Foundations o f Physics, Vol. 12, No.

7, 1982

Pauli-Dirac Matrix Generators

of Clifford Algebras
Charles P. Poole, Jr. 2 and Horaeio A. Farach z

Received August 24, 1981

This article presents a Pauli-Dirac matrix approach to Clifford Algebras, it is

shown that the algebra C 2 is generated by two Pauli matrices ion. and ia3; C3 is
generated by the three Pauli matrices a~, 62, a 3 ; C 4 is generated by four Dirae
matrices Yo, Yl, ~2, Y3 and C~ is generated by five Dirac matrices i7o, iy~, i)'2,
i~'3, i75. The higher dimensional anticommuting matrices which generate
arbitrarily high order Clifford algebras are given in closed form. The results
obtained with this Clifford algebra approach are compared with the vector
product method which was described in a recent article /Found. Phys. 10,
531-553 (1980) by Poole, Farach and Aharonov] and with the Dirac,
Rashevskii and Ramakrishnan methods of matrix generation.

1. I N T R O D U C T I O N

In three previous articles ~-3) denoted by I, II, and III, we presented a vector
product treatment of space rotations, Lorentz transformations and special
relativity. At the conclusion of the third article, we pointed out that the
vector product method m a y be considered as a particular application of
Clifford algebras. ~4 7) This article will discuss a Pauli-Dirac matrix approach
to Clifford algebras, and we will show how this is related to the vector
products employed in I, II, and III.
Early pioneer work on the theory and physical applications of Clifford
algebras was carried out by Riesz ~8-1°) and continued by Corson, (11~
Rashevskii/12) Teitler,~ 13,14) Ramakrishnan, (~5~ Hestenes, (16,t7) Sobczyk, ( ~
Salingaros, (19) and Dresden. ~z°) They emphasized the properties of the Dirac

Supported by the National Science Foundation under Grant ISP-80-1145 I.

2 Physics and Astronomy Department, University of South Carolina, Columbia, South
Carolina.
719
00t5-9018/82/0700-0719503.00/0© 1982PlenumPublishingCorporation
720 Poole and Farach

algebras in Lorentz space-time and applications of it, for example, to

classical field theory, electromagnetism and particle symmetries. Much of
this work is related to the material presented in III from a vector product
approach. Recently Salingaros and Dresden (2°) introduced a new vector
product method called vee multiplication. A number of articles have
appeared on applications of Clifford algebras to Dirac theory, (2~-31)
Maxwell's equations, (13'3z-35) spinors,(29,36 52) elementary particles, (53-58)
curved spaces, ~s9-61) transformations ~6z-64~ and other topics. (6s-74) Clifford
algebras are related to other types of algebras that have been recently applied
in physics. ~3°'48"s6'75-83)

2. CLIFFORD ALGEBRAS

A Clifford algebra (4-7) is an algebra of anticommuting operators. An n-

th order Clifford algebra is characterized by n anticommuting matrices uj
which generate the algebra and constitute its basis. This set of n matrices
forms what is called the vector (V) term of the algebra. All products uiu J,
u i u j u k , u i u j u k u t .... where i 4= j 4= k 4= l,... of these matrices generate a linear
vector space of dimension 2 n. One term of the vector space called the scalar
(S) is associated with the unit matrix I. Each pair of matrices of the basis
satisfies the anticommutation relation

uiu~ + u j u i = i 2fiij (1)

As we will show later in the article the positive sign applies for n = 0, 3, 4, 7,
8, 11, 12 ..... and the negative sign is used for n = I, 2, 5, 6, 9, 10..... The
former is said to have a positive signum and the latter a negative signum.
The notation u~ will ordinarily be employed for the case of a negative
signum, and u i for a positive signum. For even order algebras, the anticom-
muting matrices are (21/2") × (21/2n) in size, and for odd order algebras, they
are (21/2("- 1))× (21/2(,-1)) matrices. For example, when n = 5, the matrices
are 4 X 4 in size. The present article will assume a metric with elements gij
given by
gij =fiii (2)
as implied by Eq. (I).
In the next section we will show how the various orders C, of Clifford
algebras are generated by sets of matrices. More specifically, the C o and Cl
algebras are generated by one dimensional matrices 1 and i, the C 2 and C 3
algebras are generated by 2 × 2 Pauli spin matrices ~i, the C~, and C5
algebras are generated by 4 × 4 Dirac matrices yj and the C 6 and Cv
algebras are generated by 8 × 8 generalizations of Dirac matrices called 7ri
Pauli-Dirac Matrix Generators of Clifford Algebras 721

Table L Structure of the Clifford Algebra C, of Various Orders n in Terms of Their

Component Scalars (S), Vectors (V) and Tensors (T, U, W,...) and Their Pseudo Counterparts
(Sp, Vp, Tp, L~, W~,...). The Dimensionality and Signum is Given for Each Term.

