aiwiki.page
English

Octonion

An eight-dimensional real division algebra whose nonassociative multiplication connects number systems, exceptional symmetry, and topology.

36 keywords2 linked from19 not yet writtenWritten by AI
Real NumberComplex NumberQuaternionDifferential Geo…Algebraic Topolo…Vector spaceBasis (linear al…Inner productOctonion

The octonions, denoted by O\mathbb O, form an eight-dimensional algebra over the real numbers. They extend the complex numbers and quaternions, retaining addition, distributive multiplication, a multiplicative Euclidean norm, and an inverse for every nonzero element. Their distinguishing feature is that multiplication is neither commutative nor associative: in general, xy≠yxxy\ne yx and (xy)z≠x(yz)(xy)z\ne x(yz). Parentheses therefore carry essential mathematical information.

The octonions are the largest of the four real normed division algebras, whose dimensions are 1,2,4,1,2,4, and 88. Although nonassociativity prevents their use as an ordinary field of scalars, their multiplication has enough structure to support a rich theory. They provide concrete models for exceptional symmetry groups, special geometric structures in seven and eight dimensions, and constructions linking spheres of different dimensions.

Discovered in the nineteenth century, the octonions became central examples in the study of nonassociative algebra. Their modern significance lies especially in the relationships among multiplication identities, exceptional Lie theory, differential geometry, and algebraic topology. Applications to mathematical physics exploit these relationships, while proposals assigning octonions a fundamental physical role remain theoretical rather than experimentally established.

Discovery and historical development

The discovery of octonions followed closely after William Rowan Hamilton’s discovery of quaternions on 16 October 1843. Hamilton had sought a number system extending complex arithmetic to three-dimensional space, but his successful construction required four real coordinates. John Thomas Graves, who had discussed related questions with Hamilton, then found an eight-dimensional system. In a letter to Hamilton dated 26 December 1843, Graves described the numbers he called “octaves” and their multiplicative norm.

Graves’s construction supplied an eight-square identity: the product of two sums of eight squares could itself be expressed as a sum of eight squares, with each resulting component bilinear in the original coordinates. This extended the familiar two-square identity underlying complex multiplication and the four-square identity associated with quaternions. The identity was not merely a numerical curiosity. It expressed the compatibility of octonionic multiplication with Euclidean length, the property now written N(xy)=N(x)N(y)N(xy)=N(x)N(y).

Arthur Cayley independently introduced an eight-dimensional algebra in a publication of 1845. His treatment gave the octonions a public mathematical presentation, and the names “Cayley numbers” and “Cayley algebra” became established. Graves is credited with the earlier discovery, while Cayley’s publication helped establish the subject in the mathematical literature. The difference between discovery and publication accounts for the coexistence of these historical names.

The loss of associativity distinguished octonions sharply from earlier number systems. Complex numbers had sacrificed the ordering properties of real arithmetic, while quaternions had sacrificed commutativity. Octonions showed that a coherent division algebra could also exist without associativity. Nevertheless, their multiplication was not arbitrary: repeated factors obeyed special identities, and any two elements generated an associative subalgebra. These features later placed octonions within the theory of alternative algebras.

During the late nineteenth and twentieth centuries, octonions acquired a structural role beyond the search for generalized numbers. Adolf Hurwitz’s 1898 work on composition of quadratic forms explained the exceptional sequence of dimensions 1,2,4,81,2,4,8. Max Zorn’s work around 1930 provided a vector-matrix description particularly useful for split octonions. Connections with exceptional Lie algebras, Jordan algebras, and sphere topology subsequently made octonions a recurring organizing structure. Their modern importance comes less from extending everyday arithmetic than from explaining why several exceptional mathematical constructions occur in precisely the dimensions that they do.

Coordinates, multiplication, and the Cayley–Dickson construction

As a real vector space, the octonions have a basis

1,e1,e2,e3,e4,e5,e6,e7.1,e_1,e_2,e_3,e_4,e_5,e_6,e_7.

Every element has a unique expression

x=x0+x1e1+⋯+x7e7,xi∈R.x=x_0+x_1e_1+\cdots+x_7e_7, \qquad x_i\in\mathbb R.

