
TL;DR
This paper revisits the Cayley-Dickson process, presenting a generalized form that allows for nonstandard multiplication and broadens the algebraic framework of hypercomplex systems.
Contribution
It introduces a generalized Cayley-Dickson process based on units not necessarily tied to orthogonal vector spaces, expanding the algebraic formulations of hypercomplex numbers.
Findings
Generalized Cayley-Dickson process developed
Matrices with nonstandard multiplication analyzed
Broader algebraic structures for hypercomplex systems proposed
Abstract
In the theory of the hypercomplex, the laws governing the algebra are based on units that are naturally associated with an orthogonal vector space, a requirement that is far from mandatory in many algebraic formulations arising in the context of the reals or the complex numbers.In this article the complementing view is held, in that the laws of hypercomplex algebra are recast in terms of quite generally posited units. Proceeding in this manner, a generalized form of the Cayley-Dickson process is examined. The representations given are regular bimodular; the resulting matrices are standard except they are allowed nonstandard multiplication for noncommutative matrix elements.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsControl and Stability of Dynamical Systems · Mathematical and Theoretical Analysis
Cayley-Dicksonia Revisited
U. Merkel
D-70569 Stuttgart, Universitätsstr. 38, [email protected]
From a 1981 manuscript, unpublished then, now re-typeset for the net community of the hypercomplex, with my compliments
Abstract
In the theory of the hypercomplex, the laws governing the algebra are based on units that are naturally associated with an orthogonal vector space, a requirement that is far from mandatory in many algebraic formulations arising in the context of the reals or the complex numbers.1††thanks: To name but two prominent examples: algebras with units 1, in the reals, and 1, in the complex numbers, where and respectively. In this article the complementing view is held, in that the laws of hypercomplex algebra are recast in terms of quite generally posited units. Proceeding in this manner, a generalized form of the Cayley-Dickson process is examined. The representations given are regular bimodular; the resulting matrices are standard except they are allowed nonstandard multiplication for noncommutative matrix elements.
keywords:
hypercomplex algebras \sepCayley-Dickson process \sepperiodic algebras \sepregular representations \sepnonstandard representations
1 Introduction
Among the infinite set of all algebras, there is one class that stands out for its members – the algebra of real numbers, the algebra of the complex numbers, the algebra of quaternions, and the octonion (or Cayley-)algebra. These algebras, known as hypercomplex algebras, share some unique properties which, in part or altogether, are absent from either other algebras or those venturing beyond.
Frobenius’ theorem: Each associative division algebra is isomorphic to either the algebra of real numbers, the algebra of complex numbers or the quaternion algebra [1-3]; for the more general class of alternative division algebras, the algebra of octonions has to be included.
Hurwitz’ theorem: Each normed composition algebra with unit element is isomorphic to one of the algebras listed above [4-7].
The nonassociativity and/or nonalternativity of an algebra gets in the way of matrix representation in its usual form. The very uniqueness of properties of hypercomplex algebras, however, calls for a continuation of the matrix-representation approach, and in 4.2 we present matrices with nonstandard multiplication which allow accommodating nonassociativity or nonalternativity.
As to the applicability of (non-)associative or (non-)alternative hypercomplex algebras in physics, the reader is referred to [8].
2 Choice of units
It is common ground in the theory of the hypercomplex to start from a vector space from which ensue units that are orthogonal to each other. To free ourselves from the limitations of this approach, here, we seek to proceed in a different manner.
Hypercomplex algebra shall be defined as a linear algebra over the field of real numbers (or complex numbers ), with general element (summation assumed)
[TABLE]
where for quadratic algebra, for quaternion algebra, and for octonion algebra.
The units satisfy the multiplication rules
[TABLE]
which is all what is needed to describe the class of quadratic algebras, denoted here . For the quaternion algebras, , and octonion algebras, , one further rule has to be added:
[TABLE]
where , and is the Kronecker symbol and the fully antisymmetric Levi-Civita symbol. In a particular basis, we have , with cyclic triples , , , , , , for octonions, and any of these triples for the quaternion subalgebra.
From Eqs. (2) and (3) we get the commutation and anticommutation relations
[TABLE]
and the association relation
[TABLE]
(which, since vanishes identically)
[TABLE]
The quaternion units are noncommutative, although associative and alternative; the octonion units are noncommutative, nonassociative, and alternative.
