Académique Documents
Professionnel Documents
Culture Documents
2. Vector Subspaces
Contents
Definition of Vector Subspace Examples of Vector Subspaces Intersections of Vector Subspaces Unions of Vector Subspaces Sums of Vector Subspaces Distributivity It is usually easiest to study an object from the outside rather than the inside; for example, we can see the shape of a object in three-dimensional space most easily by moving around and looking at its surface from different positions in the space. Similarly, it is usually easiest to study a vector space "from the outside" by considering it to be part of a larger vector space, which leads to the idea of a vector subspace. A vector subspace is a vector space that is embedded in a larger vector space, or equivalently a vector subspace is a subset of a vector space that is itself a vector space. A subset of a vector space automatically satisfies most of the vector space axioms, so the extra conditions that are required to ensure that it is a vector subspace are essentially the closure conditions to stop vectors "escaping" from the vector subspace into the larger vector space. Examples of vector subspaces of real three-dimensional Euclidean space are any straight line or plane through the origin, as illustrated in the following two figures.
If the subspace were not flat or did not include the origin then it would be possible for vectors constructed by applying the vector space operations of addition and scalar multiplication to "escape" from the subspace, which is not allowed because the subspace must be closed. The formal definition generalizes this geometrical notion.
Other nonlinear constraints do not give vector subspaces either! As another example, U = { (x, y) | y = x2 } R2 is not a vector subspace of R2 because it is not closed. Here are two examples (i.e. simple special cases) that show that it is not closed under addition: ( 1, 1 ) U and ( 2, 4 ) U but ( 1, 1 ) + ( 2, 4 ) = ( 3, 5 ) is not in U because 5 32; ( 1, 1 ) U and ( 1, 1 ) U but ( 1, 1 ) + ( 1, 1 ) = ( 0, 2 ) is not in U because 2 02. The following figure illustrates both these failures of closure.
U is also not closed under scalar multiplication, e.g. ( 1, 1 ) U but 2 ( 1, 1 ) = ( 2, 2 ) is not in U because 2 ( 2 )2. However, any single counter-example is sufficient to prove that U is not a vector subspace.
Vector subspaces of R2 In fact, the only vector subspaces of the vector space R2 are (a) the set containing only the zero vector, i.e. U = { 0 R2 }, (b) any straight line through the origin, i.e. U = { ( x, x ) R2 | x R } for any fixed , R, and (c) the whole plane R2, i.e. U = R2.
We can formalize this as follows. Proposition: Intersections of Vector Subspaces are Vector Subspaces Let V be a vector space and let S and T be vector subspaces of V. Then their intersection S T is also a vector subspace of V. Proof
Let U = S T. We need to prove that the vector subspace conditions are satisfied for U. Zero-vector inclusion: 0 S and 0 T since S and T are vector subspaces, so 0 U; Closure under addition: u1, u2 U => (u1, u2 S and u1, u2 T) => (( u1 + u2 ) S and ( u1 + u2 ) T) since S and T are vector subspaces, so ( u1 + u2 ) U for all u1, u2 U; Closure under scalar multiplication: u U => (u S and u T) => (k u S and k u T) since S and T are vector subspaces, so k u U for all u U, k K. Example Let V = R3, S = { ( x, y, z ) R3 | x = y } and T = { ( x, y, z ) R3 | y = z }. S = { ( x, x, z ) R3 } is a vector subspace of V, since ( 0, 0, 0 ) S; (( x1, x1, z1 ) S and ( x2, x2, z2 ) S) => ( x1 + x2, x1 + x2, z1 + z2 ) S; ( x, x, z ) S => ( k x, k x, k z ) S. Similarly T = { ( x, y, y ) R3 } is a vector subspace of V. Therefore U = S T is a vector subspace of V, by the proposition on intersections of subspaces. It is easy to check this assertion explicitly. U = { ( x, y, z ) R3 | x = y = z } = { (x, x, x) | x R } is a vector subspace of V, since ( 0, 0, 0 ) U; (( x1, x1, z1 ) U and in( x2, x2, x2, U )) => in( x1 + x2, x1 + x2, x1 + x2, U ); in( x, x, x, U ) => in( k x, k x, k x, U ).
Proposition: a Sum of Vector Subspaces is a Vector Subspace If S, T are vector subspaces of a vector space V then S + T is also a vector subspace of V. Proof Let U = S + T. We need to prove that the vector subspace conditions are satisfied for U. Zero-vector inclusion: 0 S and 0 T since S and T are vector subspaces, so ( 0 = 0 + 0 ) ( U = S + T ). Closure under addition: u1, u2 U => u1 = s1 + t1, u2 = s2 + t2 for some s1, s2 S, t1, t2 T. Hence u1 + u2 = (s1 + t1) + (s2 + t2) = (s1 + s2) + (t1 + t2). But ( s1 + s2 ) S and ( t1 + t2 ) T since S and T are vector subspaces, so ( u1 + u2 ) U for all u1, u2 U. Closure under scalar multiplication: u U => (u = s + t for some s S and t T) => k u = k s + k t. But k s S and k t T since S and T are vector subspaces, so k u U for all u U, k K. Example Let V = R2, S = { (x, 0) | x R }, T = { (0, y) | y R } (i.e. the two coordinate axes as before). Then S + T = { (x, 0) + (0, y) | x R, y R } = { (x, y) | x, y R } = R2 = V. The proposition asserts that S + T = { (x, y) | x, y R } is a vector subspace of V = R2, which is clearly true because S + T = R2. Note also that S T = { (0, 0) } is a (trivial) vector subspace.
Distributivity*
An operation is distributive relative to a rule of combination if performing the operation on the combination is equivalent to performing the operation on each element of the combination and then combining the results. For example, multiplication distributes over addition in a field, i.e. a . ( b + c ) = ( a . b ) + ( a . c ), and scalar multiplication distributes over both vector addition and scalar addition in a vector space; similarly intersection distributes over union in set theory A ( B C ) = ( A B ) ( A C ). We have just considered intersections and sums of vector subspaces, which raises the question: does intersection distribute over addition of vector subspaces? The answer is no, as we now prove by giving a simple counter-example, continuing the above example. We recall that V = R2, S = { (x, 0) | x R }, T = { (0, y) | y R }. Consider the line L = { (x, x) | x R }, where S, T and L are all vector subspaces of R2. Then L S = { 0 } and L T = { 0 }, so ( L S ) + ( L T ) = { 0 }. However, L ( S + T ) = L R2 = L. Hence L ( S + T ) ( L S ) + ( L T ) and intersection does not distribute over addition of vector subspaces.