Professional Documents
Culture Documents
CLIFFORD ALGEBRAS
P. LOUNESTO
In,titute 0/ Mathematic,
Hel,inki Unil1er,ity 0/ Technology
SF-O!150 ESPOO, Finland
Abstract. This article reviews Marcel Riesz's lecture notes on Clifford Number, and
Spinor" 1958, and evaluates its effect on present research on Clifford algebras.
The article begins with a critical survey of the history of Clifford algebras. Inaccuracies
in citations are pointed out and mistaken priorities are rectified. Misconceptions about
Clifford algebras are examined and flaws are corrected. The paper focuses on controversial
issues that have been debated over the past few years, and offers definitive resolutions.
Concrete counter-examples are given to misplaced claims.
Particular attention is directed to the existence of a canonical linear isomorphism
between the exterior algebra and the Clifford algebra. ChevalIey constructed the Clifford algebra Cl(Q) of a quadratic form Q as a subalgebra of the endomorphism algebra
End(A V) of the exterior algebra A V. This construction depends on an arbitrary, not
necessarily symmetric, bilinear form B such that B(x,x) = Q(x). The choice of B
fixes the contraction u J 'II in A V and permits the introduction of a Clifford product
xu = x J u + x "u of x E V and u E A v. This gives rise to the Clifford algebra of the
symmetric bilinear form HB(x,y) + B(y, x when the characteristic =1= 2.
Marcel Riesz went backwards with his fundamental formulas
x J u=
+ 'li.x)
and re-introduced the contraction and the exterior product in the Clifford algebra, and
constructed, in alI characteristics =1= 2, a privileged linear isomorphism from the Clifford
algebra to the exterior algebra (without antisymmetric products of vectors). [In the above
'Ii. means the grade involute of u, that is, for a homogeneous element a EA V we have
i. = (-l)a.]
Key words: Exterior algebra - contraction - Kiihler-Atiyah isomorphisms
216
P. LOUNESTO
217
E.
CARTAN
Clifford's geometric algebra was reinvented in 1880, just two years after
its first publication, by Rudolf Lipschitz, who later acknowledged Clifford's
prior discovery in his book, see R. Lipschitz: Untersuchungen tiber die Summen von Quadraten, 1886 [Budinich&Trautman 1988 p. 2 r. 18 refer to 1886
but seem to be unaware of 1880]. In his study on sums of squares Lipschitz
considered a representation of rotations by complex numbers and quaternions and generalized this to higher dimensional rotations in 1Rn. Lipschitz
thus gave the first geometrical application of Clifford algebras in 1880 [Budinich&Trautman 1988 p. 54 r.10 again refer only to 1886]. He expressed a
rotation y = (1 +A)(1 -A)-IX (written here in modern notation with an antisymmetric matrix A) in the form y - Ay = x + Ax or as y +y. B = x+ Bx
where BEN 1Rn is the bivector determined by Ax = Bx (= -x B). He
rewrote this expression with an even Clifford number a E Clci,n in the form
ya = ax, thus representing the rotation as y = axa- l . (Lipschitz wrote
yal = az where z = xe l \ y = yell, al = elael l .) In modern terms, he
used the exterior exponential a = 1 + B + B /\ B + . .. of the bivector
B so that the normalized element a/lal was in the spin group Spin(n).
[Delanghe&Sommen&Soucek 1992 p.127 follow here Helmstetter 1986 and
attribute these results incorrectly to Sato&Miwa&Jimbo 1978.)
Vahlen 1897 found an explicit expression for the multiplication rule of
two basis elements in Clo,n
e "'le"'2
I 2
e"'''ef31ef32
n
I 2
e",,,+{3,,
n
where the exponents are 0 or 1 (added here modulo 2, though for Vahlen
1+ 1 = 2 so that the summation was over i > j) - a formula reinvented frequentlyafterwards [Brauer&WeyI1935, Deheuvels 1981 p. 294, or disguised
with index sets as in Brackx&Delanghe&Sommen 1982 p. 2, Carmichael 1984
p. 228, Chevalley 1946 p. 62, Artin 1957 p. 186, or then hidden among permutations like in Delanghe&Sommen&Soucek 1992 pp. 58-59 who gratuitously
218
P. LOUNESTO
attribute this result to Kiihler 1960/62]. In 1902 Vahlen initiated the study
of Mobius transformations of vectors in 1Rn (or paravectors in 1R +1Rn) by
2 X 2-matrices with entries in Clo,n. This important study was re-initiated
by Ahlfors in the 1980's.
