Escolar Documentos
Profissional Documentos
Cultura Documentos
www.elsevier.com/locate/amc
b,*
Abstract
A 16-dimensional CayleyDickson algebra is presented. Its unique algebraic properties, its zero-divisors, and the solutions to a general linear equation are found. A
theory of function is developed in terms of the regularity (monogenicity) conditions and
some such functions are constructed. 2000 Elsevier Science Inc. All rights reserved.
Keywords: CayleyDickson algebra; Sedenions; Monogenicity; Quaternions Octonions
1. Introduction
What we mean here by a sedenion algebra is a non-commutative, non-associative, non-alternative, but power-associative 16-dimensional Cayley
Dickson algebra with a quadratic norm and whose elements are constructed
from real numbers, R, by iterations of CayleyDickson process [1,2], where 1
is chosen as the eld parameter at each step of the construction. In such process
2N -dimensional algebras of complex numbers, C (N 1), quaternions, Q
(N 2), octonions, O (N 3), sedenions, S (N 4), and other power-associative hypercomplex numbers are successively obtained.
The fact that sedenions do not necessarily constitute a composition nor a
division algebra, the reason for which can be traced to the non-associativity of
O, and possessing properties that are not obvious within our current physical
theories, may have been the motivation for introducing concepts and modications to achieve or eliminate some specic metric and algebraic properties.
Some parts of our work paralells that of other authors [35]; but unlike the
0096-3003/00/$ - see front matter 2000 Elsevier Science Inc. All rights reserved.
PII: S 0 0 9 6 - 3 0 0 3 ( 9 9 ) 0 0 1 4 0 - X
78
2. Basic algebra
In this paper we denote sedenions by S, T, Z, in bold capital letters, and
octonions by capital letters, A, B, X, Y, etc... We dene a sedenion,
S : A; B 2 S;
A; B 2 O;
l 0; 1; . . . ; 7;
e8l : il e8
i0 e0 1:
79
The multiplication of the two basal elements, ej and ek (the Greek (Latin) indices take some or all values of 0; 1; . . . ; 15 1; . . . ; 15), satises
ej ek djk ejkm em ;
where the value of the structure constant, ejkm , which is totally antisymmetric in
its indices, is given by the usual permutation rule for a 3-indexed antisymmetric
tensor with values 0 or 1, and determined by the appropriate 35 sedenion
cycles j; k; m. In terms of the basis a real sedenion, S, may be written componentwise as
S : s0
15
X
sj e j ;
sl 2 R;
j1
P15
where s : j1 sj ej is termed as the ``pure sedenion''. The structure constant,
ejkm , is associated with the non-commutativity of sedenions by
X
sj tk ejkm em ;
3
S; T : ST TS 2
jkm
where the summation is over all possible permutations of sedenion cycles j; k; m. The non-associativity of three sedenions, S, T, and U, is measured
by the non-zero (not necessarily zero) associator, (S, T, U) : (S T) U S(T
U).
Although the multiplication of basal elements is strictly alternative (true for
the basis of any 2N -dimensional CayleyDickson algebra with eld parameter
1) it is not necessarily true for arbitrary sedenions. For two sedenions, S and
T, the associator
S; T
S; S; T A; B; D; A; B; C S;
m 6 k
does not always vanish. This translates to the fact that the second structure
constants, which deals with triple products of sedenions, are non-antisymmetric in its indices. The norm of a sedenion S, dened as
SS
nA nB
nS : SS
15
X
l0
s2l 2 R;
80
algebra. For a sedenion S A; B, which satises certain norm and ``vectorclosure'' conditions, there exist a sedenion, Z, which is both left and right zerodivisor of S,
ZS 0 SZ:
3. Zero-divisors
One cannot solve a general linear equation of sedenion in a usual manner
due to the non-alternativity of the algebra. However, we can write a sedenion
as an ordered pair of octonions. We rst deal with the case of nding the zerodivisors of sedenions. For the right zero-divisor, Z, of S,
SZ 0:
BX YA 0;
1
B; X ; Y ;
nZ
1
Y ; A; B;
X
nS
1
A; X ; Y ;
nZ
1
Y
X ; A; B:
nS
xY A; B Y ; A; B 0;
81
where
2
x :
nS
2 R:
nA; B
One can show that x 1, which implies that Eq. (8) for X reduces to
XAB X BA 0:
k0j xj 0;
10
j1
7
X
!
