OCTONIONIC TERNARY GAUGE FIELD THEORIES REVISITED

CARLOS CASTRO Center for Theoretical Studies of Physical Systems Clark Atlanta University, Atlanta, Georgia. 30314, USA perelmanc@hotmail.com May 2013, Revised July 2013
Abstract An octonionic ternary gauge eld theory is explicitly constructed based on a ternary-bracket dened earlier by Yamazaki. The ternary innitesimal gauge transformations do obey the key closure relations [1 , 2 ] = 3 . An invariant action for the octonionic-valued gauge elds is displayed after solving the previous problems in formulating a non-associative octonionic ternary gauge eld theory. These octonionc ternary gauge eld theories constructed here deserve further investigation. In particular, to study their relation to Yang-Mills theories based on the G2 group which is the automorphism group of the Octonions and their relevance to Noncommutative and Nonassociative Geometry.

Keywords : Octonions; Ternary Algebras; Gauge Fields.

Introduction

Exceptional, Jordan, Division, Cliord, noncommutative and nonassociative algebras are deeply related and are essential tools in many aspects in Physics, see [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15] for references, among many others. A thorough discussion of the relevance of ternary and nonassociative structures in Physics has been provided in [16], [17], [18],[19],[20], [21]. The earliest example of nonassociative structures in Physics can be found in Einsteins special theory of relativity. Only colinear velocities are commutative and associative, but in general, the addition of non-colinear velocities is non-associative and non-commutative. Great activity was launched by the seminal works of Bagger, Lambert and Gustavsson (BLG) [22], [23], [24] who proposed a Chern-Simons type Lagrangian describing the world-volume theory of multiple M 2-branes. The original BLG theory requires the algebraic structures of generalized Lie 3-algebras and also 1

of nonassociative algebras. Later developments by [25] provided a 3D ChernSimons matter theory with N = 6 supersymmetry and with gauge groups U (N ) U (N ), SU (N ) SU (N ). The original construction of [25] did not require generalized Lie 3-algebras, but it was later realized that it could be understood as a special class of models based on Hermitian 3-algebras [26], [27]. A Nonassociative Gauge theory based on the Moufang S 7 loop product (not a Lie algebra) has been constructed by [28], [29]. Taking the algebra of octonions with a unit norm as the Moufang S 7 -loop, one reproduces a nonassociative octonionic gauge theory which is a generalization of the Maxwell and YangMills gauge theories based on Lie algebras. BP ST -like instantons solutions in D = 8 were also found. These solutions represented the physical degrees of freedom of the transverse 8-dimensions of superstring solitons in D = 10 preserving one and two of the 16 spacetime supersymmetries. Nonassociative deformations of Yang-Mills Gauge theories involving the left and right bimodules of the octonionic algebra were presented by [30]. Non-associative generalizations of supersymmetry have been proposed by [31] which is very relevant to hidden variables theory and alternative Quantum Mechanics. The ternary gauge theory developed in this work diers from the work by [22], [23], [24], [28], [29] in that our 3-Lie algebra-valued gauge eld strengths F are explicitly dened in terms of a 3-bracket [A , A , g] involving a 3-Lie algebravalued coupling g = g a ta . Whereas the denition of F by [22], [23], [24] was )a (A )c (A )a (A )c . based on the standard commutator of the matrices (A c c b b cd )cd and given in terms of = (A These matrices were dened as A = Aab fab the structure constants fabcd of the 3-Lie algebra [ta , tb , tc ] = fabcd td . In the next section we shall analyze Nonassociative Octonionic Ternary Gauge Field Theories based on a ternary octonionic product with the fundamental dierence, besides the nonassociativity, that the structure constants fabcd are no longer totally antisymmetric in their indices. Thus the bracket in the octonion case [[A, B ]] [A, B, g] is not eectively a Lie bracket (as it occurs in the 3-Lie algebra case) because the bracket [[A, B ]] in the octonion case does not obey the Jacobi identity since the structure constants fabcd are no longer totally antisymmetric in their indices. This work is quite an improvement of our prior results where we focused solely on the global rigid symmetries and homothecy transformations [32]. A gauge symmetry in the conventional sense has a spacetime dependent gauge parameter. Here the spacetime dependence of the gauge parameter enters through a scalar eld via a nonlinear symmetry transformation. Hence we must emphasize that one does not have a conventional gauge symmetry as such but it is inspired from gauge theories. 1 It is shown that the octonionic-valued eld a strength F = F ea transforms homogeneously (covariantly) under gauge transformations and that the Yang-Mills-like action is indeed invariant under local gauge transformations involving ternary octonionic brackets and antisymmetric gauge parameters ab (x) = ba (x), a, b = 0, 1, 2, 3, ....7. Furthermore, there is closure of these transformations based on antisymmetric parameters
1 We