E. Cartan 1908 p. 464 identified the Clifford algebras Clp,q as matrix
algebras with entries in 1R, <C, 1H, 1RE91R, 1HE91H and found a periodicity of
8 (a re1inement of Clifford's original 4). Cartan also observed spinor modules
of orthogonal Lie algebras in 1913 (though he did not introduce the term
"spinor"). Spinors got their current name after they were reinvented and
applied to the spinning electron by Pauli 1927 and Dirac 1928.
Two-valued spinor representations of rotations in n-dimensions were reconstructed recursively by Brauer&Weyl in 1935. They applied Clifford algebras without using the term "Clifford algebra" [misleading wording in
Chevalley 1954 p.4, Crumeyrolle 1990 p. xi]. Brauer&Weyl also reinvented
(in the case of the Clifford algebra Cln ) the above explicit multiplication
formula of Vahlen, but failed to observe the connection to the Walsh functions (discovered in the meanwhile by Walsh 1923). This connection to the
Walsh functions was brought to general attention by Hagmark&Lounesto
1986.
0.3.
Maxwell's equations were condensed into a single equation by complex vectors (Silberstein 1907), complex quaternions (Silberstein 1912/1914), spinors
(Laporte&Uhlenbeck 1931) and by Clifford algebras (Juvet&Schidlof 1932,
A. Mercier 1935) - [mistaken priority in Hestenes 1966 p. 29]. In the Clifford
algebra Cia,}
M4(1R) the Maxwell equations were written as {)F = J
where {) = V - eo{)o and F = Eeo - Be12a E A21Ra,l. The single Maxwell
equation could be decomposed into two parts, {) A F = 0 and {). F = J. Similarly, {)A = -F could be decomposed into two parts, {)AA = -F and the
Lorenz condition {). A = 0 [discovered along with the Lorenz gauge by the
Danish physicist Ludwig Lorenz and not the Dutch physicist H. A. Lorentz,
who invented the Lorentz metric and the Lorentz transformations, see J. van
Bladel: Lorenz or Lorentz? IEEE Antennas and Propagation Magazine 33
(1991) p.69 and The Radioscientist 2 (1991) p.55]. Marcel lliesz 1947 expressed the Maxwell stress tensor as T/J" = - t(e/JFe"F)o and Hestenes 1966
p. 31 introduced the vectors T/J = -tFe/JF for which T/J" = T/Je" = e/J.T"
and the mapping Tx = -tFxF where (Tx)/J = T/J"z" [Juvet&Schidlof
1932 p. 141 gave
"'F.
1
F. F",{J
T/J&J = F/J "''' + 4" g/J" "'/3
=-
219
Juvet 1930 and Sauter 1930 replaced column spinors by square matrices
in which only the first column was non-zero - thus spinor spaces become
minimal left ideals in a matrix algebra. Riesz 1947 used primitive idempotents of Clifford algebras to construct spinor spaces as minimal left ideals
in Clifford algebras [Budinich&Trautman 1988 p. 12 misplaced priority to
Chevalley 1954]. Hestenes 1966 formulated the Dirac theory with the real
Clifford algebra Gl 1 ,3 M2(H) of the Minkowski time-space lR1 ,3 - this
led him in 1966/67 to the lucid but at first sight perplexing idea of replacing spinors spaces by the even subalgebra elt 3 of the real Clifford algebra
Gl1 ,3.
'
0.4.
THE ROLE OF
E.
WITT
E. Witt 1937 used Clifford algebras in his study on quadratic forms over arbitrary fields of characteristic :j:. 2. Witt's paper began the modern algebraic
theory of quadratic forms, and recognized Clifford algebras as algebras of
quadratic forms. The Witt ring W(K), of a field K, consists of similarity classes of non-singular quadratic forms over K (similar quadratic forms
have isometric anisotropic parts). In characteristic :j:. 2 the structure of Clifford algebras of some special quadratic forms was studied by Lee 1945/48
(el = 1), by Chevalley 1946 (el = -1), and by Kawada&Iwahori 1950
(el = 1) without the benefit of the Witt ring (and so they did not consider all the isometry classes of anisotropic quadratic forms).
0.5.
ISOMORPHISMS OF CHEVALLEY,
M.
220
P. LOUNESTO
"n
221
'2 (Q(x + y) -
Q(x) - Q(y))
=p + q.