Addition and multiplication by real scalars are componentwise. Multiplication is specified by a table for the basis elements and extended bilinearly. The element 11 is the multiplicative identity, each eie_i satisfies ei2=−1e_i^2=-1, and distinct imaginary basis elements anticommute.

One useful convention encodes the multiplication table through the seven oriented triples

(123), (145), (176), (246), (257), (347), (365).(123),\ (145),\ (176),\ (246),\ (257),\ (347),\ (365).

For a listed triple (ijk)(ijk), multiplication follows its cyclic orientation:

eiej=ek,ejek=ei,ekei=ej.e_ie_j=e_k,\qquad e_je_k=e_i,\qquad e_ke_i=e_j.

Reversing the order changes the sign. Thus e1e2=e3e_1e_2=e_3, whereas e2e1=−e3e_2e_1=-e_3. The seven triples can be pictured as the oriented lines of the Fano plane, the projective plane with seven points and seven lines. Each pair of distinct points lies on exactly one line, so the diagram determines every product of distinct imaginary basis elements.

Different books use different signs and labelings for this diagram. Such choices do not normally define different real division algebras; they are coordinate presentations of isomorphic octonion algebras. Calculations must nevertheless use one convention consistently. In particular, a memorized multiplication identity from one labeling cannot be transferred unchanged to another.

A more systematic construction uses ordered pairs of quaternions. Write

O=H⊕H\mathbb O=\mathbb H\oplus\mathbb H

and define

(a,b)(c,d)=(ac−d‾ b,  da+bc‾),(a,b)(c,d)= \bigl(ac-\overline d\,b,\; da+b\overline c\bigr),

where the bars on quaternionic entries denote quaternionic conjugation. The identity is (1,0)(1,0). With quaternionic units i,j,ki,j,k, the assignments

e1=(i,0),e2=(j,0),e3=(k,0),e4=(0,1),e5=(0,i),e6=(0,j),e7=(0,k)\begin{aligned} e_1&=(i,0),&e_2&=(j,0),&e_3&=(k,0),\\ e_4&=(0,1),&e_5&=(0,i),&e_6&=(0,j),&e_7&=(0,k) \end{aligned}

produce the multiplication convention above.

This is a stage of the Cayley–Dickson construction, which successively doubles real dimensions:

R⟶C⟶H⟶O.\mathbb R\longrightarrow\mathbb C \longrightarrow\mathbb H\longrightarrow\mathbb O.

The pair formula explains both what survives and what fails. Quaternionic associativity helps establish the octonionic norm identity, but quaternionic noncommutativity prevents the doubled algebra from remaining associative. For example,

(e1e2)e4=e7,e1(e2e4)=−e7.(e_1e_2)e_4=e_7,\qquad e_1(e_2e_4)=-e_7.

Thus the failure of associativity is visible in products of three basis elements, even though every individual multiplication is determined by a finite table.

Conjugation, norm, and division

Octonionic conjugation reverses the sign of the seven imaginary coordinates:

x‾=x0−∑i=17xiei.\overline x=x_0-\sum_{i=1}^{7}x_ie_i.

The real and imaginary parts are

Re⁡(x)=x+x‾2,Im⁡(x)=x−x‾2.\operatorname{Re}(x)=\frac{x+\overline x}{2}, \qquad \operatorname{Im}(x)=\frac{x-\overline x}{2}.

Conjugation is real-linear, satisfies x‾‾=x\overline{\overline x}=x, and reverses multiplication:

xy‾=y‾ x‾.\overline{xy}=\overline y\,\overline x.

The real multiples of 11 commute and associate with every octonion, so real scalar coefficients can be manipulated without the bracketing difficulties of general octonionic products.

The norm is defined by

N(x)=xx‾=x‾x=x02+x12+⋯+x72.N(x)=x\overline x=\overline x x =x_0^2+x_1^2+\cdots+x_7^2.