thank one referee for raising this important point

ab = ba

Octonionic Ternary Gauge Field Theories

The nonassociative and noncommutative octonionic ternary gauge eld theory is based on a ternary-bracket structure involving the octonion algebra. The ternary bracket obeys the fundamental identity (generalized Jacobi identity) and was developed earlier by Yamazaki [33]. Given an octonion X it can be expanded in a basis (eo , em ) as X = xo eo + xm em , m, n, p = 1, 2, 3, .....7. (2.1)

where eo is the identity element. The Noncommutative and Nonassociative algebra of octonions is determined from the relations e2 o = eo , eo ei = ei eo = ei , ei ej = ij eo + cijk ek , i, j, k = 1, 2, 3, ....7. (2.2) where the fully antisymmetric structure constants cijk are taken to be 1 for the combinations (124), (235), (346), (457), (561), (672), (713). The octonion conjugate is dened by e o = eo , e i = ei = xo eo xk ek . X and the norm is
1 1 X) | 1 2 = | (xo xo + xk xk ) | 2 . (2.4) N (X) = | < X X > | 2 = | Real (X

(2.3)

The inverse X , X1 X = XX1 = 1. < XX > The non-vanishing associator is dened by X1 = (X, Y, Z) = (XY)Z X(YZ) In particular, the associator (ei , ej , ek ) = (ei ej )ek ei (ej ek ) = 2 dijkl el 1 mnp , i, j, k.... = 1, 2, 3, .....7 ijklmnp c 3! Yamazaki [33] dened the three-bracket as dijkl = (2.7) (2.5)

(2.6)

[ u, v, x ] Du,v x =

1 ( u(vx) v (ux) + (xv )u (xu)v + u(xv ) (ux)v ) . 2 (2.8) 3

After a straightforward calculation when the indices span the imaginary elements a, b, c, d = 1, 2, 3, ......, 7, and using the relationship [34] cabd cdcm = dabcm + ac bm bc am the ternary bracket becomes [ ea , eb , ec ] = fabcd ed = [ dabcd + 2 ac bd 2 bc ad ] ed whereas e0 has a vanishing ternary bracket [ ea , eb , e0 ] = [ ea , e0 , eb ] = [ e0 , ea , eb ] = 0 (2.9c) (2.9b) (2.9a)

It is important to emphasize that fabcd = cabd cdcm otherwise one would have been able to rewrite the ternary bracket in terms of ordinary 2-brackets 1 [ [ea , eb ], ec ] and this would have defeated the whole as follows [ea , eb , ec ] 4 purpose of studying ternary structures. The ternary bracket (2.8) obeys the fundamental identity [ [x, u, v ], y, z ] + [ x, [y, u, v ], z ] + [ x, y, [z, u, v ] ] = [ [x, y, z ], u, v ] (2.10) A bilinear positive symmetric product < u, v >=< v, u > is required such that that the ternary bracket/derivation obeys what is called the metric compatibility condition < [u, v, x], y > = < [u, v, y ], x > = < x, [u, v, y ] > Du,v < x, y > = 0 (2.11)

The symmetric product remains invariant under derivations. There is also the additional symmetry condition required by [33] < [u, v, x], y > = < [x, y, u], v > (2.12)

The ternary product provided by Yamazaki (2.8) obeys the key fundamental identity (2.10) and leads to the structure constants fabcd that are pairwise antisymmetric but are not totally antisymmetric in all of their indices : fabcd = fbacd = fabdc = fcdab ; however : fabcd = fcabd ; and fabcd = fdbca . The associator ternary operation for octonions (x, y, z ) = (xy )z x(yz ) does not obey the fundamental identity (2.10) as emphasized by [33]. For this reason we cannot use the associator to construct the 3-bracket. The physical motivation behind constructing an octonionic-valued eld strength in terms of ternary brackets is because the ordinary 2-bracket does not obey the Jacobi identity [ ei , [ ej , ek ] ] + [ ej , [ ek , ei ] ] + [ ek , [ ei , ej ] ] = 3 dijkl el = 0 (2.13) 4