Co S(I+)
C1 S(1+) Sp(1 )
C2 S(l+) V(2 ) Sp(l-)
C3 S.+~ V~3+~ V~,~3_~ Sp~ ~
C4 S.+~ V~4+~ T¢6_~ ~4-~ S~¢~+~
C5 S¢1+) V<s_) T~,o_~ Tp¢io+~ V~¢~+:, Sp~1
C6 S(t+) V~6-) T(~5-) U~2o+) Tp{~5+) Vp(6-) Sp(l-)
C7 S(I+) V(7+~ T(21 ) U(35 ) Up(35+) Tp(/I-~) Vp(7 I Sp(I-}

C9 S(I+) V(9-) T(36-) U(84+) W(l;t6+) Wp(I26 ) Up(s4-) Tp(36+) Vp(9+) Sp~l-5

matrices. Each algebra has a corresponding unit matrix which is called a

scalar S, and in all orders for which n >/2 the set of n generating matrices is
called the vector V. Other terms in the algebras are pseudoscalars of dimen-
sionality 1, pseudovectors which have the same dimensionality, n, as the
corresponding vectors, and tensors and pseudotensors, which have dimen-
sionality n ( n - 1)/2. The scalars exist for all algebras, i.e., for n ~> 0,
pseudoscalars S o are present for n .>i 1, the vectors V are present for n ) 2,
the pseudovectors VR are present for n ~> 3, the tensors T are present for
n ) 4 , and the pseudotensors Tp exist for n/> 5, etc. The overall scheme
corresponding to these terms is illustrated in Table I for the Clifford algebras
up to n = 9. For each term we give the dimensionality and the signum or
eveness (+) or oddness ( - ) of the corresponding matrix product MjMj with
itself. In other words:
MjM) = +l (3)

where I is the unit matrix and j is not summed in the notation of this article.
These various properties will become clearer as we describe the generation of
the various individual algebras. A number of articles have appeared which
describe applications of particular Clifford algebras, such a s C2, (76'77'84)
C3 (19,30,47,48,85) C4(12,14,19,20,22,28,29,50,53,54) C5 (60,61) C7 (27,55, 86) C8 ,(51)
CN+2, (52) C2N, (78} and C2N+I. (sT)

3. L O W E S T O R D E R (0 ~< n ~< 3) C L I F F O R D A L G E B R A S

In this section we will describe the two trivial Clifford algebras C o and
C~, and then generate C2 and C3 using Pauli matrices.
722 Poole and Farach

Co: The identity algebra C o has one element, namely the scalar 1, and
there is no generating vector.
C~ : The complex number algebra C 1 has 2 ~ = 2 elements, namely the
scalar identity 1, and the term i = x/Z] - which is listed in Table I as a
p s e u d o s c a l a r S p ( l _ ) . The signum of Sp¢~_) is negative since i2=--1.
C 2 : The real quaternion algebra Cz which is of dimensionality 2 2 = 4
is generated by two anticommuting 2 × 2 matrices which constitute its basis.
These matrices may be selected as the Pauli matrices a 2 and CY3multiplied by
i which we denote by the symbols CY; and CY;, respectively.

0"2 (01 ;)
a; : iCY3 : 0 i (5)

For this case (a;) 2 = --I where I is given by

/(lo (6)

and the remaining element of the algebra is the matrix product a;a'3

0"203
' ' : --t71' = - 0 (7)

which is the remaining Pauli matrix a I multiplied by i. The matrix a'l

constitutes the pseudoscalar Sp(~_) listed in Table I. Thus we see that only
two of the three anticommuting Pauli matrices are used to generate this
algebra, the third being formed as a product which is the pseudoscalar.
C3: The complex quaternion algebra C 3 which we will call the Pauli
algebra has a total dimensionality of 2 3 = 8. It is generated by the three
Pauli matrices

which together constitute the vector V(3+) of the algebra. The signum is
positive since a~ = I. The three binary matrix products aia i

aloj=o~ (9)

which form the three dimensional pseudovector Vp(3 ~ are the matrices a '1,
a;, and a; that appeared in the next lower algebra C2. Thus the vector V{3+7
Pauli-Dirae Matrix Generators of Clifford Algebras 723

Table II. Matrix Representations of the Lowest Order Clifford Algebras in Terms of the
Unit Matrix I, the Pauli Matrices aj, the Dirac Matrices Y1and Dirac Matrix Products a f, fli,
and ~j. The prime Denotes Multiplying by i = ~ .

Co:

Ct: 1 i

C2: I o~- (7 ,I
G;

C~: I g2 el ff
G3 G~

a;

C4: I Y2 a3 3~

y4 G 5;

a~ al
a~ a2
0~ a3

y~ 5'1 6I Y4

5; G
5~ G

of C 3 is formed from the terms i V ( z _ ) and iSplt_ ) of C 2. The pseudo-scalar

term Sp(~_) equals i times the unit matrix which is denoted by I ' and arises
from the triple matrix product.

a 1 % o 3 = iI = i ' (10)

of all of the elements of the vector term V(3+). Since this algebra C 3 exhausts
all products of Pauli type matrices we conclude that higher order Clifford
algebras require higher order matrices as generators.
The results obtained at in this section are summarized in Table II which
lists the elements of all of the scalars, vectors, pseudovectors, and
pseudoscalars of these low order algebras in an array arranged in the manner
of Table I.
724 Poole and Faraeh

4. D I R A C A L G E B R A S (n = '1, 5)

In this section we will generate the next two higher order Clifford
algebras, namely C 4 and C 5, from the Dirac matrices.
C 4 : The Dirac algebra C 4, which is perhaps more properly referred to
as the real Dirac algebra, has a total dimensionality 24 = 16 and is generated
by the four anticornmuting 4 × 4 Dirac matrices which we choose as follows
consistent with the metric gj~ = 6ik

