Page 31 - Curvature and Homology
P. 31

1.4.  DIFFERENTIAL FORMS               13

        dccomposabb  p-vectors  eil  A ... A eiB.  Since,  by  assumption,  A (V)
        is spanned by  1 and such products, it follows that  A (V) is spanned by
        the  elements  eil  A ... A  ei9  where  (i,,  ..., 6) is  a  subset  of  the  set
        (1, ..., n) arranged in increasing order.  But there are exactly 2n  subsets
        of  (1, ..., n),  while  by  assumption  dim  A (V) = 2n.  These  elements
        must  therefore  be linearly independent.  Hence,  any element of  A (V)
        can be uniquely represented as a linear combination




        where now and in the sequel (il ... 6) implies i, < ... < &.  An element
        of the first sum is called homogeneous  of  degree p.
          It  may  be  shown  that  any  two  Grassman  algebras over  the  same
        vector space are isomorphic. For a realization of  A (V) in terms of  the
        'tensor  algebra' over  V the reader is referred to (I.C.2).
          The elements x,,  ..., x,  in  V are linearly independent, if  and only if,
        their product x,  A ... A x,  in  A (V) is not zero. The proof  is an easy
        exercise in linear algebra. In particular, for the basis elements el, ..., e,
        of  V,  el  A ... A en # 0. However, any  product  of  n + 1 elements of
        V must vanish.
          All  the elements



        for  a  fixed p  span  a  linear  subspace  of  A  (V)  which  we  denote  by
        Ap(V).  This  subspace. is  evidently independent of  the choice of  base.
        An element of  AP(V) is called an exterior p-vector or, simply a p-vector.
        Clearly,  A1(V) = V.  We define  AO(V) = R. As a vector space,  A (V)
       is then  the  direct  sum of  the subspaces  Ap(V), 0 5 p 5 n.
          Let  W be  the  subspace  of  V spanned  by y,,  ..., y,   E V.  This gives
        rise to a p-vector  q = yl A ... A yp which is unique up to a constant
       factor as one sees from the theory  of  linear  equations.  Moreover, any
       vector y  E W has the  property  that y  A q vanishes.  The subspace  W
       also determines its orthogonal complement (relative to an inner product)
       in  V,  and this subspace in turn defines a 'unique'  (n - p)-vector.  Note
       that for eachp, the spaces Ap(V) and An-p(V) have the same dimensions.
       Any  p-vector  5 and  any (n  2 p)-vector q determine  an n-vector 5 A q
       which in terms of the basis e = el  A ...   e,  of  An(V) may be expressed
       as
                                f A 7 = (f,~)                   (1.4.3)
                                           e
       where  (5, q) E R.  It  can  be  shown  that  this  'pairing'  defines  an  iso-
       morphism  of  AP(V) with  (A~-P(~))*  1.5.1  and 1I.A).
                                         (cf.
   26   27   28   29   30   31   32   33   34   35   36