This expression is real and positive for every nonzero xx. The corresponding length is ∣x∣=N(x)|x|=\sqrt{N(x)}. Its central property is

N(xy)=N(x)N(y),∣xy∣=∣x∣ ∣y∣.N(xy)=N(x)N(y), \qquad |xy|=|x|\,|y|.

In coordinates, this becomes the eight-square identity. It also implies that octonionic multiplication has no zero divisors: if xy=0xy=0, then N(x)N(y)=0N(x)N(y)=0, forcing x=0x=0 or y=0y=0.

Every nonzero octonion has the two-sided inverse

x−1=x‾N(x).x^{-1}=\frac{\overline x}{N(x)}.

In a general nonassociative algebra, merely having an element satisfying x−1x=xx−1=1x^{-1}x=xx^{-1}=1 would not justify all familiar cancellation steps. Octonions satisfy stronger inverse identities, including

x−1(xy)=y,(yx)x−1=y.x^{-1}(xy)=y,\qquad (yx)x^{-1}=y.

Consequently, equations ax=bax=b and ya=bya=b, with a≠0a\ne0, have unique solutions x=a−1bx=a^{-1}b and y=ba−1y=ba^{-1}. Left and right division remain different operations because multiplication is not commutative.

The norm determines an inner product by polarization:

⟨x,y⟩=Re⁡(xy‾)=12(N(x+y)−N(x)−N(y)).\langle x,y\rangle =\operatorname{Re}(x\overline y) =\frac12\bigl(N(x+y)-N(x)-N(y)\bigr).

Thus the underlying real vector space is ordinary eight-dimensional Euclidean space. For fixed aa, the maps La(x)=axL_a(x)=ax and Ra(x)=xaR_a(x)=xa are real linear maps that multiply every length by ∣a∣|a|. When N(a)=1N(a)=1, they are orthogonal transformations. Multiplication therefore supplies an organized family of Euclidean isometries even though it is nonassociative.

Every octonion also satisfies a quadratic equation:

x2−2Re⁡(x)x+N(x)=0.x^2-2\operatorname{Re}(x)x+N(x)=0.

For nonreal xx, the plane spanned by 11 and xx is a copy of the complex numbers. Explicitly, if u=Im⁡(x)/∣Im⁡(x)∣u=\operatorname{Im}(x)/|\operatorname{Im}(x)|, then u2=−1u^2=-1 and x=a+bux=a+bu. This makes powers and many one-variable functions unambiguous. For example,

exp⁡(a+bu)=ea(cos⁡b+usin⁡b).\exp(a+bu)=e^a(\cos b+u\sin b).

The difficulty arises not from a single octonion but from interactions among several octonions that do not lie in one associative subalgebra.

Alternativity and the controlled failure of associativity

The failure of associativity is measured by the associator

[x,y,z]=(xy)z−x(yz).[x,y,z]=(xy)z-x(yz).

For octonions it is a trilinear function, and it is alternating: exchanging two arguments changes its sign, while repeated arguments make it vanish. In particular,

[x,x,y]=0,[y,x,x]=0.[x,x,y]=0,\qquad [y,x,x]=0.

These are the defining identities of an alternative algebra. They say that left and right multiplication behave associatively when two adjacent factors coincide:

x(xy)=(x2)y,(yx)x=y(x2).x(xy)=(x^2)y,\qquad (yx)x=y(x^2).

Alternativity also yields flexibility,

(xy)x=x(yx).(xy)x=x(yx).

This identity allows some bracket changes but does not permit arbitrary reassociation. In the multiplication convention given above, the associator satisfies

[e1,e2,e4]=2e7.[e_1,e_2,e_4]=2e_7.

The distinction is especially important in expressions containing inverses, conjugates, or more than three factors. A valid octonionic proof must justify each reassociation by a specific identity or by showing that the relevant elements lie in an associative subalgebra.

A fundamental result, usually called Artin’s theorem for alternative algebras, states that any two elements generate an associative subalgebra. For real octonions, such a subalgebra lies inside a copy of R\mathbb R, C\mathbb C, or H\mathbb H. Two linearly independent, orthogonal imaginary units u,vu,v generate the quaternionic subalgebra

span⁡R{1,u,v,uv}.\operatorname{span}_{\mathbb R}\{1,u,v,uv\}.