If one has the ordinary Yang-Mills expression for the eld strength F = A A + [ A , A ] (2.14)

because the 2-bracket does not obey the Jacobi identity, one has an extra (spurious) term in the expression for [ D , D ] = [ F , ] + ( A , A , ) (2.15)

given by the crucial contribution of the non-vanishing associator (A , A , ) = (A A ) A (A ) = 0. For this reason, due to the non-vanishing condition (2.13), the ordinary Yang-Mills eld strength does not transform homogeneously under ordinary gauge transformations involving the parameters = a ea A = + [A , ] but it yields an extra contribution of the form F = [F , ] + ( , A , A ) (2.17) (2.16)

As a result of the additional contribution (, A , A ) in eq-(2.17 ), the ordinary Yang-Mills action S = < F F > will no longer be gauge invariant. Under innitesimal variations (2.17), the variation of the action is no longer zero but l receives spurious contributions of the form S = 4F i Aj Ak dijkl = 0 due to the non-associativity of the octonion algebra. For these reasons we focus our attention on ternary brackets. We dene the eld strength in terms of the ternary bracket as F = A A + [ A , A , g ] (2.18)

where g = g a ea is an octonionic-valued coupling function. One nds that under the naive innitesimal ternary gauge transformations
d a b c d a b c (Ad ed ) = ( ed ) + [ ea , A eb , g ec ] (F ed ) = [ (x)ea , F eb , g ec ] (2.19) the ordinary quadratic action

S =

1 42

dD x < F F >

(2.20)

is not invariant under ternary innitesimal gauge transformations as we shall see next. is a suitable dimensionful constant introduced to render the action dia mensionless. The octonionic valued eld strength is F = F ea , and has real 0 i valued components F , F ; i = 1, 2, 3, ....., 7. The < > operation extracting ) =< Y X >= Real(Y X ). Under the e0 part is dened as < XY >= Real(XY innitesimal ternary gauge transformations of the ordinary quadratic action one has S = 1 42 dD x < F (F ) + (F ) F > = 5

1 42 1 42 1 42

c dD x < F ec [ a ea , F

eb , g n en ] > +
c

b dD x < [a ea , F eb , g n en ] F b

ec > =

c dD x a F F

( < ec fabnk ek > + < fabnk ek ec > ) =

c dD x a g n F F b

1 2 dD x

1 22

fabnc = = 0 (2.21)

b c (a ga ) (F Fb ) (a Fa ) (gc F )

Hence, because fabnc = ( dabnc + 2 an bc 2 bn ac ) (2.22)

is not antisymmetric under the exchange of indices b c : fabnc = facnb , the variation in eq-(3.15) is not zero. Had fabnc been fully antisymmetric then the c F b is symmetric variation S would have been zero due to the fact that F under b c. Concluding, in the octonionic ternary algebra case, the naive transformations (2.19) do not leave the action (2.20) invariant : S = 0. One alternative would be to nd counter terms, if possible, to the action (2.20) S + S so that (S + S ) = 0. The authors [30] used counter terms of the form F AA+AAAA in the non-associative deformations of ordinary Yang-Mills theories based on the left and right actions by octonions and ordinary brackets. Unfortunately it does not work in our case and for this reason we shall follow a dierent approach. Another problem due to the fact that fabcd is not totally antisymmetric in all of its indices is that there is no closure of the innitesimal octonionic ternary gauge transformations A = + [, A , g]. Furthermore, the bracket in the octonion case [[A, B ]] [A, B, g] is not eectively a Lie bracket (as it is in the 3-Lie algebra case) since the bracket [[A, B ]] in the octonion case does not obey the Jacobi identity because the structure constants fabcd are no longer totally antisymmetric in their indices. As said previously, because the associator ternary operation for octonions (x, y, z ) = (xy )z x(yz ) does not obey the fundamental identity (3.10) one cannot use the associator to construct the 3-bracket and this rules out the use of the totally antisymmetric dabcd . Nevertheless, as we shall show below, the quadratic Yang-Mills-like action (2.20) is invariant under the local octonionic ternary gauge transformations dened by
ab c ( Ad ed ) = (x) [ ea , eb , A ec ]