AV
-+
Cl(Q)
Start from the exterior algebra A V. Define for a k-vector a E AIe V the
grade involution a = (-l)lea and the reversion ii = (_l)tle(le-l)a. The
symmetric bilinear form associated with Q on V can be extended to simple
k-vectors in Ale V by way of <Xl A X2 A A XIe,Yl A Y2 A . A n> =
det<xi,Yj>, where
<Xt,Yl>
<X2,Yl>
<Xt,Y2>
<X2,Y2>
for all wE /\ V.
P. LOUNESTO
222
I <y,x>
<z,x>
<y,W>
<z,w>
<Xl, Wl>
<X2,Wl>
<Xl, W2>
<X2,W2>
<Xl, Wk-l>
<X2, Wk-l>
i=l
and so
i=l
4. x J (u /\ v) = X J ((Ul/\ U2 /\ ... 1\ Ui) 1\ (Vl 1\ V2 /\ ... 1\ Vj))
i
Uk)( -1)k-l ul
/\ U2 1\ 1\ uk-l 1\ Uk+! 1\ 1\ Ui /\ v
k=l
j
+( _1)i
= (x J u) 1\ v + (-I)iu 1\ (X J v).
vr /\
223
(a)
(b)
xJ(uAv)=(xJu)Av+uA(xJv)
( c)
( U A v) J w = u J (v J w)
for u, v, w E AV
224
7.
P. LO UNESTO
u
Lv =
Vo
J tto
+ Vo J
ttl - VI
8. a J b = (_l),U-i)b L a for a E
Uo
,,'V,
9. a E ,,10 V, x E V, x J a
=0
(a J b) A b
:>
+ 111 J
UI
11
J u - 2V1 J
b E I\ i V
x = a J b for some b E
=0
Uo
,,"'+1 V
ab = a J b for a el\v,
be/\V, i:5:i
.:
225
'inner product' but his rules are not sufficient to permit the evaluation of
(X 1\ y) . u when U E /\ V, U rf. V (though they do permit a construction of
the 'inner product' with an additional rule (x 1\ y) . U = X (y . u) where
u E /\k V, k ;::: 2).
The above remark shows how the asymmetric contration solves a problem
of Hestenes&Sobczyk 1984, who postulate the 'inner product' to be 0 [p.6,
r. 12 formula (1.21b)) if one of the factors is a scalar, and run into difficulties
on p. 20 rows 8-18 formula (2.9). However, as the following example shows
the problem is deeper than that since the 'inner product' is not equal to the
contraction even though scalars would be excluded.
Example. Let el, e2 be an orthonormal basis for R 2. Compute
+ el 1\ e2) = 1 + e2
+ el 1\ e2) =
1 - el.
This shows that neither of the contractions coincide with the 'inner product'
of Hestenes&Sobczyk.
Remark. We may re-obtain the dot product in terms of the Clifford product
as follows (char i- 2)
a b
= (ab)li-il
for a E /\ V, bE /\ V
where (U)k is the k-vector part of U E /\ V ::: Cl( Q). For two homogeneous
elements a E /\i V, b E /\i V the Clifford product decomposes into k-vector
parts
ab = (ab)i+i + (ab)i+i-2 + ... + (ab)li-il
where al\b = (ab)i+i and ab = (ab)li-il' In particular, ab =I al\b+aJb,
when i =11 even for i < j, as can be seen by the counter-example
a
2 3 3
ab = e134 E /\R 4
CHEVALLEY'S IDENTIFICATION OF
Cl(Q) C End(/\ V)
= <px E End(/\ V)
v.
226
P. LOUNESTO
/\ V
->
Cl(Q)
t End(/\ V) t /\ V
was the identity mapping on A V. The faithful representation 1/J sent Cl( Q)
onto an isomorphic subalgebra of End(A V).
Chevalley's identification works fine also with a contraction defined by
an arbitrary - not necessarily symmetric - bilinear form B such that
B(x, x) = Q(x) and
= B (x, y)
for x, y E V
( a)
x J y
(b)
xJ(u/\v)=(xJu)/\v+u/\(xJv)
(c)
(u /\ v) J w
=u J
(v J w) for u, v, w E A V
i=l
and the faithful representation 1/J sends the Clifford algebra Cl( Q) onto an
isomorphic subalgebra of End(A V), which, however, as a subspace depends
onB.
Remark. Chevalley introduced his identification Cl(Q) C End(A V) in
order to be able to include the exceptional case of characteristic 2. In characteristic f= 2 the theory of quadratic forms is the same as the theory of
symmetric bilinear forms and Chevalley's identification gives the Clifford
algebra of the symmetric bilinear form <x,y> = t(B(x,y) + B(y,x))
satisfying xy + yx = 2<x, y>.
