopism is a triple of permutations on rows, columns, and symbols that transforms one square into the other.[17] Isotopic quasigroups thus share structural similarities, such as having isomorphic multiplication groups, though they may differ in properties like the presence of an identity.[8] Latin squares associated with quasigroups can be constructed using orthogonal mates: if a Latin square L has an orthogonal mate M—another Latin square such that the pairs (L(i,j), M(i,j)) are all distinct—then combining them yields a set of mutually orthogonal Latin squares, each corresponding to a quasigroup operation that can be composed to form more complex structures.[17] A prominent construction yields Steiner quasigroups from finite projective planes: given a projective plane of order n (with n^2 + n + 1 points), the points form the underlying set of a Steiner quasigroup satisfying the identities x \cdot x = x, x \cdot y = y \cdot x, and x \cdot (x \cdot y) = y, where the operation is defined geometrically using lines of the plane, and its Cayley table is a symmetric idempotent Latin square.[19] Such quasigroups exist whenever a projective plane of order n does, which occurs for all prime power n.[20] The bijection extends to infinite quasigroups: for any set Q, a quasigroup operation on Q yields a (possibly infinite) Latin square indexed by Q, where each row and column is a bijection from Q to itself, ensuring unique solvability of the quasigroup equations. In the countable infinite case, this corresponds to a countable Latin square, as seen in constructions over infinite sets like the rationals or integers under suitable operations, maintaining the bijective multiplication properties without finiteness restrictions.[18]Examples
Finite Quasigroups
The quasigroup of order 1 consists of a single element e with the operation defined by e \cdot e = e. This structure is a loop, serving as the trivial example where the unique element acts as the identity.[21] For order 2, there is a unique quasigroup up to isomorphism, given by the set \{e, a\} with the multiplication table: \begin{array}{c|cc} \cdot & e & a \\ \hline e & e & a \\ a & a & e \\ \end{array} This quasigroup is a loop with identity e, isomorphic to the cyclic group \mathbb{Z}_2.[21] Of order 3, there exist five non-isomorphic quasigroups, only one of which is a loop: the cyclic group \mathbb{Z}_3 under addition, with elements \{0, 1, 2\} and operation modulo 3. The remaining four are non-loops, illustrating early examples where the Latin square representation yields distinct algebraic structures without an identity element.[21] Quasigroups also arise from finite fields through planar functions, which generate commutative quasigroups on the field elements and coordinatize affine planes. For a finite field \mathbb{F}_q and a planar function f: \mathbb{F}_q \to \mathbb{F}_q, the operation x * y = f(x - y) + y defines a commutative quasigroup whose properties reflect the geometric structure of the associated affine plane of order q. These quasigroups are quasifields when equipped with additional division properties, enabling the construction of non-Desarguesian affine planes.[22] Steiner triple systems provide another construction of idempotent quasigroups. Given a Steiner triple system STS(v) on a set of v \equiv 1 or $3 \pmod{6} points, define the operation by x * x = x for all x, and for x \neq y, x * y = z where \{x, y, z\} is the unique triple containing x and y. The resulting quasigroup is idempotent, commutative, and totally symmetric, with the property that x * y = y * x and x * (x * y) = y. This correspondence links combinatorial designs directly to algebraic structures, as seen in the Fano plane yielding an STS(7) and its associated quasigroup.[23]Infinite Quasigroups
Infinite quasigroups arise in various algebraic contexts, providing structures on infinite sets where the operation ensures unique solvability of equations without requiring associativity or an identity element. A prominent example is the set of integers \mathbb{Z} equipped with the subtraction operation, defined by x \cdot y = x - y. This forms a quasigroup because, for any fixed a, b \in \mathbb{Z}, the equations a - x = b (solving for x = a - b) and y - a = b (solving for y = a + b) have unique solutions, and similarly for other combinations. However, it is not a loop, as there is no identity element satisfying x - e = x and e - x = x for all x \in \mathbb{Z}.[24][14] Similarly, the real numbers \mathbb{R} under subtraction x \cdot y = x - y constitute an infinite quasigroup. The operation is closed, and the left and right multiplications are bijective: solving a - x = b yields x = a - b, and y - a = b yields y = a + b, with unique solutions in \mathbb{R}. Like the integer case, it lacks an identity element, rendering it a non-loop quasigroup. This structure extends the additive group (\mathbb{R}, +) via conjugation, highlighting how quasigroups can be derived from groups by altering the operation.