Analysis for Applications 13 1.5 Vector spaces A vector space over an algebraic eld K is a set X together with two algebraic operations: vector addition and scalar multiplication. Elements x 2 X are called vectors; elements 2 K are called scalars. For our purposes, the eld K will always be either IR or C. Vector addition is a function, denoted \+" which maps X X into X and satises for every x; y; and z in X : 1. x + y = y + x (commutativity), 2. x + (y + z) = (x + y) + z (associativity). Furthermore, there exists a vector in X , denoted 0, and for every vector x 2 X there exists a vector ,x such that x + 0 = x; and x + (,x) = 0: Scalar multiplication is a function mapping K X into X , which satises for all x; y 2 X and all ; 2 K , 1. (x) = ( )x, 2. 1x = x (1 denotes the multiplicative identity in K ), 3. (x + y) = x + y, 4. ( + )x = x + x. A vector space dened with respect to real scalars (that is, K = IR) is called a real vector space; a vector space dened with respect to complex scalars is called a complex vector space. Examples (check that these satisfy the denitions) IR and C are themselves vector spaces, as are IRn and Cn with the usual denitions of vector addition and scalar multiplication. The spaces l , 1 p 1 are vector spaces with vector addition dened by x + y = (x1 + y1 ; x2 + y2 ; : : :) for x = (x ); y = (y ) and scalar multiplication dened by x = (x1 ; x2 ; : : :): p j j D. Dobson 14 C [a; b] is a vector space with (f + g)(x) = f (x) + g(x) for x 2 [a; b]; and (f )(x) = f (x) for x 2 [a; b] for any f; g 2 C [a; b]. It is important to note that both vector addition and scalar multiplication must dene values in the vector space, that is, x 2 X and x + y 2 X for all scalars and all x; y 2 X . You may recall that in metric spaces, we dened a subspace to be any subset of elements of the original space, paired with the induced metric. In vector spaces, we will not refer to arbitrary subsets as subspaces. To be a called a subspace, we require that the set is \algebraically closed" under the induced operations of vector addition and scalar multiplication. More denitions. A subspace of a vector space X is a subset Y X such that for all y1; y2 2 Y , and all scalars we have y1 2 Y and y1 + y2 2 Y . For any subset M X , the set of all (nite!) linear combinations of vectors in M 1 x1 + 2 x2 + + x is called the span of M , written span M . Note that span M is a subspace of X . The nite set of vectors x1; x2 ; : : : ; x 2 X is called linearly independent n if the only solution to the equation n n 1 x1 + 2 x2 + + x = 0 n n is 1 = 2 = = = 0. If the vectors x1 ; x2 ; : : : ; x are not linearly independent, they are called linearly dependent. An arbitrary subset of vectors M X is called linearly independent if every nonempty nite subset of M is linearly independent. M is called linearly dependent if it is not linearly independent. A vector space X is said to be nite dimensional if there exists a natural number N such that X contains a linearly independent set of N vectors, but any set of N + 1 or more vectors in X is linearly dependent. N is called the dimension of X . If X is not nite dimensional, it is called innite dimensional. n n Analysis for Applications 15 Examples. IRn and Cn are nite dimensional, with dimension n. l and p C [a; b] are innite dimensional. Denition. A linearly independent subset called a Hamel basis for X . B X which spans X is Theorem 1.5.1 Every nonempty vector space X has a Hamel basis. Proof. A proof can be found in most functional analysis texts but is outside the scope of our discussion. The innite dimensional case requires Zorn's lemma. Note: If B is a basis, then every element x 2 X has a unique representation as a (nite) linear combination of elements of B . 1.6 Banach spaces A normed vector space is a vector space together with a norm. A norm on a vector space X is a real-valued function dened on X , whose value at x 2 X is denoted kxk and which satises 1. kxk 0, 2. kxk = 0 if and only if x = 0, 3. kxk = jjkxk for all scalars , 4. kx + yk kxk + kyk. Note that a norm on X denes the metric d(x; y ) = kx , y k; called the metric induced by the norm. You should check for yourself that this is always a metric. Other metrics can of course be dened in terms of the norm, but they are generally less important in what follows. Denition. A Banach space is a normed vector space which is complete in the metric induced by the norm. D. Dobson 16 Examples. IRn and C are Banach spaces with the norm as we dened in our n discussion of linear algebra. We have already dened the following spaces as metric spaces. To show that they are Banach spaces requires that we rst check that the norm as dened is indeed a norm, and second that the space is complete in the metric induced by the norm. We will study closely related spaces in more detail soon, so for now we put o the details. The spaces l , 1 p 1 are Banach spaces with the norms p 01 11 X kxk = @ jx j A ; p < 1 =p p p j j and =1 kxk1 = sup jx j; p = 1: j j The space C [a; b] is a Banach space with the norm kf k1 = max 2[ ] jf (t)j: t a;b 1.7 Compactness A metric space X is compact if every sequence in X has a convergent subsequence. A subset M X is called compact if every sequence in M has a convergent subsequence whose limit is in M . It should be noted that compactness as we have dened it is known as sequential compactness in general topological spaces. In metric spaces this denition is equivalent to other standard notions of compactness. It is usually proved (or dened) in advanced calculus courses that subsets of IRn are compact if and only if they are closed and bounded. The same fact holds in normed vector spaces provided the space is nite dimensional: Theorem 1.7.1 If X is a nite dimensional normed vector space then each subset M X is compact if and only if M is closed and bounded.