(0 °:1
7j = --a)OS , j = i, 2, 3 74 ~-" ('0 °,1
The matrix 7j is a direct product of the 2 X 2 matrix (_°~o) with the Pauli
matrices.
Here we adopt the notation 74 for what is usually referred to as 70. This
unconventional notation for 74 makes it easier to write down the expressions
for higher order Clifford algebras. These four 7,i constitute the v e c t o r V(4 ÷)
which generates the algebra C 4. The signum is positive since

72 = +I j = l, 2, 3, 4 (12)

The six binary matrix products 7jYk form the six components of the tensor
T(6_ ). We will adopt the notation

(o°
= (~ O) i,j,k=l,2,3, cyclically (14)

where the definition of a i is standard in the literature and was used originally
by Dirac. (87'88) Thus the tensor T(6_ ) has three a~ and three fl[ as its
elements. The three triple products Y4YjTk where j ~ : k = 1, 2, 3, plus 7172Y3
form the pseudovector Vp(4_ ) and we write them as follows

(15)

where the symbol 6 i is adopted from Arfken (89) and the elements + I of
Eq. (16) are unit 2 × 2 matrices. Therefore, the pseudovector Vp(4_ ) has the
PaulbDirae Matrix Generators of Clifford Algebras 725

four elements 6'1,6;, c~, 0Z. Finally the pseudoscalar Vp~t+~ is the quadruple
product

~5 = ~)lY2~3Y4= ( 0 / ) (17)

and it has a positive signum. Thus we see that four of the five anticommuting
Dirac matrices generate the C 4 algebra, the fifth one forming the
pseudoscalar.
C5: The full Dirac Algebra C 5 which can also be referred to as the
complex Dirac algebra is generated by selecting the five Dirac matrices yj
defined by

yj = ,v/--1 yj, j= 1,2,3,4,5 (18)

as the elements of the vector term V{5_). The binary products Yl Y~ generate
the ten component tensor To0_ I with the elements a~, with k = 1, 2, 3, il~. with
k = 1, 2, 3, and ~ with k = 1, 2, 3, 4. The ternary products y~ y~y~ form the
ten component pseudotensor Tp(lO+) with the elements a k, ilk, and ~ . The
quarternary products y~y~7~7~ form the pseudovector Vp(5+) with the five
components y~. with k = 1 , 2 , 3 , 4 , 5 . Finally the quinquinary product
Y'IY~Y~Y~7~ gives

?#1Y2
. .73
. . Y4'}'5= I ' (t9)

which constitutes the pseudoscalar Spo_ ~.

The results of this section are summarized in the arrays that appear in
Tables I and II. A number of regularities are evident in these tables, such as
the fourfold repetition period of sigma along the horizontal, vertical and
slanted columns and the binomial coefficients along horizontal rows (e.g., 1-
4-6-4-1 for C4) of Table II.

HIGHER-ORDER (n i> 6) C L I F F O R D ALGEBRAS

The algebras that have been introduced up to this point all arose from
vectors V(n+) with well known matrices for elements. To generate higher
order Clifford algebras it is necessary to start with higher order matrices.
Accordingly we define the following seven 8 × 8 anticommuting matrices zrj
constructed from the Dirac matrices 7j and unit 4 X 4 matrices denoted by I

(
0 7~)
ztj= -YJ 0
) j = 1, 2, 3, 4, 5 zt6 =
0)
-I
rc7 =
(0,)
I 0
(20)
726 Poole and Farach

o~ with 7j.
The matrix 7rj is thus a direct product of the 2 × 2 matrix (-~0)
These seven zcj.matrices and their counterparts nj

n~ = V/-Z--I zcj (21)

satisfy the anticommutation rules

7rjrk + nknj = 2djk (22a)

njzr~, + zr~nj = --2c~jk (22b)

The six matrices n~ with j = 1, 2, 3, 4, 5, 6 form the vector V(6_ ) which

generates the C 6 Clifford algebra, and the seven nj matrices form the vector
V(7+) which generates the C 7 Clifford algebra. The sixfold product of the
matrices zrj and the seven fold product of the matrices nj
! ! ! ! ! t !

(23)

7rl ~2 ~3 7C47~5 7~67C7 = I t

form the pseudo scalars n~ and I', respectively, of C6 and C 7.

Following this procedure we can define nine 16 × 16 anticommuting
matrices

0 , n) j=1,2,3,4,5,6,7, P8= P9= (24)

&= -re) 0 - 0

and their counterparts

pj = ~ - 1 pj. (25)

where the P1 anticommute in accordance with Eq. (22a) and the pj.' satisfy
Eq. (22b). The eight pj matrices with j = 1, 2, 3 ..... 8 form the vector V(8+)
which generates the algebra C8 and the nine pj with j = 1, 2 ..... 9 constitute
V(9 ) which generates C 9. The product of the eight matrices pj forms the
pseudoscalar P9 of C s and the product of the nine pj matrices forms the
pseudoscalar I ' of C9.
Higher order sets of anticommuting matrices and higher order algebras
may be generated by induction. Thus we have a general way to construct the
set of matrices which will generate any order n of a Clifford algebra C, from
the matrices of the Clifford algebra C,_ 2 two orders lower. This is
summarized in Table III. We note from the first row of the table that the
Pauli matrices - a 2 , 0 3 , and el would have to be relabeled o 1, 02, and os,
respectively, to conform to the nomenclature of the Table. We also see why
the notation Y4 is used for the matrix which is generally denoted by Yo.
Pauli-Dirac Matrix Generators of Clifford Algebras 727