Due to linearity we can calculate products, commutators and associators for arbitrary quaternions and octonions of the form (1). For an arbitrary hypercomplex element , the *conjugate element *is defined by
[TABLE]
The conjugation mapping is an involution, i.e. , .
Also, the norm of is
[TABLE]
with symmetric quantities
[TABLE]
and the inverse element is
[TABLE]
In order that a quadratic form (7) becomes positive definite, the well known Hurwitz criterion requires that the main minors of the associated symmetric matrix be all positive and real. In the case of the matrix , these are simply non-negative powers of . Hence we can say that algebras , and are division algebras for real and positive, with nonisomorphic companions when and render the norm (7) degenerate or non-definite. To be precise, is the solution of the equation , and is the solution to . It is easily verified that the norm (7) satisfies the decomposition property . Hence hypercomplex algebras, with dimensions necessarily specified as 1,2,4 or 8, are composition algebras. When is nondegenerate and positive definite, the norm serves as a definition of the scalar product , and the algebra becomes a normed algebra.
3 The Cayley-Dickson process
Let be a -dimensional linear algebra over ( or ) with an identity element and involution :
[TABLE]
Then by the Cayley-Dickson process [14] a -dimensional linear algebra can be formed with identity element and involution and as subalgebra:
for all possible ordered pairs formed of , addition and scalar multiplication is defined componentwise and multiplication by the formula
[TABLE]
The set of all pairs forms a -dimensional algebra over . The identity element of , , is also the identity of , and will be denoted . The set of all elements , , forms a subalgebra of isomorphic to . Adjoined to is the special element by which we can write for an element of :
[TABLE]
multiplication is defined by
[TABLE]
and the involution mapping is defined by
[TABLE]
[TABLE]
Setting out from the real numbers , the algebras that result at the first step are with general element
[TABLE]
where , provided that . The norm is
[TABLE]
Taking values , three important nonisomorphic algebras are seen to emerge from class :
complex numbers 2. 2.
split complex numbers 3. 3.
dual numbers .
The transformations , (where in the latter case , hence ) yield what may be termed periodic algebras,111see appendix A a special subclass each, each of whose members is well-defined except for poles at .
Proceeding from , we have at the second step algebras , with general element
[TABLE]
whence we validate the relations
[TABLE]
and also,
The three nonisomorphic algebras that arise at this stage are222were it not for the evidence produced by regular bimodular representation (the subject of the next section), the existence of simple dual numbers could be revealed only from the vantage point of quaternions,* vi*z.
Hamilton’s quaternions ; 2. 2.
-quaternions ; 3. 3.
duodual numbers .
At step three, proceeding from , we are led to the algebras , with general element
[TABLE]
Eq. (17) is accompanied by the relations
[TABLE]
Again, three important nonisomorphic algebras can be discerned:
Cayley’s octonions ; 2. 2.
-octonions ; 3. 3.
tridual numbers .
At each step, up to equivalence within the class involved, we have one division algebra , one split algebra , and one algebra with nilideal , where the problem of isomorphism can be tackled in the following way: According to Jacobson [9], two Cayley-Dickson algebras and are isomorphic if their bilinear forms are equivalent, i.e., if there is a linear mapping , of into such that , , for all .
The Cayley-Dickson-type normed composition algebras allowed by Frobenius’ and Hurwitz’ theorems are exhausted after the first three steps. After step 3 is also brought to a halt alternativity, since the Cayley-Dickson process yields alternative algebras only if the initial algebra is associative; meta-Hurwitz algebras therefore fulfil but much weaker identities such as the flexibility law [12-14].
4 Representation of hypercomplex algebra
To keep this section brief, we just give three specific examples:
-rep by matrices over the field ; 2. 2.
-rep by matrices over the quaternionic skew field , and 3. 3.
-rep by matrices over the octonionic aq-field (or alternative quasi-field) .
For examples (2) and (3), rep matrices are nonstandard, although in the first example they are in perfect compliance with representation lore [10].