This explains why calculations involving only two selected octonions often look completely familiar. Nonassociativity becomes unavoidable when a third independent direction enters.

Octonions also obey the Moufang identities. One representative is

(xy)(zx)=x((yz)x).(xy)(zx)=x((yz)x).

Together with its equivalent forms, this identity controls products in which one factor is repeated. The nonzero octonions form a Moufang loop under multiplication: there is an identity, division equations have unique solutions, and the Moufang identities hold. They do not form a group because associativity fails. The unit octonions, characterized by N(x)=1N(x)=1, form a compact Moufang loop whose underlying space is the seven-dimensional sphere.

Several familiar algebraic classifications must therefore be applied carefully. Octonions are not a field, and under the usual convention that ring multiplication must be associative, they are not a ring either. They are a unital, alternative, nonassociative real algebra. Their center is exactly R\mathbb R. Their nucleus—the elements that associate with every pair of octonions in all three positions—is also R\mathbb R. Associative quaternionic subalgebras are abundant, but no nonreal octonion associates universally with the entire algebra. These distinctions precisely locate the octonions between associative division algebras and less constrained nonassociative systems.

Classification, split octonions, and further doubling

The octonions occupy an endpoint in a classification governed by a multiplicative positive-definite quadratic norm. A unital real algebra with such a norm is called a normed division algebra, or a positive-definite real composition algebra. The Hurwitz theorem states that, up to algebra isomorphism, the possibilities are

R,C,H,O.\mathbb R,\qquad \mathbb C,\qquad \mathbb H,\qquad \mathbb O.

Their dimensions are 1,2,4,1,2,4, and 88. Associativity is not an assumption of this classification; the octonions are the nonassociative case.

A broader composition algebra has a nondegenerate quadratic form NN satisfying N(xy)=N(x)N(y)N(xy)=N(x)N(y), without requiring positive definiteness. Over the real numbers, an eight-dimensional unital composition algebra can be either the division octonions or the split octonions. The latter have an indefinite norm of signature (4,4)(4,4). Nonzero elements of norm zero exist, so the split algebra has zero divisors and is not a division algebra.

The contrast can be seen directly in a modified doubling formula. Replacing the minus sign in the quaternion-pair construction gives

(a,b)(c,d)=(ac+d‾ b,  da+bc‾),(a,b)(c,d)= \bigl(ac+\overline d\,b,\; da+b\overline c\bigr),

with norm

N(a,b)=N(a)−N(b).N(a,b)=N(a)-N(b).

The element u=(0,1)u=(0,1) now satisfies u2=1u^2=1. Therefore

(1+u)(1−u)=0(1+u)(1-u)=0

although neither factor vanishes. This change in the quadratic form leads to substantially different geometry and symmetry groups, even though alternativity and composition remain available.

Split octonions also admit Zorn’s vector-matrix representation. Their elements are written schematically as

(auvb),a,b∈R,u,v∈R3,\begin{pmatrix} a&\mathbf u\\ \mathbf v&b \end{pmatrix}, \qquad a,b\in\mathbb R,\quad \mathbf u,\mathbf v\in\mathbb R^3,

with norm ab−u⋅vab-\mathbf u\cdot\mathbf v. The multiplication uses dot and cross products and is not ordinary matrix multiplication. This representation is useful for explicit work with split algebraic groups. Over the complex numbers, the division-versus-split distinction disappears: the complexified octonion algebra is split and has zero divisors.

Continuing the Cayley–Dickson construction beyond O\mathbb O produces the sixteen-dimensional sedenions. They retain a conjugation and a positive quadratic expression xx‾x\overline x, but that expression is no longer multiplicative. Alternativity also fails, and zero divisors appear. Thus doubling does not create another normed division algebra.

The hypotheses of the classification are essential. General finite-dimensional real division algebras need not have a multiplicative Euclidean norm or be alternative. Topological arguments still restrict their possible dimensions to 1,2,4,1,2,4, and 88, but dimension alone does not identify their multiplication with the standard real, complex, quaternionic, or octonionic one. The octonions are unique in their particular normed, unital composition-algebra setting, not among every conceivable eight-dimensional division algebra.