Table IIL Generating Matrices for the Lowest Order Cliffor Algebra

Matrix
Algebra Dimensions Generating Matrices

C~,C 3 2×2 oz= i °3= 0 -t cq.= 1 0

C4,C~ 4×4
(
0 oj
Ys= --aj
)
0 , j = 1,2,3 Y4= 0 (, 0) (0 ;)
Y~= I

C6, C 7 8X8 :zj= --7; , j = 1,2,3,4,5 ~r6=(O

(0,
lrv= I O)

Cs,C 9 16×16 &=

(0-Tr~ 0] ' j = 1 ' 2 ' 3 ' 4 ' 5 ' 6 ' 7 (, o) P~= (0I 0,)
Ps= 0

6. R E A L A N D C O M P L E X A L G E B R A S

We have seen how each even order Clifford algebra C n and the next
higher odd order algebra Cn+ ~ have the same size matrices as a basis. For
example the pair C2 and C 3 both use Pauli type 2 × 2 matrices and the pair
C a and C 5 use Dirac type 4 × 4 matrices. Since we can write

aa; + bok = (a -- i b ) o ;
(26)
ay k + by; = (a + ib)7k

it follows that the higher or odd order vector space of a pair where the
vectors have real magnitudes is equivalent to the lower or even order vector
space of the pair where the various vectors are allowed to have complex
magnitudes. In each of the two above expressions we wrote the complex form
in terms of the basis matrices of the lower order algebra, i,e., we used o~ and
Yk, respectively, instead of a~ or y;. This property can be a great
simplification in working with odd order Clifford algebras.

7. R E L A T I O N TO P R E V I O U S V E C T O R P R O D U C T ARTICLE

In the previous article, III, we worked out a formulation of special

relativity and electromagnetism in terms of a vector product approach and at
the end of the article we noted that it was equivalent to a Clifford algebra
formulation. In this section we wilt show how to write various expressions
used there in a Clifford algebra notation.
728 Poole and Farach

In the notation of I I I a general 8-component vector or aggregate G of

C 3 is written

G = (G' + iG", ig -- g") = (G, g) (27)

where it is assumed that the C2 basis matrices ~ , a;, a~ plus the identity I
are used

G = (G'~ + iGx)O ~ ÷ (G~ + iG;)a; + (G'~ + iG~')a'3

(28)
- ( G x 'a , +
' G'y cr2t + G z a ; ) - ( G x a" l + G v a "z + G " a 3 )

g = ( - g " + ig')I (29)

= - g " I + g'I'

A preferable notation for a general 8-component vector in C a is as follows:

G = (S, V, Vp, SR) (30)

where we understand that the scalar S is multiplied by the identity I, the

vector V is expressed in terms of the usual Pauli matrices aj, the
pseudovector Vp is expressed in terms of a~ and the pseudoscalar term Sp is
assumed to be multiplied by I'. Only real coefficients are permitted for the
eight terms of Eq. (30) Comparing Eqs. (27) to (29) permits Eq. (30) to be
written

G = ( - g " , - G " , G', g') (31)

Likewise Eq. (30) in the notation of Ill becomes

G = (V. - N , iSp + S) (32)

Every vector of III is in the form of Eq. (27) and we can make use of
Eq. (31) to put them in the form of Eq. (30).
We showed in III that the vectors of special relativity may be divided
into two types in terms of their transformation properties. Odd order vectors
transform under a general Lorentz transformation in such a way that the
components of S and V mix together and the components of S~ and Vp mix
together. Such a general vector will be written

God~ = {S, V; V,, Sp} (33)

Examples of this are the velocity four vector and the energy-momentum four
vector. Even order vectors transform in such a way that the components of V
Pauli-Dirae Matrix Generators of Clifford Algebras 729

Table IV. Particular "Vectors" of Special Relativity Expressed in the Notation of III
(column 2), and in Clifford Algebra Notation (Column 3)

'Vector' of Special Vector Product Clifford Algebra

Relativity Notation Notation

Energy-momentum four {p, iE/e} t0, 0; p, E / c }

vector
Pseudo four vector {t~', -v'} /-v ', - V ' : 0, 0}
Electromagnetic Six {H -- iE, 0} t0; E, H; 0}
Vector
Space rotation about [/~ sin ½0, cos ~0] lcos ½0; 0,/~ sin 40; 0t
angle 0
Improper Space Rotation [i/~ sin 40, i cos ~0] [0: -/~ sin 10, 0: cos ½0]

SpecialLorentz [:~ / y - I /y+l 1 [ /7+ I ,~ / 7 - - 1 ~ ~I

Transformation [ fP \t ~ , ks' ---~--J [ks, ---~--; --pks' ~ - - , u: uj
[y= (I --/~')-'/' I
Inversion [0, i] [0; 0, 0; 1 ]
Identity 10, 1] [l: 0, 0; 01

and lip mix while S and S n remain unchanged. The notation for this general
vector is

Ceven = IS; V, vp; s A (34)

An example of this even order case is the electromagnetic field tensor in

which H is identified with V and E with Vp. Space rotations and Lorentz
transformations are also of even order.
Table IV lists various vectors of III in the notation of that article and
also in a Clifford algebra notation. We see form the table that the particular
vectors of special relativity constitute various ways of combining the four
parts of the algebra S, V, Vo, and Sp in pairs. More general vectors such as
a general Lorentz transformation L

z = [i/;' + 1;-, l] = [i; - c , , L"; 01 (35)

which arises from the successive application of a space rotation and a special
Lorentz transformation contains more than two parts of the algebra.