4.1 Representation of
Eq. (2) can be regarded as a quadratic equation which resolves into
[TABLE]
Thus we have
[TABLE]
[TABLE]
[TABLE]
4.2 Representation of and
Proceeding further, we let nonstandard multiplier positions assume the multiplication of rep matrices
[TABLE]
(a deviation entailing failure of associativity for noncommutative entries, as well as failure of alternativity for nonassociative entries). Writing for the units of the quaternionic skew field defined by
[TABLE]
we observe
[TABLE]
[TABLE]
and so on, the last in the line being
[TABLE]
and, of course,\;e_{0}\leftrightarrow\left(\begin{array}[]{cc}1&0\\ 0&1\end{array}\right).
In the same vein, writing for the units333as did Graves and Hamilton in their famous correspondence of the octonionic alternative quasi-field , where
[TABLE]
the analogue of the previous result is obtained, with nonstandard rep matrices characterizing the nonalternative sedenionic algebra ,
[TABLE]
which also concludes our journey into the realm of hypercomplex algebra.
The author is indebted to Jens Köplinger whose enthusiasm for the subject provided the inducement to rifle through piles of dust-gathering manuscripts and rescue this specimen from oblivion.
Appendix A
From Euler’s formula
[TABLE]
it follows
[TABLE]
Exponentiating both sides of (A.2) with , one obtains the power law of :
[TABLE]
(A.3) can be recast in terms of nontrivial roots of unity, . Substituting , it reads
[TABLE]
whence for
[TABLE]
The power law in the guise of (A.5) may then be generalized in such a way that it lends itself to the description of two disjoint subclasses of periodic algebras :
The power law
[TABLE]
describes two disjoint subclasses of periodic algebras, i.e., for ,* the subclass of complex periodic algebras* , and for , the split-complex subclass .* *
Proof: Case holds trivially by way of identification . Case follows from a tedious but straightforward application of (A.6):
Let
[TABLE]
or
[TABLE]
[TABLE]
[TABLE]
By definition, the relation holds for . Whence we have
[TABLE]
Which, using and , reduces to
[TABLE]
Keeping in mind that was assumed, for one verifies
[TABLE]
and for ,
[TABLE]
[TABLE]
The periodic algebras* and , , are uniquely linked to *their respective orthogonal forms, derived from (A.6) by setting :
[TABLE]
[TABLE]
where , , .
The periodic algebras and , , provide a dual description of their respective orthogonalizations (A.7) and (A.8) by way of their specific representations.
*Proof: *As forms a subalgebra of , it suffices to verify the assertion for . As was shown in section 4.1, with is associated the regular representation
[TABLE]
Settling for the lower sign, has representation
[TABLE]
and
[TABLE]
Reinterpreting, in accordance with (A.7) ((A.8)), () as ((), the following representations – under the substitution – hold true:
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
Up to poles for even, and all afford the required orthogonality and thus lay the basis for a dual description of and . In fact, the units satisfy the Hamilton relations for quaternions,
[TABLE]
while the units satisfy the Pauli relations for spin matrices,
[TABLE]
[TABLE]
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1(1) Kurosh A.G., “Lectures on general algebra”, Chelsea, New York (1963)
- 2(2) Frobenius G., “Über lineare Substitutionen und bilineare Formen”, J. f. reine u. angew. Math., 𝟖𝟒 84 \mathbf{84} , Nr 1, 1-63 (1888)
- 3(3) Bott R., Milnor J., “On the parallelizability of the spheres”, Bull. Amer. Math. Soc., 𝟔𝟒 64 \mathbf{64} , Nr 1 (Part 1), 87-89 (1958)
- 4(4) van der Blij F., “History of the octaves”, Simon Stevin, 𝟑𝟒 34 \mathbf{34} , Nr 3, 106-127 (1960/61)
- 5(5) Hurwitz A., “Über die Komposition der quadratischen Formen von beliebig vielen Variablen”, Nachr. Ges. Wiss., 309-316 (1898)
- 6(6) Albert A., “Quadratic forms permitting composition”, Ann. of Math., 𝟒𝟑 43 \mathbf{43} , Nr 1, 161-177 (1942)
- 7(7) Dubish R., “Composition of quadratic forms”, Ann. of Math., 𝟒𝟕 47 \mathbf{47} , Nr 3, 510-527 (1946)
- 8(8) Sorgsepp L., Lômus J., “About nonassociativity in physics and Cayley-Graves’ octonions”, Hadronic J., 𝟐 2 \mathbf{2} , 1388-1459 (1979)