I
For arbitrary Q but char K f= 2 there is the natural choice of the unique
symmetric bilinear form B such that B(x,x) = Q(x) giving rise to the
canonical/privileged linear isomorphism Cl( Q) -7 A V. The case char K =
2 is quite different. In general there are no symmetric bilinear forms such
that B( x, x) = Q (x) and in the case that there is such a symmetric bilinear
form, then it is not unique, since any alternating bilinear form (is also symmetric and) could be added to the symmetric bilinear form without changing
227
"V.
= 0100)
= aOOO)
(first column
= el J el = a
(second column
ele2 = el A e2 + el J e2 = el A e2 + b
(third column = bOOl)
el (el A e2) = el J (el A e2) = (el J ed A e2 - (el J e2) A el = ae2 el el
bel
= (el A e2) L el = el A (e2 L ed - e2 A (el L el) = eel (el A e2)e2 = (ele2 - el J e2)e2 =
- (el J e2)e2 = del - be2
(el A e2)(el A e2) = (ele2 - el J e2)(el A e2)
= el(e2 A (el A e2) + e2 J (el A e2 - (el J e2)(el A e2)
= el(ee2 - ded - b(el A e2) = -ad + be + (-b + c){el A e2)'
(el A e2 )el
b
0 0
0 0
0 1
e
0
e,
a
0
0
el A ea
e
-a
0
or in general
U
C
Ul
U2
U12
'
0
d
-b
0
aUl + eU2
Uo + eU12
- au 12
-U2
e,
0
0 -1
d
0
0
0
-aYbO)
-b+ e
bUl + dU2
dU12
Uo - bU12
Ul
-(ad
- bolu"
-(bUl + dual
Uo
aUl + cU2
+ (-b + e)ul2
ae2
228
P. LOUNESTO
:f
=d
-bel + ae2
-del + ee2
-ad + be + (-b + e)el 1\ e2
-el 1\ e2 + e
eel - ae2
In characteristic
= a,
2 we find
1
1
'2 (el e2 - e2 e d = el 1\ e2 + '2 (b - e)
'2 (xy -
'2 (b -
e)(xlY2 - x2yd
= x 1\ Y + A( x, y)
aXlYl +
Example. Take a special case of the previous example, the Clifford algebra
with Q 0 and B(x,y) b(XlY2 - 2:2Yl). Send the matrix of the exterior
product (with the synunetric bilinear form = 0) to a matrix of the isomorphic
Clifford product (determined by the alternating bilinear form = B)
Uo
229
In this case
J3(x)J3(y) = J3(x A Y + B(x, y)).
The above example shows that those who refute the existence of k-vectors
in a Clifford algebra Cf(Q) over K, char K f- 2, should also exclude fixed
subspaces A k V c A
v.
Ul
Ul
Uo
U2
U12
-U2
C'
Uo -
U2
for B 2 :
Ul
U2
U12
Uo
+ Ul2
0
-U2
-"1 )
U12
Ul
Uo
0
0
o0 )
Uo
Ul
U12
U2
Uo
U12
230
P. LOUNESTO
Bl
el
e2
el A e2
B2
el
e2
el A e2
el A e2
el
e2
0
-el
1 + el A e2
0
0
-el A e2
0
-e2
-el A e2
e2
el A e2
el
0
0
el A e2
0
1- el A e2
e2
0
el
el A e2
with respect to the basis 1, el, e2, el Ae2 for " V. In this case there are only
two linear isomorphisms "V -+ Cl(Q) which are identity mappings when
restricted to K + V and which preserve parity (send even elements to even
elements and odd to odd). It is easy to verify that the above multiplication
tables of Cl( Q) are actually the only representations in
In this case
there are no canonical linear isomorphisms
-+ Cl(Q), in other
words, neither of the above multiplication tables can be preferred over the
other. In particular, N V cannot be canonically embedded in Cl(Q), and
there are no bivectors in characteristic 2!
"V "V.
The above example shows that Lawson&Michelsohn pp. 10-11 are wrong
when they allow characteristic 2 and claim that there is a canonical linear
isomorphism" V -+ Cl(Q). Chevalley used unnecessarily complicated and
abstract methods, since he wanted to include the case char K = 2 and
employed endomorphisms of
to get a faithful representation of Cl(Q).