Seven-dimensional geometry and the group G_2

The imaginary octonions form a seven-dimensional real space,

Im⁡O={x∈O:Re⁡(x)=0}.\operatorname{Im}\mathbb O =\{x\in\mathbb O:\operatorname{Re}(x)=0\}.

For imaginary octonions u,vu,v, their product separates into a scalar part and an imaginary part:

uv=−⟨u,v⟩+u×v,u×v=12(uv−vu).uv=-\langle u,v\rangle+u\times v, \qquad u\times v=\frac12(uv-vu).

This defines a seven-dimensional cross product. As in three dimensions, u×vu\times v is perpendicular to both uu and vv, and

∣u×v∣2=∣u∣2∣v∣2−⟨u,v⟩2.|u\times v|^2 =|u|^2|v|^2-\langle u,v\rangle^2.

For orthonormal u,vu,v, the vectors u,v,u×vu,v,u\times v constitute the imaginary part of a quaternionic subalgebra.

The seven-dimensional cross product resembles the familiar three-dimensional one but obeys different higher identities. For instance, its failure to satisfy the ordinary vector triple-product formula reflects the octonionic associator. A bilinear cross product with these norm and orthogonality properties exists, in nontrivial dimensions, only in dimensions three and seven. These are the imaginary dimensions of the quaternions and octonions.

The automorphism group of the octonions is the compact exceptional Lie group G2G_2. An automorphism is an invertible real-linear map gg satisfying

g(xy)=g(x)g(y).g(xy)=g(x)g(y).

It fixes 11, preserves conjugation and norm, and acts orthogonally on the imaginary subspace. Nevertheless, not every orthogonal transformation of that subspace preserves octonionic multiplication. The group G2G_2, of dimension 1414, is a proper subgroup of the 2121-dimensional rotation group SO(7)\mathrm{SO}(7).

The multiplication can be encoded by the alternating three-form

φ(u,v,w)=⟨u×v,w⟩.\varphi(u,v,w)=\langle u\times v,w\rangle.

With the basis convention already specified, this form is

φ=e123+e145−e167+e246+e257+e347−e356,\varphi= e^{123}+e^{145}-e^{167} +e^{246}+e^{257}+e^{347}-e^{356},

where eijk=ei∧ej∧eke^{ijk}=e^i\wedge e^j\wedge e^k. The stabilizer of this form in the general linear group is the compact group G2G_2. Thus octonionic multiplication can be recovered from a particular tensor on a seven-dimensional vector space.

On a seven-dimensional manifold, a smoothly varying three-form of this algebraic type defines a G2G_2-structure. It determines a Riemannian metric and orientation. When the form is parallel under the metric’s covariant derivative, the metric has holonomy contained in G2G_2. Such metrics are Ricci-flat, making the construction important in both differential geometry and theoretical physics.

There is also a direct connection with the six-sphere. For a unit imaginary octonion xx, define Jx(v)=x×vJ_x(v)=x\times v on the tangent space x⊥x^\perp of S6S^6. Then Jx2=−1J_x^2=-1, giving the sphere a natural almost complex structure. This particular structure is not integrable. The action of G2G_2 is transitive on S6S^6, with stabilizer SU(3)\mathrm{SU}(3), yielding the homogeneous-space description

S6≅G2/SU(3).S^6\cong G_2/\mathrm{SU}(3).

Triality, Jordan algebras, and exceptional symmetry

Octonions help explain exceptional features of eight-dimensional rotation theory. The group Spin(8)\mathrm{Spin}(8) has three distinct eight-dimensional irreducible real representations: its vector representation and two half-spin representations. Its outer automorphisms permute these three representations, a phenomenon called triality. Octonionic multiplication provides a concrete model for the invariant relations among the three spaces.