x j a j br
j1
7
X
!
x k b k ar 0
8r 1; . . . ; 7;
k1
P
where k0j : km ak bm ejkm :
The vector-closure conditions for zero-divisors imply
x0 0;
AX 0
7
X
aj xj 0;
j1
X AB0
7
X
BX 0
7
X
bj xj 0;
11
j1
k0j xj 0;
j1
k02
a2
b2
k03
a3 :
b3
82
Similarly for x2 and x3 . The corresponding solution for Y can now be obtained
using the appropriate equation or simply from
Y
X A; B
:
nS
S
Z;
nS
A 6 0 6 B:
: F ; G, we
Using the identity (4) and dening octonions, F and G, by ST
obtain
nSX A; B; Y F ;
12
nSY A; B; X G:
By substitution we have for X,
13
xX A; B A; B; X 2K;
where
K :
nSF A; B; G A; B
;
2nA; B
x :
nS
:
nA; B
jm
where we had used the fact that for each xed j and k,
X
X
ap bq ejkpq
k0m ejkm kjk ; kjk : aj ; bk :
pq
83
15
which satises
X
ajk a0k 0;
j 1; . . . ; 7:
w :
7
X
j1
17
bj x j ;
7
X
k0j xj k0 :
j1
The equations are now solved in a conventional manner as follows. Note that
the seven equations of (17) are imposed on x0 , v, and w, and that the rank of D
and M, where
0
1
0
1
k01 a1 b1 k1
k01 a1 b1
M : @ k02 a2 b2 k2 A;
L : @ k02 a2 b2 A;
k07 a7 b7
k07 a7 b7 k7
should be equal to or be less than three.
(i) The rank of L is three: by re-indexing the ak and bk (strictly for notational
purposes only), so that
k01
D k02
k03
a1
a2
a3
b1
b2 6 0;
b3
84
we nd
x0
D1
;
D
D2
;
D
D3
;
D
k0k xk k0 ;
k1
7
X
D2
;
D
aj x j
j1
7
X
bj x j
j1
D3
:
D
Dj
D
8j 1; 2; 3;
where
k01
D : a1
b1
k02
a2
b2
k03
a3 ;
b3
and Dj is the determinant D in which the three entrices in the jth column are
replaced by
7
7
7
X
D2 X
D3 X
k0j xj ;
aj x j ;
bj x j ;
k0
D
D
j1
j1
j1
where xj j 4; . . . ; 7 denote the free parameters.
(ii) The rank of L is two: we consider B aA; a 2 R, and conclude from
Eq. (12) that
X
F
;
nS
G
:
nS
ST
:
nS
5. Theory of functions
As found by many authors both past and present the usual Cauchy
Riemann type dierentiability conditions, when extended to functions of a
hypercomplex variable, are too restrictive to yield interesting, if not useful,
85
functions. Our path is to follow one of the most successful approach in constructing functions, that of regular functions; a theory originally developed for
quaternions by Fueter [10], which was then rened, generalized, and extended
to the cases of a variable in other CayleyDickson algebras [1116] as well as in
Cliord algebras, and which are successfully used in physics by authors too
many to list them here. We dene the regularity (monogenicity) condition
for sedenions in terms of the variation of an integral of a function in a 16dimensional
Euclidean space. For a function, F(Z), of a sedenion variable,
P15
Z: l0 zl el ; zl 2 R, consider an integral over a closed hypersurface S15 ,
Z
15
dZ F Z:
J S :
S 15
15
X
ozl el
l0
15
X
l0
cl drel :
Using Stoke's theorem we can write J[S15 ] as a volume integral over V16 enclosed in S15 as
Z
DF ZdV ;
J S 15
V 16
where
D :
15
X
ol e l ;
l0
dV :
15
Y
dzl :
l0
to S15
in
Now consider an innitesimal variation of the surface S15
A
B
15
the neighbourhood of a point Z on SA . The volume enclosed by the two
surfaces is
Z
dV 16 2 R:
dV :
SA SB
Dene
dJ S
J SA15 J SB15
: lim
lim
dV !0
dV !0
dS
dV
R
dV
DF dV
dV
DFZ:
Then the function F(Z) is left regular at Z0 on S15 if J[S] is stationary under the
innitesimal variation of the surface at Z0 ,
dJ S
DFZ
0:
dS ZZ0
ZZ0
86
The denition is independent of the choice of S15 for DF(Z) 0. The right D
regularity as well as D-regularity
conditions can be dened in a similar manner.