The need for a simplification of Chevalley's presentation is obvious. For
instance, B. L. van der Waerden: On Clifford algebras, 1966, said that 'the
ideas underlying Chevalley's proof (p. 40) are not easy to discern' and gave
another proof, equivalent but easier to follow. (Also Crumeyrolle 1990 p. xi
claims that 'Chevalley's book proved too abstract for most physicists' and
in a Bull. AMS review Lam 1989 p. 122 admits that 'Chevalley's book on
spinors is ... not the easiest book to read. ') It might be helpful to get acquainted with a simpler and more direct method of relating " V and Cl( Q)
due to Marcel lliesz: Clifford Numbers and Spinors, 1958, pp. 61-67. lliesz introduced a second product in Cl( Q) making it isomorphic with" V without
resorting to the usual completely antisymmetric Clifford product of vectors
and constructed a privileged linear isomorphism Cl( Q) -+ " V.
"V
1.3.
Cl(Q)
Start from the Clifford algebra Cl( Q) over K, char K -::J. 2. The isometry
x -+ -x of V is extended to an automorphism of Cl( Q) called the grade
involution u -+ U. Define the exterior product of x E V and u E Cl( Q)
231
by
(xu + ux),
x Au =
u Ax =
(ux + xu)
and extend this exterior product by linearity to all of Cl( Q) which then
becomes isomorphic to AV. The exterior products of two vectors x A y =
(xy - yx) are simple bivectors and they span A2 V. The exterior product
of a vector and a bivector x A B = (xB +Bx) is a 3-vector in A3 V. The
subspace of k-vectors is constructed recursively by
1
(x A y) A z = 4" (xyz - yxz
+ zxy -
zyx).
= xyz -
zyx + (zy
+ yz)x -
x(yz + zy)
= yzx -
xzy.
when char K -=f 2, 3 (note the resemblance with anti symmetric tensors).
[Similarly, we may conclude that xyz + zyx = x(yz + zy) - (xz + zx)y +
z(xy + yx) is a vector in V.] lliesz's construction shows that bivectors
exist in all characteristics -=f 2.
Introduce the contraction of u E Cl( Q) by x E V so that
xJ u
(xu - ux)
(xuv - Uvx)
= !2 (xuv -
(xuv - uvx)
232
P. LOUNESTO
Thus one and the same contraction is indeed a derivation for both the exterior product and the Clifford product. (Kahler 1962 p. 435 (4.4) and p. 456
(10.3) was aware of the equations x J (u 1\ v) = (x J u) 1\ v + it 1\ (x J v)
and x J (uv) = (x J u)v +it( x J v).) Provided with the scalar multiplication
( u 1\ v) J w = u J (v J w) the exterior algebra A V and the Clifford algebra
Cl(Q) are linearly isomorphic as left A V-modules.
Exercises. Show that (char K
f:.
2)
5. (xyz - zyx)2 E K, x 1\ Y 1\ z
= Hxyz -
zyx)
6. a 1\ b
7.
u,v E A V,
Uo
= even(u),
Vo
= even(v),
UI
= odd(u),
VI
2VI
1\ UI where
= odd(v)
(u J v)o (=
u J v)
Riesz's construction shows that Crumeyrolle 1990 pp. 42-45, p. 292 is mistaken when he excludes characteristic 2 and claims that there is no canonical
linear isomorphism A V -+ Cl(Q). Crumeyrolle seems to take it both ways
when he agrees that in char f:. 2 there is a natural choice for Cl(Q) c
End(A V) given by the unique symmetric B such that B(x,x) = Q(x).
Crumeyrolle and Lawson&Michelsohn have opposite opinions about the existence of a canonical linear isomorphism A V -+ Cl(Q) but both are wrong:
Lawson&Michelsohn because they allow char 2 and Crumeyrolle because
he excludes char 2. Curiously, Crumeyrolle excludes char 2 even though he
uses Chevalley's method developed specifically for char 2. Neither Crumeyrolle nor Lawson&Michelsohn have understood that Chevalley presented his
regular-type representations of Cl(Q) in End(A V) in order to be able to
handle also the exceptional case of char 2. [This motive of Chevalley has
also escaped Harvey 1990 p. 180 and Budinich&Trautman 1988 p. 58.]
(Roughly speaking, Lawson&Michelsohn claim that two numbers x, y in
a field K always have a mean !(x + y) within that field K, but this is false
in the field of two elements K = {O, I} and Crumeyrolle claims that no
233
1.4.