[14] Vector spaces over fields also yield infinite quasigroups when equipped with componentwise subtraction. Consider a vector space V over a field F (such as \mathbb{R} or \mathbb{Q}), viewed as F^\kappa for some infinite cardinal \kappa (e.g., countable dimension), with the operation (v \cdot w)_i = v_i - w_i for each component index i \in \kappa. Since F under subtraction is a quasigroup, the componentwise extension preserves the property: each equation has a unique componentwise solution, as the operations act independently in each coordinate. This construction generalizes to arbitrary infinite-dimensional spaces, providing quasigroups of any desired infinite cardinality.[14] Free quasigroups offer another fundamental class of infinite examples. The free quasigroup generated by a set X is the free algebra in the variety of quasigroups on the generators X, constructed as the free extension of the empty partial Latin square on X. For infinite |X| = \lambda, the cardinality of this free quasigroup is \lambda, as the terms are finite expressions in the three binary operations (multiplication, left division, right division) modulo the quasigroup axioms, yielding a set of size at most \lambda^{\aleph_0} = \lambda under the axiom of choice. These structures are universal objects embedding any map from X into a quasigroup.[14] Beyond specific constructions, the diversity of infinite quasigroups is vast. For each infinite cardinal \lambda, there exist uncountably many pairwise non-isomorphic quasigroups of cardinality \lambda. This abundance is exemplified in subclasses like Steiner quasigroups, where strongly minimal Steiner triple systems (each coordinatizable by a quasigroup) include uncountably many non-isomorphic models of countable cardinality.[25]Symmetries
Semisymmetry
A quasigroup (Q, \cdot) is semisymmetric if it satisfies the identity (x \cdot y) \cdot x = y for all x, y \in Q.[26] This condition ensures a form of cyclic symmetry in the multiplication table under the action of the cyclic group C_3.[26] Equivalent characterizations include the identity x \cdot (y \cdot x) = y, or y \cdot (x \cdot y) = x, or (y \cdot x) \cdot y = x.[27] In terms of translations, semisymmetry holds if and only if, for every x \in Q, the left translation L_x: y \mapsto x \cdot y and the right translation R_x: y \mapsto y \cdot x are mutual inverses, satisfying L_x \circ R_x = \mathrm{id} and R_x \circ L_x = \mathrm{id}.[28] Examples of semisymmetric quasigroups include structures derived from abelian groups (G, +) equipped with the operation x \cdot y = -x - y.[28] For instance, the integers \mathbb{[Z](/page/Z)} under this operation form an infinite semisymmetric quasigroup. Semisymmetric quasigroups also arise in combinatorial designs, such as extended Mendelsohn triple systems, where blocks correspond to cyclically ordered submultisets of Q.[26] Semisymmetric quasigroups are flexible, satisfying the identity x \cdot (y \cdot x) = (x \cdot y) \cdot x for all x, y \in Q, since both sides equal y by the defining identities.[27]Triality
In quasigroup theory, triality refers to a specific cyclic symmetry in the language of quasigroups, arising from the natural action of the alternating group A_3 (the cyclic subgroup of order 3 in S_3) on the six parastrophes (conjugate operations) of the quasigroup, preserving the structure under cyclic permutations of the operation symbols.[29] This symmetry is equivalently expressed in binary terms as x / y = y \setminus x for all x, y \in Q, where / denotes right division and \setminus denotes left division; this equality implies that the left and right division operations coincide after a swap of arguments, reflecting the cyclic interchange.[29] Quasigroups exhibiting triality thus possess a balanced divisibility that aligns the solving mechanisms symmetrically. Examples of quasigroups with triality include commutative Moufang loops of exponent 3, where the operation satisfies x^3 = e (the identity) for all x, ensuring the multiplication group admits a triality automorphism.[30] Certain Steiner quasigroups, derived from Steiner triple systems and characterized by total symmetry (full S_3-invariance), also possess triality as a subgroup symmetry, with the quasigroup operation yielding idempotent, commutative structures where every pair of distinct elements appears uniquely in a "triple."[26] Historically, the concept of triality in algebraic structures like quasigroups draws analogy from Élie Cartan's 1925 introduction of triality for the exceptional Lie group of type D_4, later connected to the automorphism group of the octonions, which exhibit similar cyclic symmetries in their multiplication; this link has influenced studies of Moufang loops and exceptional geometries arising from quasigroups with triality.[31]Total Symmetry