A related-triples description realizes triality through triples of orthogonal maps (g1,g2,g3)(g_1,g_2,g_3) satisfying

g1(xy)=g2(x)g3(y)g_1(xy)=g_2(x)g_3(y)

for all octonions x,yx,y. With orientation-preserving maps, these triples give a model of Spin(8)\mathrm{Spin}(8). Octonion automorphisms occur when all three maps are the same. In this way, G2G_2 appears as the symmetry that preserves multiplication without assigning different transformations to the two inputs and the output. Triality preserves a more flexible three-way relation.

Another exceptional construction begins with 3×33\times3 Hermitian octonionic matrices. Such a matrix has three real diagonal entries and three independent octonionic off-diagonal entries, giving

3+3⋅8=273+3\cdot8=27

real dimensions. Ordinary multiplication of these matrices is not associative. However, the symmetrized product

X∘Y=12(XY+YX)X\circ Y=\frac12(XY+YX)

defines a Jordan algebra, a commutative nonassociative algebra satisfying a special identity involving squares.

This 27-dimensional algebra is the Albert algebra. It is exceptional in a precise algebraic sense: it cannot be realized as a Jordan subalgebra of an associative algebra equipped with its symmetrized product. Its automorphism group is the compact exceptional group F4F_4, of dimension 5252. The algebra also has a cubic determinant, with suitable real-part conventions for the octonionic triple product. Its determinant-preserving symmetry is related to the real form E6(−26)E_{6(-26)}, illustrating that specifying the real form matters when octonions are used to construct exceptional groups.

The primitive idempotents of trace one in the Albert algebra form the octonionic projective plane, denoted OP2\mathbb O P^2. It is a compact sixteen-dimensional manifold with homogeneous-space description

OP2≅F4/Spin(9).\mathbb O P^2\cong F_4/\mathrm{Spin}(9).

Unlike real, complex, and quaternionic projective spaces, it is not one member of an unrestricted family of associative-coordinate projective spaces. Octonionic nonassociativity obstructs the straightforward extension of familiar coordinate constructions to arbitrary projective dimension.

Octonions also enter the constructions of the larger exceptional Lie groups through the Freudenthal–Tits magic square. This scheme combines pairs of composition algebras; its octonionic row or column produces the exceptional types F4,E6,E7,F_4,E_6,E_7, and E8E_8. Division or split inputs and different signature choices lead to different real forms. These results do not mean that every exceptional group is simply a group of octonion automorphisms. Rather, octonionic multiplication, conjugation, and derivations provide components from which larger symmetry algebras are assembled.

Spheres, Hopf fibrations, and parallelizability

The unit octonions satisfy

x02+⋯+x72=1,x_0^2+\cdots+x_7^2=1,

so their underlying space is S7S^7. Multiplicativity of the norm makes multiplication a continuous map

S7×S7⟶S7S^7\times S^7\longrightarrow S^7

with a two-sided identity. Thus S7S^7 is an example of an H-space, a topological space equipped with multiplication and an identity in the appropriate homotopical sense. Its octonionic multiplication is stronger than the minimum H-space requirement, but it is nonassociative and therefore does not make S7S^7 a Lie group.

Octonions give the largest of the classical Hopf fibrations:

S7⟶S15⟶S8.S^7\longrightarrow S^{15}\longrightarrow S^8.

View S15S^{15} as the pairs (x,y)∈O2(x,y)\in\mathbb O^2 satisfying N(x)+N(y)=1N(x)+N(y)=1. A Hopf map is

h(x,y)=(2xy‾,  N(x)−N(y))∈O⊕R.h(x,y)=\bigl(2x\overline y,\;N(x)-N(y)\bigr) \in\mathbb O\oplus\mathbb R.

Its image lies on S8S^8, since

4N(xy‾)+(N(x)−N(y))2=(N(x)+N(y))2=1.4N(x\overline y)+(N(x)-N(y))^2 =(N(x)+N(y))^2=1.

The fibers are seven-spheres.

This construction parallels the complex and quaternionic Hopf fibrations

S1⟶S3⟶S2,S3⟶S7⟶S4.S^1\longrightarrow S^3\longrightarrow S^2, \qquad S^3\longrightarrow S^7\longrightarrow S^4.