We now outline the construction of some regular functions of a sedenion
variable.
(I) Intrinsic functions: Starting with an analytic function,
gz ux0 ; x ivx0 ; x;
of a complex variable, z x0 ix, and replacing i
P15
j1 xj ej
;
i)
x
p
1 by a pure sedenion
where
v
u 15
uX
x2j ;
x : t
j1
: DD
15
X
l0
o2l ;
is left-regular.
(II) Polynomial functions: Consider a function
"
#n
15
X
XT TXn
x0 ej xj tj ;
Fn X; T
2
j1
where
X :
15
X
xl el ;
l0
T :
15
X
tj ej 2 S:
j1
The power associativity ensures without ambiguity that Fn (X, T) can be written
as a power series in tk ,
X
n15
n!Pn1 n2 . . . n15 x0 ; x1 ; . . . ; x15 t1n1 t2n2 . . . t15
;
Fn X; T
P15
nr
where n :
j1
87
ExpiFX; T :
i n
F X; T:
n!
X
;
nX
Cn HX
;
nXn
Cn 1 2n n!
n 2 Z ;
n
Y
m 2k;
k1
: DD
15
X
o2
;
ox2l
l0
88
A : a0
7
X
aj ij ;
A : a0
15
X
j1
nA : AA AA
aj i j ;
j1
7
X
l0
a2l ;
A:1
: A=nA;
where the second structure constant, ejkmn , is antisymmetric in its indices, and is
determined from the seven distinct associator cycles, j; k; m; n by the usual 4indexed tensor rule: ejkmn (+1, 1, or 0), depending on the (even, odd, or
otherwise) permutation of the octonion associator cycle.
The following identities hold for octonion associators:
B; C B; A; C;
A; B; C A;
A; B; A; B; C A; BA; B; C A; B; CA; B;
A; B : AB BA;
A; AB; C A; B; CA:
A:2
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
R.D. Schafer, An Introduction to Nonassociative Algebras, Academic Press, New York, 1966.
L.E. Dickson, J. de Math. Pures et Appliq. 2 (1923) 281.
K. Carmody, Appl. Math. Comput. 84 (1997) 27.
K. Carmody, Appl. Math. Comput. 28 (1988) 47.
L. Sorgsepp, J. L~
ohmus, J. Hadronic 4 (1981) 327.
C. Muses, Appl. Math. Comput. 60 (1994) 25.
F. G
ursey, in: Y. Ne'eman (Ed.), To Fulll a Vision, Addison-Wesley, Reading, MA, 1981.
M. Imaeda, in: J.S.R. Chisholm, A.K. Common (Eds.), Cliord Algebras and their
Applications in Mathematical Physics, Reidel, Dordrecht, 1986.
K. Imaeda, M. Imaeda, Bull. Okayama Univ. Sci. 20A (1985) 133.
R. Fueter, Comm. Math. Helvetici. 7 (1935) 307.
K. Imaeda, H. Tachibana, M. Imaeda, S. Ohta, Nuovo Cim. 100B (1987) 53.
A. Sudbery, Math. Proc. Camb. Phil. Soc. 85 (1979) 199.
K. Imaeda, Nuovo Cim. 32 (1976) 138.
K. Imaeda, Memoirs of the Faculty of Liberal Arts and Education 8 (1957) 131.
K. Imaeda, Memoirs of the Faculty of Liberal Arts and Education 2 (1951) 111.
P. Dentoni, M. Sce, Rend. Sem. Mat. Univ. Padova 50 (1973) 251.