825/12/7-6
730 Poote and Faraeh

8. G E N E R A T I N G METHODS FOR HIGHER ORDER CLIFFORD

ALGEBRAS

The present article has proposed a general method for generating the
matrix representations of arbitrarily high order Clifford Algebras.
Ramashrishnan (58'9°) summarized three other general methods of generation,
and we will review these three for comparison with the present one.
In the first method originating from Dirac (87"88} two sets of three
anticommuting matrices of higher dimension can be obtained from left (Ixaj)
and right (ajxI) direct products of a high dimensional unit matrix I with the
three Pauli matrices aj, and when I is a 2 × 2 matrix we have

Y5 = (~1 xI) = (? ') 0

(36)

--34 = (cr2XI) = iI 0

)J4 = (t73XI) =
(,o) 0 --I

Dirac selected the set of four anticommuting matrices aj = - 7 4 ~ ) where

j = 1, 2, 3 and Y4 in his pioneer formulation (Sv'Ss) corresponding to the n = 4
Clifford algebra case. The fifth anticommuting matrix y5 can be added for
the next higher algebra n = 5.
The second method of generation due to Ramakrishnan (58"59} is called
the L-matrix or a-operation method. It starts with the primitive L-matrix
called L 3.

2]'3 21 -- i22 ) (37)

L 3= 21 + i2 a --2 3

=210" 1 -I-22(72 @2303 (38)

where the )~j are arbitrary coefficients, and it generates the matrix L2,,+ ~
from L 3 by the successive application of the a-operation defined as follows:

= { "~'2m+ 1/ L2m-l--i22rnl) (39)

L2m+l \L2m_~ + i22mI --22m+i / /
Pauli-Dirae Matrix Generators of Clifford Algebras 731

where I is the 2m × 2m unit matrix. The matrix L?rn. 1 is the sum of 2m + 1

anticommuting matrices u~m+
2m+1
L2m+l = ,X'~ Aj"uj2m+l (40)
j=l

which in our notation form a basis for the algebra of odd order n = 2m + t.
For the complex Dirac case m = 2 we have explicitly

L5 = 2 i a l + 2 2 a z + ~-3a3 + 24(-c~;) +25y4 (41)

where we have made use of the matrices of C 5 that were defined earlier. To
write the corresponding even order L-matrix L2m we "unsaturate" L2m+l by
omitting the 2:m+ ~I terms on the diagonal, to obtain

( 0 L2m_loi,~2mI ) (42)
Lzm = Lzm_ 1 q- i]~2ml

which gives the same summation as Eq. (40) with the j = 2m + 1 term
omitted. Thus L 4 is identical with L 5 with the 25 Y4 term omitted. In both the
unsaturated L2m and saturated L2,~+ ~ cases the basis matrices u ~-'+z are of
dimension 2 m × 2 m. This ~-operation method of generation corresponds to
cases of even signum since

2 iXj~l
Z2m+l =- ~, I (43)

and similarly for the square of L2m.

The third method of generation devised by Rashevskii 112~ forms a set of
basis matrices u~ of the even order algebra C n from the 2 n/2 elements of the
auxiliary algebra Cn/2 written in the form of an "aggregate vector" with 2 n/2
components.
For example, if we wish to generate the matrices of C 4 we write for the
aggregate of C 2

A = 2 0 1 + 212(O'~ a~) -I- )~i G'I + ).2if; (44)

= Aeve. + Aoaa (45)

where Aeven is the sum of all of the terms with even products of basis
matrices

Aeven = )~o[ + ,~,i2(~7t~(7;) (46)

732 Poole and Farach

and Aod d contains only odd products of basis matrices

Aod d =210" [ -t-226; (47)

In the general case Aeven and Aoa ~ both contain the same number of terms.
Each term of A is the product of a coefficient 2 and a matrix.
The first n matrices of C, are obtained through the mapping

A -, (A.ve. + Aod~) u;/2 (48)

and for the present case there are two u 2 basis matrices, namely a'1 and a~
which gives

A ~ Ao i = -21I- 22(0' 1a;) + 2oa i + 2,2a ~

(49)
A ~ Aa~ = - 2 f l + 2,(a i o;) -- ).,2a'~ + 2oa;

These two equations may be written in matrix form as follows

0 0 - ;t~ -22
0 0 21
1 "~2 21
(50)
0 1 [2'2/=1 21/
0--1 0 / 21 ! I-2'21
1 0 \22/ \ 2 o ff
the first matrix corresponding to 5~ and the second to - a ; . The remaining n
basis matrices for C n are obtained from the mapping

A ~ iu~./Z(Aeve, -- Aodd) (51)

and in the case of C4 they give - % and - % . This process may be

summarized in terms of the following mappings

A~ (Aeven ~- Aodd)O'tl
-- i

A _Z22, (Aeven + Aodd)O-~

i
(52)
A~ i a~(Aev¢, -- Ao~d)
/
--~1) • !
A- z G2(A~,~ - Aodd)
Pauli-Dirac Matrix Generators of Clifford Algebras 733

carried out by four basis matrics ~;, a '1, a2,

' and a 3.
' The product of these
four mapping matrices provides the fifth anticommuting matrix ?:4 of the set.