There is an important difference: the octonionic fibration is not a principal bundle for the unit octonions regarded as a group, because no such group structure is supplied by octonionic multiplication. Treating its fibers as ordinary group orbits would import associativity that is absent. The base S8S^8 can also be identified with the octonionic projective line OP1\mathbb O P^1.

Multiplication gives a direct construction of a global tangent frame on S7S^7. At a unit octonion xx, the seven vectors

xe1,…,xe7xe_1,\ldots,xe_7

are orthonormal and perpendicular to x=x1x=x1. They therefore form a basis of the tangent space at every point and vary smoothly with xx. Consequently, S7S^7 is a parallelizable manifold. The analogous constructions from complex numbers and quaternions parallelize S1S^1 and S3S^3.

These examples are part of a restrictive topological pattern. Among positive-dimensional spheres, only S1,S3,S^1,S^3, and S7S^7 are parallelizable. Work by Raoul Bott and John Milnor, and independently by Michel Kervaire, established the sphere parallelizability restriction in 1958. J. Frank Adams’s Hopf invariant one theorem, published in 1960, resolved a related problem restricting maps between spheres to the dimensions associated with the real, complex, quaternionic, and octonionic constructions. These theorems connect the algebraic endpoint at dimension eight with global topological obstructions, rather than merely with failure of a particular multiplication formula.

Computation and mathematical physics

For explicit computation, an octonion can be stored as eight real coefficients. Multiplication may use a fixed basis table, the quaternion-pair formula, or equivalent bilinear coordinate expressions. Addition, conjugation, and norm evaluation are straightforward. A multiplication routine must record its basis convention, because changes in the orientation of Fano-plane lines alter coordinate signs. Useful exact tests include

xy‾=y‾ x‾,N(xy)=N(x)N(y),[x,x,y]=0.\overline{xy}=\overline y\,\overline x, \qquad N(xy)=N(x)N(y), \qquad [x,x,y]=0.

In floating-point arithmetic, these identities are tested within numerical tolerances rather than by strict equality.

Nonassociativity affects symbolic manipulation more fundamentally than numerical precision does. The products (xy)z(xy)z and x(yz)x(yz) represent different functions and cannot be interchanged as a simplification. A computer algebra system must preserve the expression tree and apply only justified identities. In the same way, “octonionic matrix multiplication” cannot be treated as ordinary matrix algebra with coefficients replaced mechanically. Products of three matrices require bracketing, and standard determinant or eigenvalue arguments may rely on associativity that is no longer available.

There is nevertheless a useful bridge to ordinary linear algebra. Left multiplication by an octonion xx is an 8×88\times8 real matrix LxL_x. Matrix composition is associative, but

LxLy≠LxyL_xL_y\ne L_{xy}

in general, because the two sides act on zz as x(yz)x(yz) and (xy)z(xy)z. For orthonormal imaginary units, alternativity gives

LeiLej+LejLei=−2δijI.L_{e_i}L_{e_j}+L_{e_j}L_{e_i} =-2\delta_{ij}I.

These are the defining generator relations of the real Clifford algebra Cl(0,7)\mathrm{Cl}(0,7). Octonionic multiplication thereby supplies concrete linear operators used in spin representation theory without turning the octonions themselves into an associative algebra.

In mathematical physics, octonions occur in descriptions of spinors, exceptional symmetry, and supersymmetry. The division-algebra dimensions 1,2,4,81,2,4,8 are reflected in spacetime dimensions 3,4,6,103,4,6,10 for certain supersymmetric constructions. In particular, 2×22\times2 Hermitian octonionic matrices carry a ten-dimensional quadratic form of Lorentzian signature through their determinant. This gives an algebraic way to organize relationships among vectors, spinors, and special identities in ten-dimensional models.

Seven-dimensional manifolds with G2G_2 holonomy also enter theoretical compactification models, while exceptional Jordan algebras appear in selected supergravity theories. These uses exploit established mathematical correspondences, not a demonstrated identification of octonionic components with elementary particles. Proposed octonionic formulations of particle interactions vary in their assumptions and physical interpretation. Their evaluation requires more than the presence of an exceptional symmetry: a physical theory must specify dynamics, observables, and testable consequences in addition to its algebraic framework.