19. Subspaces

19. Subspaces

In this video we will explain in more detail what a subspace is. Recall from the last video that subspaces are generalisations of lines/planes and the set of solutions of a system of simultaneous equations in n variables forms a subspace of 𝐑n .

Linear subspaces

Definition:

A subset V𝐑n is a linear subspace if:

  • for all v,wV , v+wV (closed under addition),

  • for all vV and λ𝐑 , λvV (closed under rescaling).

Note that nonempty linear subspaces always contain the origin: if vV then 0vV by the second axiom, and 0v=0 is the zero-vector (the origin). Note that there are systems of simultaneous equations for which (0,,0) is not a solution; likewise, there are lines and planes that do not pass through the origin. So for our purposes, we need something more general than linear subspaces.

Affine subspaces

Definition:

V𝐑n is called an affine subspace if there exist a vector w𝐑n and a linear subspace U𝐑n such that V=w+U:={w+u:uU} ; in other words, V is obtained by translating U by the vector w .

Line V obtained by translating line U by the vector W
Lemma:

Given a system of simultaneous equations in matrix form Av=b , the set of solutions v=(v1vn) form an affine subspace of 𝐑n , (n is the number of variables). It is a linear subspace if and only if b=0 .

First, let's assume b=0 . We'll prove that the set of solutions satisfies the axioms for being a linear subspace. If v,v are solutions then Av=Av=0 , so A(v+v)=0+0=0 , and if λ𝐑 then Aλv=λAv=λ0=0 . Therefore both v+v and λv are solutions and the set of solutions is a linear subspace. It's also nonempty because v=0 is always a solution to Av=0 .

Now suppose b0 . If there are no solutions then the set of solutions is empty, and the empty set is an empty subspace. So suppose there is at least one solution w . Let U be the set of solutions to Au=0 . Let S be the set of solutions to Av=b . We will show that S=w+U :

  • We first prove that w+US . This is because if uU then A(w+u)=Aw+Au=b+0=b , so w+uS .

  • Then we prove that Sw+U . This is because Av=b implies A(v-w)=Av-Aw=b-b=0 , so v-wU and ww+U .

More properties

You'll see a lot more about subspaces in future courses on linear algebra, but I'll just give you a couple more nice facts you can see from the definitions and which are related to what we've been doing.

Lemma:

If V,W𝐑n are linear subspaces then the intersection VW is also a linear subspace.

We'll check that the two axioms hold.

Suppose a,bVW . Then aV and bV , so a+bV because V is a linear subspace. Similarly aW,bW implies a+bW . Therefore a+bVW .

Similarly, aVW implies aV and aW , so λaV and λaW , so λaVW .

Another fact which is good to know but which I won't prove is:

Lemma:

A nonempty affine subspace is linear if and only if it contains the origin.

You now have the beginnings of the language you need for talking about lines, planes and their generalisations in higher dimensions.