9. C O M P A R I S O N OF M A T R I X G E N E R A T I O N M E T H O D S

The method of generating basis matrices presented in this article is

closest to the a-operation method described above. It is equivalent to
generating the higher order matrices from the following modified a-
operations

= { "~2mI ~2m+11 2- iZ2m_l ) (53)

Lzm+l iLzm_ 1
\)].2m + 1/ __ --22mi
for the odd order n = 2m + 1 case and

=( 22mi iL2m-1) (54)

L ;rn \--itzm - 1 --)C2mI

for even order n = 2m. For example L~ is given by

L~ =)~l ?:1 +227z +23?:3 + ~-474 + ~57:5 (55)

Note that unsaturation is brought about by removing the 2~m+~I terms from
the off diagonal positions of the matrix L~m+l. The primitive L-matrix L 3
remains unchanged in our method of generation, the modified L-matrices
(53) and (54) being used only for m > 1.
Ramakrishnan ~15'9°) pointed out that in the Dirac method of generation
we compose matrices of a high order Clifford algebra from those of a lower
order one, while in the a-operation method we decompose an L-matrix into
the basis of a higher order algebra. From this viewpoint the method
presented in this article composes the matrices of the higher order algebras
from those of lower order, and yet it is mathematically closer to the a-
operation method. In the Rashevskii method on the other hand, the basis
matrices of a higher order algebra are obtained through mappings involving
lower order bases.
The method of basis matrix generation introduced in this article has the
advantage that it provides directly the sets of anticommuting matrices widely
employed in physical applications of spinors, namely the three Pauli spin
matrices aj, the five Dirac gamma matrices 7j given by Eq. (55) and their
higher order analogues. The other three methods of generation do not give all
of these particular matrices directly, as can be seen by comparing Eqs. (11)
and (55) with Eqs. (36), (41), and (52), respectively.
734 Poole a n d F a r a c h

10, CLASSIFICATIONS OF C L I F F O R D A L G E B R A S

There are a number of ways to classify Clifford algebras, and in this

section we will comment upon how the present algebras fit into broader
schemes of classification.
There are n + 1 types of Clifford algebras of dimension n, and each is
designated by its signature (p, q) where p and q are positive integers or zero
which satisfy the relation

p + q= n (56)

Each algebra has n anticommuting orthonormal basis elements u~,/'/2 IAn

.....

with the property that p of them have the square uj2 = +1 and q of them have
the square uj-2--1. In the notation of this article the algebras (n, 0) have a
positive signum and (0, n ) h a v e a negative signum. Other algebras with p 4:0
and q 4:0 are said to have a mixed signum and those with p = q are called
neutral.
The various high order algebras are constructed from three basic
algebra types associated with the real numbers (R), the complex numbers
(C) and the quarterions (H). Porteous (7) uses the notation R(n') and H(n')
for algebras of even dimension n where n' = 2 "/2 for R(n') and n' = 2 ~ n - l ) / 2
for H(n'), and he uses the symbol C(n') for the odd dimensional algebras n
where n ' = 2 ("-1)/2. For odd n there are also double field algebras 2R(n')
and 2H(n') where n' = 2 ~"-1)/2 for the former and n' = 2~"-3)/2 for the latter.
The label (n') is omitted in each case when n ' = 1, and hence, for example,
we write C in lieu of C(1).
In an alternate classification scheme Salingaros (~9) employs the symbol
S i for C type algebras and A) for both R and H type algebras, where each
particular algebra is assigned a particular value o f j . Salingaros (~9) uses the
same symbols R, C, and H for the n' = 1 cases of Porteous, and in addition
he employs the special symbol S for the Pauli or quaternian algebra S~
(which Porteous designates as C(2)) and the special symbol D for the Dirac
algebra S 2 (which Porteous calls C(4)). The relationships between these two
classification schemes and the notation employed in this article are
summarized on Table V. This table continues with a periodicity of 8 in the
sense that if we write C, ~A(n') where A is R, C, or H then C~+ 8 ~ A ( 1 6 n ' ) .

11. M I X E D A N D N E U T R A L C L I F F O R D A L G E B R A S

The present article dealt only with Clifford algebras having a positive or
negative signum, i.e., it was only concerned with (0, n) and (n, 0) signatures.
 ~7

Table V. Comparison of the Various Names and Notations for the Algebras Treated in this Article

Various Notations

Order Signature Present

n (P, q) Signum work Porteous ~7) Salingaros ~9) Basis Name of Algebra

(0, o) Co R(1) =R N0=R Real Number

(0, l) Cl C(1) = C S0 = C i Complex Number
(0, 2) Cz H(1): H N2:H a'2,a ~ Quaternion
¢
(3, o) + C3 C(2) SI = S a~, oh,or 3 Spinor, Pauli or
Complex Quaternion
(4, o) + C4 H(2) N4 yj, j = 1, 2, 3, 4 Dirac
(0, 5) C~ C(4) S 2= D yj, j --= 1, 2, 3, 4, 5 Full Dirac
(0, 6) Ca R(8) N~ ~zJ, j = 1, 2 ..... 6
(7, o) + C7 C(8) S~ zci, j - - 1, 2 ..... 7
(8, o) + C8 R(16) N7 p~, j = l, 2 ..... 8
(0, 9) C9 C(16) S, pj, j = 1, 2 ..... 9
736 Poole and Farach