Definition. An associative algebra over K with unit 1 is the Clifford algebra Cl of a non-degenerate Q on V if it contains V and K = K . 1 as
distinct subspaces so that
= Q(x)
for any x E V
(1)
x2
(2)
(3)
while the condition (3) becomes Porteous' 1969 requirement ele2'" en f:.
1 which is needed only in the signatures p - q = 1 mod 4. The relations
(La) also generate a lower-dimensional non-universal algebra in any signature p - q = 1 mod 4 in which all the basis elements ei commute with
el e2 ... en and (el e2 ... e n )2 = 1. However, no similar counter-examples
exist in even dimensions. Therefore, whereas it is correct to introduce the
Clifford algebra of the Minkowski space-time without the condition (3), in
arbitrary dimensions it is controversial to omit condition (3).
The above definition gives a unique algebra only for non-degenerate (nonsingular) quadratic forms Q. In particular, it is not good for a degenerate
Q as is shown by the following two counter-examples in case Q = O.
1. Define for x, y E V the product xy
K Efl V an algebra of dimension n + 1.
= O.
+ +
This shows that it is not possible to replace the condition (3) by the requirement that only parallel vectors commute. We could include arbitrary
quatratic forms Q by requiring instead of condition (3) that the product
of any set of linearly free vectors in V should not belong to K. However,
even this would leave some 'ambiguity' in the definition by generators and
relations. The above definition results in a unique algebra only 'up to isomorphism'. Here are two more examples to clarify the meaning of this statement:
234
P. LOUNESTO
o
o
o
A
el
e2
el /\ e2
el
e2
el /\ e2 + b
-el /\ e2 - b
-bel
-be2
el /\ e2
-bel
-be2
_b 2 - 2bel /\ e2
where b > o. Denote the second product by u Av. Rearrange the multiplication table of the second product into the form
A
el /\ e2 + b
o
o
0
0
which shows that we have generated a new exterior algebra Arn. 2 on rn. 2 ,
different from Arn. 2 but isomorphic with Arn. 2 In other words, we have
introduced a linear mapping a: Arn. 2 -+ Arn. 2 for which a(ed = ei, i =
1,2 and a( el /\ e2) = el Ae2 = el /\ e2 + b so that it is the identity on rn. 2
and gives an isomorphism between the two products a( u /\ v) = a( u) Aa( v).
4. An orthonormal basis el, e2 for rn. 2 satisfying eiej+ejei = 2liij generates
the Clifford algebra Cl 2 = Cl 2 ,o with basis 1, el, e2, e12 = el e2 (= el/\ e2).
We have the following multiplication table for Cl 2
el
e2
e12
el e2 e12
1 e12 e2
-e12 1 -el
-e2 el -1
el
e2
el2
1
e2 - bel
e12 + b
-e12 - b
1
-el - be2
-e2 - bel el - be2 -1 - b2 - 2be12
235
1.5.
x 2 = Q(X)
= Q(x) IA
for all x E V
VLCl
cp'\.
11/1
A
This definition says that all Clifford maps may be obtained from I : V
which is thereby universal.
-+
Cl
236
P. LOUNESTO
V/I(Q)
Chevalley 1954 p. 37 constructs the Clifford algebra Ci(Q) as the quotient
algebra V/I(Q) of the tensor algebra V with respect to the two-sided
ideal I( Q) generated by the elements x x - Q(x) where x E V. See also
N. Bourbaki 1959 p.139 and T. Y. Lam 1973 p.l03. The tensor algebra approach gives a proof of existence by construction - suitable for an algebraist
who is interested in rapid access to the main properties of Clifford algebras.
* * * *
In characteristic zero we may avoid quotient structures by making the exterior algebra AV concrete as the subspace of antisymmetric tensors in V.
For example, if x,y E V, then x A y =
y - y x) E A2 V. More
generally, a simple k-vector Xl A X2 A ... A Xk is identified with
Alt(Xl X2 ... Xk)
;! E
1f
thus splitting the Clifford algebra Cl(Q) into fixed subspaces of k-vectors
Any orthogonal basis ell e2, ... , en of V gives a correspondence
Ak V c Cl( Q).
237
In the text the years 1986 and 1992 refer to the Proceedings of the Canterbury (1985) and Montpellier (1989) workshops on Clifford algebras. For
historical references see the excellent surveys of Budinich&Trautman 1988,
Bolinder 1986, 1987 and van der Waerden 1985.
R. Ablamowicz, P. Lounesto, J. Maks: Conference Report, Second Workshop
on 'Clifford Algebras and Their Applications in Mathematical Physics,'
238
P. LOUNESTO
239
240
P. LOUNESTO
241