A totally symmetric quasigroup is a quasigroup (Q, \cdot) in which the relation x \cdot y = z holds if and only if it holds after any permutation of x, y, and z.[32] Equivalently, for all x, y, z \in Q, x \cdot y = z implies y \cdot x = z, x \cdot z = y, z \cdot x = y, y \cdot z = x, and z \cdot y = x. This condition ensures that all six parastrophes of the quasigroup—the original multiplication and the four division operations—coincide as the same binary operation on Q.[33] Consequently, a totally symmetric quasigroup is both commutative, satisfying x \cdot y = y \cdot x for all x, y \in Q, and semi-symmetric, satisfying x \cdot (y \cdot x) = y for all x, y \in Q; the latter is a weaker property than total symmetry, as it does not require commutativity.[26] The commutativity of a totally symmetric quasigroup implies that its left multiplication maps L_a: y \mapsto a \cdot y coincide with its right multiplication maps R_a: y \mapsto y \cdot a for every a \in Q. Moreover, the semi-symmetry condition forces each such map to be an involution, satisfying L_a^2 = \mathrm{id}_Q = R_a^2. These symmetries extend to the division operations, making the quasigroup highly symmetric in its algebraic structure. While the variety of totally symmetric quasigroups is defined by these permutation identities, not all members satisfy additional identities such as mediality, (x \cdot y) \cdot (u \cdot v) = (x \cdot u) \cdot (y \cdot v), though medial totally symmetric quasigroups form an important subclass with applications in abelian group isotopes.[34] Examples of totally symmetric quasigroups include the elementary abelian $2-groups equipped with their group operation; for instance, the Klein four-group (\mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z}, +)satisfies the total symmetry conditions because addition over\mathbb{F}_2is commutative andx + (y + x) = yholds due to2x = 0. More generally, any vector space over the field \mathbb{F}_2 with componentwise addition yields a totally symmetric quasigroup, providing infinite examples as well. A finite non-group example arises from the cyclic group of order $3 with a modified operation, but such constructions are limited; in fact, the only groups that are totally symmetric quasigroups are precisely the elementary abelian $2$-groups.[35] Totally symmetric idempotent quasigroups, which additionally satisfy x \cdot x = x for all x \in Q, are closely connected to combinatorial designs. Specifically, given such a quasigroup on a set S, the collection of triples B = \{\{x, y, x \cdot y\} \mid x, y \in S, x \neq y\} forms a Steiner triple system of order |S|, where every pair from S appears in exactly one triple. Conversely, every Steiner triple system on S determines a unique idempotent totally symmetric quasigroup on S by defining x \cdot y = z if \{x, y, z\} is a triple in the system (and x \cdot x = x). This correspondence links total symmetry to finite geometries and design theory.[13]Total Antisymmetry
A totally antisymmetric quasigroup, also known as a totally anti-symmetric (TA) quasigroup, is an idempotent quasigroup (Q, ·) in which x · x = x for all x ∈ Q, and the operation is anticommutative in the weak sense that x · y = y · x implies x = y for all x, y ∈ Q.[36] To achieve "total" antisymmetry for applications such as error detection, the structure additionally satisfies that for all c, x, y ∈ Q, (c · x) · y = (c · y) · x implies x = y, ensuring that adjacent transpositions alter the result unless the transposed elements are equal.[36] In additive notation, the operation often takes the form x + y ≡ - (y + x) \pmod{n} for finite orders n, reflecting the skew-symmetry of the multiplication table where entries in symmetric positions sum to 0 modulo n. Such quasigroups exist for every finite order n except n=2 and n=6.[36] These quasigroups are related to anticommutative magmas, where the operation opposes commutativity, and the idempotence property aligns with involutory behavior in loop contexts, though TA quasigroups are typically not loops due to the fixed-point-free nature of loop identities.[36] If the quasigroup is a loop with identity e, the idempotence generalizes to x · x = e, making the operation involutory and reinforcing the anticommutative structure by ensuring each element is its own inverse in a paired sense. The division operations in quasigroups provide the unique solvability required for this symmetry, as the right division y / x solves y · z = x, linking the antisymmetry to the bijectivity of multiplications. Examples include finite TA quasigroups of order 10 used in the Damm check digit algorithm, where the multiplication table is constructed to satisfy the antisymmetry conditions for error detection in decimal codes.[36] Infinite examples arise in vector spaces equipped with cross product-like operations that can be extended to quasigroup structures in higher odd-dimensional spaces with compatible bilinear forms.[36]Properties