The basis vectors presented in T a b l e s III a n d V c a n also be e m p l o y e d to

generate m i x e d a n d n e u t r a l s i g n u m algebras. F o r e x a m p l e , the basis set e'~,
a ; , a 3 generates the mixed a l g e b r a with the s i g n a t u r e (1, 2) a n d the basis set
crI , a ; generates the n e u t r a l a l g e b r a with the s i g n a t u r e (1, 1). T h e M a r j o r a n a
a l g e b r a M is a m i x e d type with the s i g n a t u r e (3, 1).

REFERENCES

1. Y. Aharonov, H. A. Farach, and C. P. Poole, Jr., Amer. J. Phys. 45,451 (1977); referred
to as I.
2. H. A. Farach, Y. Aharonov, and C. P. Poole, Jr., Amer. J. Phys. 47, 247 (1979); referred
to as II.
3. C. P. Poole, Jr., H. A. Farach, and Y. Aharonov, Found. Phys. 10, 718 (1980); referred
to as III.
4. W. K. Clifford, Amer. J. Math. 1, 350 (1878).
5. B. L. van der Waerden, Algebra (Springer Verlag, Berlin. 1967).
6. S. Lang, Algebra (Addison Wesley, Reading, Mass. 1965).
7. R. Porteous, Topological Geometry (Van Nostrand Reinhold, Co., New York, 1969).
8. M. Riesz, Comptes Rendus du DixiOme Congrds des Math. Scan. (Copenhagen, 1946),
p. 123ff.
9. M. Riesz, Comptes Rendus du Douzidme Congrds des Math. Scan (Lund, 1953), p. 241ff
10. M. Riesz, Clifford Numbers and Spinors Chap. I-IV, Lecture Series No. 38 (University of
Maryland, 1958).
11. E. M. Corson, Introduction to Tensors, Spinors and Relativistic Wave Equations (Blackie
and Son, London 1953).
12. P, K. Rashevskii, Amer. Math. Soc. Transl. 6, 1 (1957).
13. S. Teitlen Suppl. Nuovo Cimento 3, I, 15 (1965).
14. S. Teitler, J. Math. Phys. 6, 1976 (1965); 7, 1730, t739 (1966).
15. A. Ramarishnan, L-Matrix Theory and the Grammar of Dirac Matrices (Tata-McGraw
Hili, Bombay. 1972).
t6. D. Hestenes, Am. J. Phys. 47, 399 (1979).
17. D. Hestenes, & Math. Phys. Appl. 24, 313 (1968).
18. D. Hestenes and G. Sobczyk, Clifford Algebras and Geometric Calculus: A Unified
Language for Mathematics and Physics, Mathematics Lecture Note Series (Addison
Wesley, Reading, Mass. 1980).
19. N. Salingaros, J. Math. Phys. 22, 226 (1981).
20. N. Salingaros and M. Dresden, Phys. Rev. Lett. 43, 1 (1979).
21. R. H. Good, Jr., Rev. Mod. Phys. 27, 187 (1955).
22. D. Hestenes, Space 77me Algebra (Gordon & Breach, New York 1966).
23. G. Dixon, J. Math. Phys. 18, 2204 (1977).
24. K. Konco, M. Kawaguehi, and K. Kondo, Raag Memoirs of the Unifying Study of Basid
Problems in Engineering and Physical Science by Means of Geometry, IV (1968).
25. J. Palmer, or. Math. Anal. and Appl. 64, 189 (1978).
26. A. B. Pestov, ProbL Teor. H Gravitas. H and Elem. Chastits 5, 54 (1974).
27. J. Jayaraman, J. Phys. A 9, 1131 (1976).
28. P. A. Collins and R. W. Tucker, Nucl. Phys. BI21, 307 (1977).
29. K. Greider, Phys. Rev. Lett. 44, 1718 (i980).
Pauli-Dirae Matrix Generators of Clifford Algebras 737