(2.23a) (2.23b)

and (g d ed ) = ab (x) [ ea , eb , g c ec ] where one introduces a local spacetime depedence on the antisymmetric gauge parameters ab (x) = ba (x). One may notice now that the coupling g c ec is 6

not inert under the transformations (2.23b). Only the real part of the coupling g 0 is inert. After some straightforward algebra one can verify that the ternary eld strength F dened in terms of the 3-brackets transforms properly (homogeneously) under the local transformations (2.23)
m c c m c (F em ) = ab [ ea , eb , F ec ] = ab F fabcm em F = ab F fabcm (2.24) if the following conditions are satised ij k l [ ( ij ) Ak ( ) A ] fijk = 0

(2.25)

Due to the key presence of the imaginary parts of the couplings g i we shall i prove below that one can partially gauge the elds by setting Ai = g (x) (x) (in terms of an auxiliary scalar eld (x)) and still leave room for residual symmetries. One should note that even if the coupling functions g i are chosen to be constants g i = constant, it must be kept in mind that after gauge transformations the new couplings g i will acquire a spacetime dependence via the x -dependence of the ab (x ) parameters. For this reason we should not set a priori the couplings to constants. Furthermore, the eld strength will not become trivially zero. It can be rewritten, when the gauge elds are partially l gauged as Al = g (x) (x), in the following way
l i j k l l l F = [ Al ] + A A g fijk = ( g ) ( ) ( g ) ( ) = 0 (2.26)

after using the conditions (g i )(g j )fijkl = 0 due to the antisymmetry of fijkl = fjikl and the symmetry g i g j = g j g i . Therefore, due to the xdependence of the imaginary parts of the couplings g i (x), the eld strength l are not zero and their contribution to the action (2.20) is not components F trivially zero. After this detour let us look for nontrivial solutions to (2.25). ij = constant are the trivial solutions leading to global rigid symmetries. A partial gauge xing k k k k provided by Ak = g , A = g yields in eq-(2.25), after factoring g and l = f l ij , the following equation dening k ijk l l ( k ) ( ) ( k ) ( ) = 0 eq-(2.27) can also be rewritten as l l ( k ) ( k ) = 0 a solution to eq-(2.27) can be easily obtained when l l l l l l (x) , = 0 (2.29a) k = k , k = k k (x) = Ck e l = f l ij one can infer that is an arbitrary nonzero constant. Since k ijk
l k ij (x) = C ij e(x) , C = fijkl C ij , = 0

(2.27)

(2.28)

(2.29b)

 l and C ij are the constant entries of the arbitrary constant 7 7 where C k , C when the range of indices is l, k = 1, 2, 3, ......, 7. Inserting the matrices C solutions (2.29a) back into eq-(2.27) one immediately can verify that l (2.29c) k ( ) = 0 l ( x) = C l e(x) is indeed a correct and nontrivial therefore, the expression k k solution to eq-(2.27). When = 0 one recovers in eqs-(2.29a, 2.29b) the trivial l = constants, ij = constants corresponding to the global (rigid) solutions k transformations. To sum up, after partially gauging the eld
k Ak (x) = g (x)

(2.30)

the gauge parameters which solve eq-(2.25) are given by ij (x) = C ij e(x) and, consequently, one has found nontrivial solutions for the gauge parameters. From eq-(2.30) one can infer the explicit relation between and Ak given by the following line integral
x

o =
0

gk (x )Ak (x ) dx g 2 (x )

(2.31a)

The exponential of the constant o (x = 0) can be reabsorbed into the denition of the entries of the constant matrix so that the eld dependence of the parameters becomes

x )

ij (x ) = C ij e(x

= C ij exp

gk (x )Ak (x ) dx g 2 (x )