Multiplication Operators
In a quasigroup (Q, \cdot), the left multiplication operator by a fixed element a \in Q is the map L_a: Q \to Q defined by L_a(x) = a \cdot x for all x \in Q, while the right multiplication operator is R_a: Q \to Q given by R_a(x) = x \cdot a. These operators arise naturally from the binary operation and capture the action of multiplication from each side.[13][8] The bijectivity of L_a and R_a follows directly from the quasigroup axioms, which guarantee unique solvability of the equations a \cdot x = b and x \cdot a = b for any b \in Q; thus, both maps are permutations of Q. This property embeds the quasigroup into the symmetric group \mathrm{Sym}(Q) via the left regular representation \phi_L: Q \to \mathrm{Sym}(Q) given by \phi_L(a) = L_a, and analogously via the right regular representation \phi_R(a) = R_a. These representations provide a permutation-based view of the quasigroup structure, where the image of \phi_L (or \phi_R) acts regularly on Q.[13][37][8] For loops, which are quasigroups equipped with an identity element e \in Q satisfying e \cdot x = x = x \cdot e for all x \in Q, the operators L_e and R_e coincide with the identity permutation on Q, thereby fixing the identity. In general quasigroups (without requiring an identity), the sets \{L_a \mid a \in Q\} and \{R_a \mid a \in Q\} generate the translation group of the quasigroup, also termed the multiplication group \mathrm{Mlt}(Q) \leq \mathrm{Sym}(Q), which is the subgroup spanned by all left and right multiplications and encodes the transitive permutation action central to quasigroup theory.[13][8][37]Inverse Properties
In quasigroups, the left inverse property (LIP) is defined by the existence of a permutation \lambda on the underlying set Q such that x^\lambda \cdot (x \cdot y) = y for all x, y \in Q.[38] This condition ensures that left multiplication by x can be "undone" on the left in a consistent manner, facilitating unique recovery of y.[38] Analogously, the right inverse property (RIP) holds if there exists a permutation \rho such that (x \cdot y) \cdot y^\rho = x for all x, y \in Q.[38] These properties strengthen the unique solvability inherent to quasigroups by introducing inverse mappings that align with the binary operation. A loop possessing both the LIP and RIP is termed an inverse property loop (IP-loop), where the permutations \lambda and \rho coincide with the two-sided inverse map x \mapsto x^{-1}.[38] In IP-loops, the inverse operation interacts seamlessly with the groupoid structure, leading to identities such as x^{-1} (x y) = y = (y x) x^{-1}.[39] Such loops exhibit enhanced algebraic behavior; for instance, under additional conditions like the Moufang identities, IP-loops are power-associative, meaning that powers x^n are well-defined independently of parenthesization for all x \in Q and integers n. Quasigroups inherently satisfy left and right cancellation laws due to their unique solvability: if x \cdot y = x \cdot z, then y = z, and similarly for right multiplication.[40] These laws follow directly from the existence of unique solutions to the equations x \cdot y = b and a \cdot y = b for fixed a, b \in Q.[40] In the context of inverse properties, cancellation reinforces the invertibility aspects, ensuring that distinct elements remain distinguishable under multiplication. The cross inverse property extends these ideas by requiring identities that mix division and multiplication, such as (x \backslash y) \cdot z = x \backslash (y \cdot z) for all x, y, z \in Q, where \backslash denotes left division (the unique w satisfying x \cdot w = y).[39] Equivalently, this can be expressed via permutations as x y \cdot x^\rho = y or x \cdot y x^\rho = y.[39] This property captures a form of "crossed" invertibility, where right inverses enable associative-like behavior in divisions, and it holds in structures like cross inverse property quasigroups (CIPQs).[39]Morphisms
Homotopy and Isotopy