30. N. lonescu-Pallas, "Abstracts of the Conference on Magnetic Resonance and Telated

Phenomena," Inst. A tom. Phys. ( 1971).
3t. D. Hestenes, J. Math. Phys. 8, 798 (1967).
32. D. Hestenes and R. Gurtler, Amer. J. Phys. 39, 1028 (1971).
33. A. Ercoli-Finzi and C. Morosi, Atti. Aecad. Naz. Lincei, Rend. Cl. Sci. Fis. Mat. and
Nat. 59, 421 (1975)~
34. A. E.Finzi and C. Morosi, Atti. Accad. Naz. Lineei, Rend. Cl. Sci. Fis. Mat. and .Nat. 59
(1975).
35. M. Novello, & Math. Phys. 12, 1039 (1971).
36. J. A. Brooke, J. Math. Phys. 19, 952 (1978).
37. J. A, Brooke, J. Math, Phys, 21, 617 (1980).
38. A. Crumeyrolle, Ann~ Inst. H. Poincard A I1, 19 (1969).
39. F. A. Doria, Lett. Nuovo Cimento 8, 994 (1973).
40. F. A. M. Frescura and B. J. Hiley, Found. Phys. 10, 7 (1980).
41. W. Graf, Ann. Inst. H. Poincar~ Sect. A 29, 85 (1978).
42. D. Hestenes, Amer. J. Phys, 39, 1013 (I971).
43. R. Kerner, J. Math. Phys. 21. 2553 (1980).
44, V. E. Stepanov, Vestn. Leningr. Univ. Fiz. and Khim 1, 104 (1979).
45, M, Sirugue-Collin and M. Sirugue, Commun. Math. Phys. 48, 131 (1976).
46, J. O. Winnberg, J. Math. Phys. 18, 625 (1977).
47. K. Bugajska, Int. J. Theor. Phys. 18, 95 (1979).
48. P, Lounesto, Celestial Mech. 17~ 207 (1978).
49. P. Lounesto, Spinors and Brauer-Wall Groups Report HTKK-MAT-A124 Helsinki,
University of Technology (1978).
50. G, Sobczyk, Aeta Phys. Pol. B l l , 579 (1980).
51. D. Dominici and G. Longhi, Nuovo Cimento A 46A, Set. 2, 213 (1978).
52. A. P. Hristev, J. Math. Phys. New York 12, 118 (1971).
53. J. D. Edmonds, Lett. Nuovo Cimento 17, Ser. 2, 465 (1976).
54. A. Ramakrishnan, Madras, India, New York, Plenum, Jan. (1969).
55. L. P. Horwitz and L. C. Biedenharn, J, Math. Phys. 20, 269 (1979).
56. S. A. Basri and L. P. Horwitz, Phys. Rev. D 11, 572 (1975).
57. T. S. Santhanam, J. Math. Phys. 11, 1122 (1970),
58. A. Ramakrishnan, The Weak Interaction Hamiltonian in L Matrix Theory (Seminar at
Rand Corp., July 1970).
59. E. V. Pavlov, Izv. Vuz Mat. 22, 64 (1978).
60. J. D. Edmonds, Jr.. Int. J. Theor. Phys. 15. 911 (1976).
61. J. P, Fillmore, Int. J. Theor. Phys. 16, 937 (1977).
62. R. W. Hayward, The Dynnamies o f Fields of Higher Spin (NBS Report Mono-154, Aug.
1976).
63. G, G. Markov and A. P. Norden, Izv. Vuz Mat. 19, 82 (1975).
64. T. S. Santhanam, P. S. Chandrasekaran, and N. B. Menon. J. Math. Phys~ 12, 377
(1971).
65. F. A, Berezin, The Method of Second Quantization Frame, N. Mugibayashi (Academic
Press, New York 1966).
66. R. Delanghe, Meded. Kon. Vlaamse Aead. Wet. 32, 3 (1970),
67. R. Delanghe and F. Brackx, J. Funct. Anal. 37, 164 (1980).
68. G. Dixon, J. Math. Phys. 19, 2103 (1978).
69. D. Hestenes, Math. Anal. and AppL 24, 467 (1968).
70. D. Hestenes, J. Math. Phys. 15, 1768 (1974).
71. M. Kenmoku and S. Naka, Prog. Theor. Phys. 59, 1397 (1978).
738 Poole and Farach

72. C. N. Ktorides, J. Math. Phys. 16, 2123 (1975).

73. J. T. Lewis and P. N. M. Sisson, Commun. Math. Phys. 44, 279 (1975).
74. J. Manuceau and A. Verbeure, Commun. Math. Phys. 18, 319 (1970).
75. M, Gfinaydin, Nuovo Cimento A 29, Ser. 2, 467 (1975).
76. F. N. Ndili and G. C. Chukwumah, Int. J. Theor. Phys. 11, 261 (1974).
77. R. Penney, Nuovo Cimento B 3B, Ser. 2, 95 (1971).
78. H. D. Doebner and T. D. Palev, C. R. Acad. Bulg. Sci. 26, 151 (1973).
79. A. Van Dael, Commun. Math. Phys. 21, 171 (1971).
80. H. Tilgner, Ann. Inst. H. Poineard, A 14, 179 (1971).
81. U. E. Schroder, Acta Phys. Austriaea 33, 328 (1971).
82. R. M. Lovely and F. J. Bloore, Lett. Nuovo Cimento 6, Ser. 2, 302 (1973).
83. R. Jagannathan and R. Vasudevan, J. Math. Phys. 19, 1493 (1978).
84. N. Salingaros, Hadronic J. 3, 339 (1979).
85. A. Schober, W. Beiglbock, A. Bohm, and E. Takasugi, Seventh International Colloquium
and Integrative Conf. on Group Theory and Math. Phys. (Sept. 1979).
86. L. P. Horwitz, D. Sepunaru, and L. C. Biedenharn, Ann. Isr. Phys. Soc. 3, 300 (1980).
87. P. A. M. Dirac, The Principles of Quantum Mechanics, 2nd. Ed. (Oxford, New York,
1935) Chapter XII.
88. P. A. M. Dirac, Proe. Royal Soc. A l l 7 , 610 (1928).
89. G. Arfken, Mathematieal Methods for Physicists (Academic Press, New York 1970).
90. A. Ramakrishnan. Syrup. Theor. Phys. Math. 10, 63 (1970).