(2.31b)

and the gauge transformations (2.23) are now nonlinear in the elds. Hence, we nally have arrived at eqs-(2.30, 2.31b) which are the nontrivial solutions to eq-(2.25) when = 0. It is interesting that (2.31b) is given by an expression which resembles the Wilson loop form with the main dierence that one has a line integral instead of a loop. The value of the line integral is path-independent and solely depends on the initial point (the origin of coordinates) and the nal point x . k To conclude, when one partially gauges the elds as Ak = g , there is still a residual symmetry that remains such that the gauge transformations k (2.23) become nonlinear in the elds and the nonvanishing eld strength F given by eq-(2.26) transforms homogeneously (2.24) under the local gauge transformations (2.23) due to the vanishing conditions imposed by eq-(2.25). The 0 component associated with the unit element of the octonion algebra F is inert 0 F = 0 under the gauge transformations. k The reason the eld strength F transforms homogeneously (2.24) when the inhomogeneous terms (2.25) vanish is a direct consequence of the fundamental identity (2.10) because the 3-bracket (2.8) is dened as a derivation [ [ea , eb , A ], A , g ] + [ A , [ea , eb , A ], g ] + [ A , A , [ea , eb , g] ] = 8

[ ea , eb , [A , A , g] ]

(2.32)

Another important condition due to the antisymmetry fijkl = fjikl = fijlk , and symmetry g l g k = g k g l , is the invariance of g 2 = gl g l under gauge transformations g 2 = (gl g l ) = 2 gl g l = 2 gl ij g k fijkl = 2 g l g k ij fijkl = 0 (2.33) Therefore, the full octonionic norm-squared (g o )2 + gi g i of the octonionic-valued coupling function g = g o eo + g i ei is invariant under gauge transformations (2.23b). An important remark is in order. There is a plausible caveat about the conditions (2.25). One must ensure that such conditions, which do not appear to be explicitly gauge covariant, will not break the gauge covariance (invariance) of the theory one is trying to construct. In particular, after performing a gauge variation of the conditions C[l ] = 0 in (2.25) one would introduce the secondary conditions C[l ] = 0. Performing yet another gauge variation of the secondary conditions ( C[l ] ) = 0 .... , and so forth, one obtains a hierarchy of equations to be satised by the gauge parameters ij (x). It is clear that the trivial solutions ij = C ij = constants will satisfy automatically all the equations suggesting, perhaps, that octonionic ternary eld theories cannot be gauged. Nevertheless, as we shall show next, there is a very natural way to bypass this problem such that the gauge variation of the conditions (2.25) remains zero without introducing additional constraints on the parameters ij that might have forced them to be constants. From the gauge variations
ij k l ij k l l Al = (x) A fijk = (x) g fijk = (g ) =

(g l ) + g l () = ij (x) g k fijkl + g l () = 0 (2.34) one learns that = 0, and in turn, we can infer from (2.31b) that ij = 0 so i ij that the variation of the condition (2.25) (when Ai = = 0) = g , remains zero without introducing further constraints on the parameters. A variation of (2.25) gives then fijkl ( ij ) (g k ) ( ) ( ij ) (g k ) ( ) = 0 (2.35)

therefore, after simple factorization of g k in (2.35) it leads to the exact same equation (2.27) at the beginning obtained from a factorization of g k in (2.25) and which admits the solutions described above. Hence, the variations of (2.25) do not impose additional constraints on the parameters ij which might have forced us to have the trivial constant solutions for these gauge parameters. a Finally, given the octonionic valued eld strength F = F ea , with 0 i real valued components F , F ; i = 1, 2, 3, ....., 7, one can verify that the quadratic action (2.20) is indeed invariant under the ternary innitesimal local

gauge transformations (2.23) when the eld strength transforms as provided by eq-(2.24) S = 1 42 1 42 1 42 1 42 dD x < F (F ) + (F ) F > =
c dD x < F ec ab [ea , eb , F c dD x < ab [ea , eb , F ec ] F n n

en ] > + en > =

c dD x ab F F

( < ec fabnk ek > + < fabck ek en > ) = 0. (2.36)

this is a direct result of < ec fabnk ek > + < fabck ek en > = (fabnk ck + fabck kn ) = (fabnc + fabcn ) = 0 (2.37) due to the property fabnc + fabcn = 0 which can be explicitly veried as follows [ dabnc + 2 an bc 2 bn ac ] + [ dabcn + 2 ac bn 2 bc an ] = 0 (2.38) because dabnc + dabcn = 0; dnabc + dcabn = 0, due to the total antisymmetry of the associator structure constant dnabc under the exchange of any pair k 2 ) = of indices. A shortcut to prove the invariance S = 0 is simply (F ij lk lk kl 2 fij Fl Fk = 0 due to the antisymmetry fij = fij and symmetry Fl Fk = Fk Fl under the exchange of indices k l. The variation of 0 2 ) =0 the components associated with the e0 generator is trivially zero (F because [ei , ej , e0 ] = 0. The Bianchi identities are also satised. This work is not complete until we show the closure of the innitesimal transformations (2.23). To achieve this one needs rst to recast them as derivations
ab c ab 1 A = 1 ( Ak ek ) = 1 [ ea , eb , A ec ] = 1 Dea ,eb A cd l cd 2 A = 2 ( Ak ek ) = 2 [ ec , ed , A el ] = 2 Dec ,ed A