In quasigroup theory, a homotopy between two quasigroups (Q, \cdot) and (P, *) is defined as a triple of functions (\alpha, \beta, \gamma): Q \to P satisfying the equation \alpha(x \cdot y) = \beta(x) * \gamma(y) for all x, y \in Q.[41] This relation generalizes the notion of a homomorphism by allowing three independent maps rather than a single structure-preserving function, capturing weaker structural similarities between the operations.[42] An isotopy is a special case of a homotopy where \alpha, \beta, and \gamma are all bijective.[41] Isotopy defines an equivalence relation on the class of quasigroups: reflexivity holds via the identity maps, symmetry by inverting the bijections, and transitivity by composition of the triples.[37] A principal isotopy occurs when \beta is the identity map on P, effectively conjugating the operation by permutations on the left and right factors while fixing the output labeling.[37] Isotopies preserve key structural properties of quasigroups, such as the type defined by identities or the presence of an identity element. Specifically, if (Q, \cdot) is a loop (a quasigroup with a two-sided identity), then any quasigroup isotopic to it is also a loop, as the image of the identity under the appropriate bijection serves as the identity in the isotope.[43] Conversely, every quasigroup is isotopic to some loop, allowing the normalization of quasigroups to this form without loss of essential structure.[43] For example, consider two quasigroups on the same finite set corresponding to Latin squares; an isotopism arises from permuting the rows (via \beta), columns (via \gamma), and symbols (via \alpha) in the multiplication table, yielding isotopic quasigroups with equivalent combinatorial properties.[41] Such transformations demonstrate how isotopy equates quasigroups that differ only in labeling, as seen in the case where a cyclic group quasigroup is principal-isotopic to another via left and right translations.[42]Parastrophe
In quasigroup theory, a parastrophe of a binary operation ⋅ on a set Q is obtained by rearranging the roles of the variables in the defining equation x ⋅ y = z, yielding up to six possible operations corresponding to the permutations of S_3, though for quasigroups these typically reduce to three distinct forms due to the unique solvability of equations. Specifically, given the left division x \ y (the unique z such that x ⋅ z = y) and right division x / y (the unique z such that z ⋅ y = x), the parastrophes include operations defined as x * y = y / x (where y / x is the unique z with z ⋅ x = y), x *' y = x \ y, and x ** y = y \ x (where y \ x is the unique z with y ⋅ z = x).[44] All parastrophes of a quasigroup are themselves quasigroups, as the latin square property is preserved under these rearrangements of the multiplication table.[45] Furthermore, if the original structure is a loop (a quasigroup with identity), each parastrophe retains the loop property, ensuring the existence of a two-sided identity in the new operation.[44] A conjugation of a quasigroup is an isotopy composed with a parastrophe, combining variable permutations with operation rearrangement to relate structures while preserving quasigroup axioms.[45] For example, in the case of groups—which are special quasigroups with associativity and inverses—all parastrophes yield structures isotopic to the original group, and thus isomorphic as groups.[45]Isostrophe
An isostrophe is defined as the composition of an isotopy and a parastrophe in quasigroup theory, specifically representing an isotopy between two parastrophes of quasigroups.[46] Parastrophes arise from reinterpreting the binary operation of a quasigroup (Q, \cdot) as one of its five conjugate operations, such as left division x \backslash y or right division y / x, yielding up to six distinct quasigroups associated with the original.[47] An isostrophe thus connects quasigroups that differ both in their underlying permutations (via isotopy) and in the choice of operation (via parastrophe), providing a broader equivalence relation than isotopy alone.[46] Isostrophes preserve additional structural features compared to isotopies, particularly in the multiplication groups and symmetry properties of quasigroups. For instance, if a loop is an isostrophe of a quasigroup, their middle multiplication groups coincide, and the left and right multiplication groups of the loop form normal subgroups within this structure.[46] The autotopism group of a quasigroup, which consists of triples of bijections preserving the operation, extends to include isostrophic automorphisms that account for parastrophic rearrangements, enhancing the analysis of symmetries across conjugate operations.[46] In the context of Latin squares, which are equivalent to the multiplication tables of quasigroups, isostrophes correspond to combined manipulations of orthogonal arrays. These include permuting rows, columns, and symbols (from the isotopy component) alongside conjugating the array to reflect a different operation (from the parastrophe), thereby preserving orthogonality properties in sets of mutually orthogonal Latin squares.[48]Applications
