Chapter II.
Affine Spaces
II.1. Spaces
Note. In this section, we define an affine space on a set X of points and a vector
space T. In particular, we use affine spaces to define a tangent space to X at
point x. In Section VII.2 we define manifolds on affine spaces by mapping open
sets of the manifold (taken as a Hausdorff topological space) into the affine space.
Ultimately, we think of a manifold as pieces of the affine space which are “bent
and pasted together” (like paper mache). We also consider subspaces of an affine
space and translations of subspaces. We conclude with a discussion of coordinates
and “charts” on an affine space.
Definition II.1.01. An affine space with vector space T is a nonempty set X of
points and a vector valued map d : X × X ! T called a difference function, such
that for all x, y, z 2 X:
(A i) d(x, y) + d(y, z) = d(x, z),
(A ii) the restricted map dx = d|{x}×X : {x} × X ! T defined as mapping
(x, y) 7! d(x, y) is a bijection.
We denote both the set and the affine space as X.
II.1. Spaces 2
Note. We want to think of a vector as an arrow between two points (the “head”
and “tail” of the vector). So d(x, y) is the vector from point x to point y. Then
we see that (A i) is just the usual “parallelogram property” of the addition of
vectors. Property (A ii) can be thought of in terms of putting vectors in a “standard
position” with their tails all at point x. In fact, we show below in Lemma II.1.A
that d(x, x) = 0 2 T and so it could be convenient to thank point x as the origin
for all vectors with their tails at x. We will see in Exercise II.1.1 that all such
vectors in T of the form d(x, y) form a vector space. Notice that an affine space
consists of a set X, vector space T, and difference function d; however, an affine
space (per se) is not itself a vector space.
Examples. With X = R (as a point set) and T = R1 (as a vector space), we can
define d(x, y) = y − x. We can similarly let X = Rn (as a point set), T = Rn (as
a vector space) and define
d((x1, x2, . . . , xn), (y1, y2, . . . , yn)) = (y1 − x1, y2 − x2, . . . , yn − xn).
These are examples of affine spaces where d(x, y) is the vector in Rn from point x
to point y (with the usual arrow with magnitude-an-direction interpretation, valid
in Rn). Notice that in these examples, (A i) simply expresses the summation of
vectors in Rn and (A ii) sets up a bijection where x is mapped to the 0 vector (as
we now show).
Lemma II.1.A. In an affine space with difference function d we have
(a) d(x, x) = 0 for all x 2 X, and
(b) d(x, y) = −d(y, x) for all x, y 2 X.
II.1. Spaces 3
Note II.1.A. Notationally, in an affine space if d(x, y) = t then we write y = x+t;
think that we start at point x and go “directed distance” t ending at point y. If
V T then we write x + V = {d(x, y) = t | t 2 V } = {x + t | t 2 V }.
Basis, Dimension, Rank
A basis for a subspace S of Rn is a set of vectors in S that
1. span S
2. are linearly independent
An example of a basis is fe1; e2; : : : ; eng, which spans Rn and is linearly independent.
This is called the standard basis of Rn.
Example:
Find a basis for row(A), col(A) and null(A), where
A = any 4x4 matrix
In general, to _nd a basis for row(A), col(A) and null(A), do the following:
1. Find the reduced row echelon form R of A.
2. The nonzero row vectors of R form a basis for row(A).
3. The column vectors of A corresponding to columns of R with leading 1's (the
pivot
columns) form a basis for col(A).
4. Solve for the leading variables of Rx = 0 in terms of the free variables. Write the
solution x as a column vector expanded into a linear combination with the free
vari-
ables being the coe_cients. The vectors in terms of which the linear combination is
written form a basis for null(A).
Although the same subspace can have many di_erent bases, the number of vectors
in
each must be the same.
Theorem: Let S be a subspace of Rn. Then the number of vectors in any basis of S
is the
same and is called the dimension of S.
For the row and column space of a matrix the following property holds.
Theorem: The row and column space of a matrix A have the same dimension.
The rank of a matrix A, denoted rank(A), is the dimension of its row and column
spaces.
The nullity of a matrix A, denoted nullity(A), is the dimension of its null space.
It is easy to see that rank(AT ) = rank(A). The rank and the nullity of a matrix have
the following relation.
The Rank Theorem: If A is an m _ n matrix, then
rank(A) + nullity(A) = n:
1
The fundamental theorem for invertible matrices can be now extended using some
of
these new notions.
The Fundamental Theorem for Invertible Matrices: Let A be an n_n matrix. The
following are equivalent.
(a) A is invertible
(b) Ax = b has a unique solution for every b in Rn.
(c) Ax = 0 has only the trivial solution.
(d) The reduced row echelon form of A is In.
(e) A is a product of elementary matrices.
(f) rank(A) = n.
(g) nullity(A) = 0.
(h) The column vectors of A are linearly independent.
(i) The column vectors of A span Rn.
(j) The column vectors of A form a basis for Rn.
(k) The row vectors of A are linearly independent.
(l) The row vectors of A span Rn.
(m) The row vectors of A form a basis for Rn.
An application of the Rank Theorem and the Fundamental theorem gives the
following.
Theorem: Let A be an m _ n matrix. Then:
_ (a) rank(ATA) = rank(A).
_ (b) The n _ n matrix ATA is invertible if and only if rank(A) = n.