(2.39a) (2.39b)

by recurring to the fundamental identity (2.10) in order to evaluate the commutator of two derivations, and after relabeling indices, one arrives at
ab [1 , 2 ] A = cd 2 1 [ Dec ,ed , Dea ,eb ] A = ab cd 2 1

D[ec ,ed ,ea ],eb + Dea ,[ec ,ed ,eb ]

A =

ab cd 2 1 ( [ [ec , ed , ea ], eb , A ] + [ ea , [ec , ed , eb ], A ] ) = ab cd ab cd 2 1 fcdak [ ek , eb , A ] + 2 1 fcdbk [ ea , ek , A ] =

10

ak b cd ab k kb ( cd 2 1 fcda 2 1 fcda ) [ ek , eb , A ] = 3 [ ek , eb , A ] = 3 A (2.40) Therefore the antisymmetric parameter resulting from the closure of two transformations is given by ak b cd ab k kb = ( cd 3 2 1 fcda 2 1 fcda )

(2.41)

Inserting the solutions given by eq-(2.29b) into (2.41) give

cd ak cd ab kb 3 kb = ( C2 C1 fcdab C2 C1 fcdak ) e(1 +2 ) = C3 e (2.42) 3

where 3 = 1 + 2 and the constant entries of the antisymmetric C3 matrix are given in terms of the constant entries of the antisymmetric matrices C1 , C2 as follows
kb cd ak cd ab b ak 2 k ab C3 = ( C2 C1 fcdab C2 C1 fcdak ) = ( C a C1 C2a C1 ) (2.43) kb 3 Therefore, because the solution for kb given by (2.42) has the same 3 = C3 e functional form as that required by eq-(2.29b) in order to solve eq-(2.25), there is closure of two gauge transformations as shown in the relations (2.42, 2.43) ij ij (3 = 1 + 2 ) among the parameters ij 1 , 2 and 3 . One may write such relation among the parameters symbolically as 3 = 1 2 such that [1 , 2 ]A = 3 A . The nite ternary transformations can be obtained by exponentiation as follows

= F + F +

1 1 (F ) + ( ( (F ))) + .... 2! 3!

(2.44)

c c m ec ] ]; ec ]; (F ) = mn [ em , en , ab [ea , eb , F em ) = ab [ea , eb , F where (F ...... To show that the action is invariant under nite ternary local transformations requires to follow a few steps. Firstly, one denes

< x y > Real [ x y] =

1 (x y+y x ) < x y > = < y x > (2.45) 2

Despite nonassociativity, the very special conditions x( xu) = (xx )u; x(ux ) = (xu) x; x(xu) = (xx)u; x(ux) = (xu)x (2.46) are obeyed for octonions resulting from the Moufang identities. Despite that (xy )z = x(yz ) one has that their real parts obey Real [ (x y ) z ] = Real [x (y z ) ] (2.47)

Due to the nonassociativity of the algebra, in general one has that (U F )U 1 = U U = UU = 1, as a result of the U (F U 1 ). However, if and only if U 1 = U the very special conditions (2.46) one has that F = (U F )U 1 = U (F U 1 ) =