Combinatorics and Design Theory
Quasigroups find prominent applications in combinatorics and design theory through their intimate connection to Latin squares, which are precisely the multiplication tables of finite quasigroups.[21] This bijection allows quasigroup theory to underpin the algebraic study of combinatorial structures like orthogonal arrays and block designs. A historical cornerstone is Euler's 36 officers problem, posed in 1779, which seeks to arrange 36 officers—representing one from each of 6 ranks and 6 regiments—into a 6×6 square such that each row and each column contains exactly one officer from every rank and every regiment.[49] This arrangement is equivalent to constructing two mutually orthogonal Latin squares of order 6, or equivalently, two orthogonal quasigroups of order 6, where orthogonality means that the map sending each pair (x, y) to (x \cdot y, x * y) (with \cdot and * the respective operations) is bijective, ensuring every ordered pair of symbols appears exactly once in the superposition of their tables.[49] Euler conjectured no solution exists for order 6 (or more generally for orders congruent to 2 modulo 4), a claim later confirmed for n=6 by exhaustive enumeration, though disproven for larger such orders.[49] More broadly, sets of mutually orthogonal Latin squares (MOLS) of order n arise from sets of k mutually orthogonal quasigroups on an n-element set, where every pair of distinct quasigroups in the set satisfies the orthogonality condition.[50] Such constructions are central to design theory, enabling the formation of orthogonal arrays OA(n, k+2) of strength 2, which in turn yield resolvable balanced incomplete block designs and affine geometries. For instance, a complete set of n-1 MOLS corresponds to a projective plane of order n, and quasigroup prolongations—extensions preserving Latin square properties—facilitate explicit constructions of orthogonal pairs, as demonstrated for order 10 using T-quasigroups.[50] In the realm of Steiner triple systems, idempotent commutative quasigroups of order v = 6n+3 provide an algebraic model for certain STS(v), which are collections of 3-element blocks on a v-element set such that every unordered pair appears in exactly one block.[51] Specifically, a Steiner quasigroup is a totally symmetric idempotent quasigroup satisfying x \circ x = x, x \circ y = y \circ x, and x(yx) = y; its multiplication table encodes the triples via x \circ y = z where \{x, y, z\} is a block (with z = x if x = y).[51] This equivalence holds isomorphically: STS(v) exist precisely when v \equiv 1 or $3 \pmod{6}, and the quasigroup operation uniquely recovers the design, with subsystems corresponding to subquasigroups.[51] Such quasigroups are instrumental in enumerating and constructing large sets of STS, including extensions via idempotent commutative operations. Quasigroup-based error-correcting codes, often derived from Latin square transformations, offer robust mechanisms for detecting and correcting errors, including bursts.[52] These nonlinear codes exploit quasigroup operations to generate codewords with low autocorrelation, enabling correction of multiple errors in finite fields; for example, transformations using isotopic quasigroups yield codes with minimum distance proportional to the quasigroup order.[52] In burst error scenarios, cryptcodes constructed from quasigroups—such as random quasigroup-based encodings—facilitate fast decoding for image transmission over noisy channels, correcting contiguous error bursts by leveraging the bijective properties of the operations to recover original data with minimal redundancy.[53] Simulations show these codes achieve bit-error rates comparable to linear codes while resisting adversarial bursts, with performance scaling with quasigroup size.[54]Cryptography
Quasigroups find significant application in cryptography through quasigroup string transformations (QST), which leverage the non-associative binary operation to mix input strings in a nonlinear manner, providing a foundation for stream ciphers such as those proposed by Gligoroski et al. in 2004.[55] In these ciphers, QST processes plaintext sequences by iteratively applying the quasigroup operation, often combined with modular arithmetic over large primes, to generate keystreams that scramble data without relying on associative structures, thereby enhancing diffusion properties.[56] A notable example is the quasigroup encryptor described by Satti et al., which uses indexed quasigroup matrices to achieve high-entropy output even for repetitive inputs, suitable for symmetric stream encryption in resource-constrained environments.