11

 is unambiguously dened. One can equate the result of the U F U 1 = U F U exponentiation procedure in eq-(2.44) to the expression = ek (ab )ek (F c tc ) ek (ab )ek ; = U F U 1 = U F U

k = 1, 2, 3, ...., 7. (2.48) where k (ab )ek is a complicated function of ab . It yields the nite transformations which agree with the innitesimal ternary ones when ab are inf initesimals. For instance, to lowest order in ab , one has that k satises 2k ckcd = ab fabcd and which follows by comparing the transformations in (2.44) to those in (2.48), to lowest order. Dropping the spacetime indices for convenience in the expressions for F , F , , the action denand by repeated use of eqs-(2.45, 2.46, 2.47), when U 1 = U sity is also invariant under (unambiguously dened) transformations of the form , F = U F U 1 = U F U F ] = Re [(U F U 1 ) (U F U 1 )] = Re [(U F ) ( U 1 (U F U 1 ) )] = < F F > = Re [F ) (U 1 U ) (F U 1 )] = Re [(U F ) (F U 1 )] = Re [(F U 1 ) (U F )] = Re [(U F ) )] = Re [F (U 1 U ) F ] = Re [F F ] = Re [F F] = < F F > . Re [F ( U 1 (U F (2.49) The real part of the coupling g 0 is inert under the transformations (2.23b) and it decouples from the denition of the eld strength F because e0 has a vanishing 3-bracket with other elements of the octonion algebra. The coupling g 0 = constant can be incorporated into the eld strength in the same fashion as it occurs in ordinary Yang-Mills. One may rewrite the physical coupling g 0 as a prefactor in front of the 3-bracket as F = A A + g 0 [A , A , g], and reabsorb g 0 into the denition of the A eld as F = 1 1 0 0 0 0 g 0 (g A ) (g A ) + [g A , g A , g] . Thus F g 0 F and the ac1 tion is rescaled as S (g0 )2 S as it is customary in the Yang-Mills action. i 2 To conclude this work, when Ai = g we have an action 1 4 2
0 i dD x ( F F0 F Fi ), i = 1, 2, 3, ....., 7

S =

(2.50)

the (g 0 )2 coupling squared can be reabsorbed into the denition of 2 . The kinetic terms are explicitly given as
i i j k l 2 i i 2 F Fi = [ [ Al ] + A A g fijk ] = [ ( g ) ( ) ( g ) ( ) ] =

( g i )2 ( )2 ( g i ) ( ) ( gi ) ( )
0 F F0

= 0 A 0)

(2.51) (2.52)

( A0

A0 )

A 0

F > for the action so that the F i F terms also appear with can choose < F i a positive sign inside the integrand. 2 One

12

and which are invariant under the transformations (2.23a, 2.23b) due to eqs(2.27) and = 0 (2.34). The solutions to eqs-(2.25, 2.27) are provided by eqs-(2.30, 2.31b). It would have been desirable to avoid the conditions (2.25, 2.27) which force the gauge parameters ij (x) to be eld-dependent in the sense that they are given by eq-(2.31b) and which has a Wilson-loop-like expression. There is closure of two gauge transformations as indicated by eqs-(2.40-2.43). After lengthy but straightforward algebra one can also verify that the Bianchi i i i identities D[ F ] = 0 are obeyed when A = g and DF = dF + [A, F, g ]. The nontrivial role of the 7 + 1 scalars in the action (2.51) given by the coupling functions g i (x) and (x) warrants to be studied further. We should i emphasize that the kinetic terms F Fi for the latter scalars do not appear in the action in the usual canonical form (D i )2 . One reason being that j k l D l = l + Ai g fijk does not transform homogeneously under gauge transformations unless the gauge parameters are trivially constant. However, when the scalars g i (x), (x) appear in the action in the form displayed by eqsl k l i fijk as they = ij F terms transform homogeneously F (2.50, 2.51), the F 0 should, whereas F = 0. The terms (2.52) have the same functional form as the Maxwell action but 0 the transformations laws for A0 ( A = 0) dier from the U (1) gauge eld ( A = ). The inclusion of potential terms for the scalar elds and ChernSimons actions will be the subject of future investigation. These octonionc ternary gauge eld theories deserve further investigation. In particular, to study their relation to Yang-Mills theories based on the G2 group which is the automorphism group of the Octonions and their relevance to Noncommutative and Nonassociative Geometry [35]. Acknowledgments We thank M. Bowers for her assistance and to one of the referees for pointing an error and raising important points to improve this work.

References
[1] P. Jordan, J von Neumann and E. Wigner, Ann. Math 35 (1934), 2964. [2] K. MacCrimmon, A Taste of Jordan Algebras (Springer Verlag, New York 2003). [3] H. Freudenthal, Nederl. Akad. Wetensch. Proc. Ser 57 A (1954), 218. [4] J. Tits, Nederl. Akad. Wetensch. Proc. Ser 65 A (1962), 530. [5] T. Springer, Nederl. Akad. Wetensch. Proc. Ser 65 A (1962), 259. [6] J. Adams, Lectures on Exceptional Lie Groups, (Chicago Lectures in Mathematics, Univ. of Chicago Press, Chicago 1996).