[57] Beyond stream ciphers, quasigroups underpin key agreement protocols and hash functions, often via their equivalence to Latin squares. For instance, public-key schemes like Xifrat employ restricted-commutative quasigroups to enable secure key exchange resistant to quantum attacks. Hash functions such as the Edon-R family utilize QST over quasigroups to compress inputs into fixed-length digests, ensuring collision resistance through the quasigroup's permutation properties derived from Latin square representations.[58] A comprehensive 2020 survey by Markovski highlights these uses, noting quasigroups' role in designing primitives for data integrity, digital signatures, and commitment schemes.[59] In 2024, a symmetric encryption scheme based on quasigroups with dynamic S-boxes was proposed, improving resistance to differential and linear cryptanalysis.[60] The non-associativity of quasigroups confers key advantages in cryptography, particularly resistance to linear cryptanalysis, as the lack of associative laws prevents straightforward linear approximations of the encryption function.[59] This property disrupts attacks that exploit linearity in group-based ciphers, making quasigroup operations ideal for substitution-permutation networks. Recent developments post-2020 integrate quasigroups into secure multi-party computation (MPC) protocols, enhancing privacy in distributed systems. For example, a 2025 MPC scheme for resilient coordination selects random elements from quasigroups to perform secure operations like multiplication and division, supporting applications in fault-tolerant environments such as blockchain networks.[61]Knot Theory
Quandles, defined as idempotent right-distributive quasigroups satisfying x \triangleright x = x and (x \triangleright y) \triangleright z = (x \triangleright z) \triangleright (y \triangleright z) for all x, y, z, provide algebraic invariants for knots and links. Introduced by Joyce in 1982, quandle colorings assign elements of a quandle to knot arcs such that the operation respects crossings, distinguishing knots that groups cannot, such as the trefoil and figure-eight knot.[62] Quandle homology theories, developed in the 2000s, further enhance these invariants by capturing topological features, with applications in classifying knots up to concordance and studying link homologies.[63] This connection bridges quasigroup theory with low-dimensional topology, enabling computational tools for knot recognition and enumeration.Generalizations
Multiary Quasigroups
A multiary quasigroup, also known as an n-ary quasigroup for n \geq 2, is an algebraic structure consisting of a set Q equipped with an n-ary operation q: Q^n \to Q such that, for each position i = 1, \dots, n, the equation q(x_1, \dots, x_{i-1}, y, x_{i+1}, \dots, x_n) = z has a unique solution y \in Q for any fixed x_1, \dots, x_{i-1}, x_{i+1}, \dots, x_n, z \in Q.[64] This unique solvability condition generalizes the divisibility property of binary quasigroups to higher arities, ensuring that the operation allows for invertible "divisions" in each variable.[65] When n=2, an n-ary quasigroup reduces precisely to a binary quasigroup, recovering the standard definition with left and right division.[64] For n=3, it yields a ternary quasigroup, where the operation satisfies unique solvability in each of the three positions, often studied for its connections to geometric and combinatorial structures.[66] A key property is reducibility: an n-ary quasigroup is completely reducible if and only if it arises as an iterated binary group operation, such as q(x_1, \dots, x_n) = x_1 \cdot x_2 \cdots x_n in an abelian group, without needing parenthesization due to associativity.[64] Examples of multiary quasigroups frequently derive from iterative applications of binary operations. In particular, heaps provide a canonical construction of ternary quasigroups from binary groups or quasigroups: given a group (G, \cdot) with identity e, the ternary operation [x, y, z] = x \cdot y^{-1} \cdot z defines a heap, which satisfies the unique solvability condition along with para-associativity [x, y, [z, w, v]] = [[x, y, z], w, v] and the heap axiom ensuring symmetry in certain divisions.[65] This structure captures "group-like" behavior without a specified identity, and its retracts (by fixing one variable) often yield binary quasigroups.[64] Recent advancements extend multiary quasigroups to hyperstructural settings through polyquasigroups and polylops, introduced in 2025. A polyquasigroup is a polygroupoid—a set with a hyperoperation producing subsets—equipped with hyperdivisions satisfying inclusion-based solvability conditions, such as x \in (x \cdot y) / y for all x, y, generalizing unique solvability to multi-valued outputs.[67] A polylop further requires a hyperidentity e such that x \cdot e = e \cdot x = x. Examples include finite sets like \{1,2,3,4,5,6,7\} with explicitly tabulated hyperoperations forming polyquasigroups, and subsets thereof yielding polylops with identity 1, illustrating applications in hypergroup theory and generalized algebraic systems.[67]Quasigroupoids