13

[7] R. Schafer, An introduction to Nonassociative Algebras (Academic Press 1966). [8] C. H Tze and F. Gursey, On the role of Divison, Jordan and Related Algebras in Particle Physics (World Scientic 1996). [9] S. Okubo, Introduction to Octonion and other Nonassociative Algebras in Physics (Cambridge Univ. Press, 2005). [10] T. Springer and F. Veldkamp, Octonions, Jordan Algebras and Exceptional Groups (Springer Verlag 2000) [11] G. Dixon, Division Algebras, Octonions, Quaternions, Complex Numbers, and the Algebraic Design of Physics (Kluwer, Dordrecht, 1994). [12] G. Dixon, J. Math. Phys 45 (2004), 3678. [13] J. Baez, Bull. Amer. Math. Soc. 39 (2002), 145. [14] G. Dixon, Division algebras, Spinors, Idempotents, the Algebraic structure of Reality, arXiv : 1012.1304. G. Dixon, Octonions, the E8 Lattice to 16 , arXiv : hep-th/9501007. [15] L. Boya, Octonions and M-theory (Talk Presented at the the 24th International Colloquium on Group-Theoretical Methods in Physics. Paris, July 15-22, 2002) hep-th/0301037. [16] Y. Nambu, Phys. D 7 (1973), 2405. [17] R. Kerner, Ternary Algebraic Structures and their applications in Physics, math-ph/0011023. [18] R. Kerner, Communications in Math. Phys. 91 (1983), 213. [19] R. Kerner, Classical and Quantum Gravity 14 (1997), 203. [20] M. Dubois-Violette, R. Kerner and J. Madore, Jour Math Phys 31 (1990), 316. [21] M. Dubois-Violette and M.Henneaux, Comm. in Math. Phys. 226 (2002), 393. [22] J. Bagger and N. Lambert, Phys. Rev. D 75 (2007), 045020. [23] J. Bagger and N. Lambert, Phys. Rev D 77(2008), 065008. [24] A. Gustavsson, JHEP 04 (2008), 083. Nuc. Phys B 811 (2009), 66. [25] D. Jaeris, O. Aharony, O. Bergman and J. Maldacena, JHEP 10 (2008), 091.

14

[26] J. Bagger and N. Lambert, Three-Algebras and N=6 Chern-Simons Gauge Theories, arXiv: 0807.0163 [hep-th]. [27] P. de Medeiros, J. Figueroa-O Farrill, E. Mendez-Escobar and P. Ritter, On the Lie-algebraic origin of metric 3-algebras, arXiv:0809.1086 [hep-th]. [28] T. Ootsuka, E. Tanaka and E. Loginov, Nonassociative Gauge Theory, hepth/0512349. [29] E. Loginov, Nucl. Phys B 606 (2001), 636. [30] R. Foot and G. Joshi, Phys. Rev D 36 (1987), 1169. Int. J. of Theor. Phys 28 (1989), 263. [31] V. Dzhunushaliev, Nonassociative generalization of Supersymmetry, arXiv.org : 1302.0346. V. Dzhunushaliev, J. Gen. Lie. Theory and Applications vol 2, no. 4 (2008), 269. V. Dzhunushaliev, J. Gen. Lie. Theory and Applications vol 3, no. 1 (2009), 33. [32] C. Castro, Int. J. Mod. Phys A 26, no. 18 (2011), 2997. [33] M. Yamazaki, Octonions, G2 and generalized Lie 3-algebras, Phys. Letts B 670 (2008), 215. [34] L. Borsten, D. Dahanayake, M. J. Du, H. Ebrahim and W. Rubens, Black Holes, Qubits and Octonions, arXiv : 0809.4685. [35] R. Wulkenhaar, The Standard Model within Non-Associative Geometry, hep-th/9607096. Grand Unication in Non-Associative Geometry, hepth/9607237. S. Farnsworth and L. Boyle, Non-Associative Geometry and the Spectral Action Principle, arXiv : 1303.1782 [hep-th].

15