A quasigroupoid is a generalization of a groupoid to non-associative settings, often defined in categorical terms as a magmoid where every span and cospan admits a unique factorization.[68] In algebraic terms, as studied in recent work, a quasigroupoid \mathcal{Q} \rightrightarrows \mathcal{I} consists of a set \mathcal{Q} \supset \mathcal{I} (with \mathcal{I} \neq \emptyset) equipped with partial binary multiplication and inversion, source map \sigma: \mathcal{Q} \to \mathcal{I}, and target map \tau: \mathcal{Q} \to \mathcal{I}, such that multiplication is defined only when \tau(x) = \sigma(y), with associativity holding whenever defined, every element having an inverse, and appropriate identity conditions via the maps.[69] This structure ensures local solvability properties akin to quasigroups but in a partial, directed setting. Quasigroupoids generalize quasigroups by relaxing the totality of the operation and incorporating categorical structure, allowing undefined products while maintaining divisibility where defined.[69] They encompass loopoids as a subclass, which are partial loops featuring a partial identity element satisfying e \cdot x = x \cdot e = x whenever defined.[69] Examples of quasigroupoids include partial Latin squares, where the defined entries form a multiplication table satisfying the local quasigroup axioms, with undefined cells representing non-computable products.[70] They also appear in incomplete block designs, where the partial operation encodes incidence relations that are not specified for every pair of elements, yet satisfy solvability for defined blocks. Recent developments, published online in January 2025, introduced matrix representations for quasigroupoids, generalizing classical matrix models of groupoids to non-associative contexts by employing a parameterized family of quasigroups \Theta(\mathcal{P}, \mathcal{I}) with source \sigma and target \tau maps to define partial multiplications, such as (h, x, j) \cdot (k, y, l) = (h, \theta_{hjl}(x, y), l) when j = k.[69] This approach proves that every connected quasigroupoid admits a non-canonical matrix representation (Theorem 5), facilitating the study of their embeddings and decompositions, including loopoids into pair groupoids and loops.[69]Enumeration
Small Quasigroups
The enumeration of quasigroups of small finite orders up to isomorphism reveals a rapid increase in complexity even for modest sizes. For order 1, there is a single trivial quasigroup on the singleton set, where the operation maps the unique element to itself. For order 2, there is exactly one quasigroup up to isomorphism, which coincides with the cyclic group of order 2 and thus is a loop. For order 3, there are five quasigroups up to isomorphism, of which one is a loop (the cyclic group of order 3). These five arise from the 12 Latin squares of order 3, classified into isomorphism classes via their Cayley tables. For order 4, there are 35 quasigroups up to isomorphism, including two loops (the cyclic group of order 4 and the Klein four-group). The classification into these 35 classes was obtained by enumerating the 576 Latin squares of order 4 and grouping them by quasigroup isomorphisms, which preserve the algebraic structure. The following table summarizes the counts of quasigroups and loops up to isomorphism for orders 1 through 4:| Order | Quasigroups | Loops |
|---|---|---|
| 1 | 1 | 1 |
| 2 | 1 | 1 |
| 3 | 5 | 1 |
| 4 | 35 | 2 |
Small Loops
The enumeration and classification of small finite loops up to isomorphism focus on their structure as quasigroups with a two-sided identity element. For orders 1 through 3, there is precisely one loop of each order, and each is associative, coinciding with the cyclic group \mathbb{Z}_n. Specifically, the loop of order 1 is the trivial group, that of order 2 is \mathbb{Z}_2, and that of order 3 is \mathbb{Z}_3.[72] For order 4, there are two non-isomorphic loops, both associative and thus groups: the cyclic group \mathbb{Z}_4 and the Klein four-group \mathbb{Z}_2 \times \mathbb{Z}_2 (also known as the dihedral group of order 4). It is a known result that all loops of order at most 4 are associative.[72][73] Non-associative loops first appear at order 5, where there are six loops up to isomorphism: one associative loop, the cyclic group \mathbb{Z}_5, and five non-associative examples. The complete counts of loops up to isomorphism for small orders are summarized in the following table:| Order n | Number of loops up to isomorphism |
|---|---|
| 1 | 1 |
| 2 | 1 |
| 3 | 1 |
| 4 | 2 |
| 5 | 6 |