Linear Algebra › Section D · Vector spaces › Topic D.4

Coordinate systems

Once a space has a basis, every vector in it can be described by a list of numbers, its coordinates. That list turns polynomials, matrices and planes in space into ordinary vectors in Rⁿ, where row reduction does the work.

Why a basis gives coordinates

A spanning set lets you build every vector of a space; a basis lets you build each one in exactly one way. That uniqueness is what makes a coordinate system possible. On ordinary graph paper, the point (3, 2) means "3 steps along the first axis and 2 along the second". Replace the two axes with any two basis vectors and the same idea works, even when the grid is slanted. This topic makes that precise, shows how to compute coordinates, and explains why an n-dimensional space with a chosen basis behaves exactly like \(\mathbb{R}^n\).

The unique representation theorem

Let \(\mathcal{B} = \{\mathbf{b}_1, \ldots, \mathbf{b}_n\}\) be a basis for a vector space \(V\). Then every \(\mathbf{x}\) in \(V\) can be written as \(\mathbf{x} = c_1\mathbf{b}_1 + \cdots + c_n\mathbf{b}_n\) for exactly one list of scalars. Existence holds because \(\mathcal{B}\) spans \(V\). For uniqueness, suppose \(\mathbf{x}\) also equals \(d_1\mathbf{b}_1 + \cdots + d_n\mathbf{b}_n\). Subtracting the two expressions gives

\mathbf{0} = (c_1 - d_1)\mathbf{b}_1 + \cdots + (c_n - d_n)\mathbf{b}_n

Because \(\mathcal{B}\) is independent, every weight \(c_i - d_i\) is 0, so the two lists are the same.

Coordinates and the coordinate vector

The weights \(c_1, \ldots, c_n\) are the coordinates of \(\mathbf{x}\) relative to \(\mathcal{B}\), or the \(\mathcal{B}\)-coordinates of \(\mathbf{x}\). Stacked as a column they form the coordinate vector, written \([\mathbf{x}]_{\mathcal{B}}\), which lives in \(\mathbb{R}^n\). The order of the basis matters: listing the same vectors in a different order permutes the coordinates. Standard coordinates in \(\mathbb{R}^n\) are simply coordinates relative to the standard basis \(\{\mathbf{e}_1, \ldots, \mathbf{e}_n\}\), which is why the coordinate vector of \((4, 9)\) in that basis is \((4, 9)\) itself.

Worked example: coordinates in the plane

Let \(\mathcal{B} = \{\mathbf{b}_1, \mathbf{b}_2\}\) with \(\mathbf{b}_1 = (2, 1)\) and \(\mathbf{b}_2 = (-1, 3)\). Find \([\mathbf{x}]_{\mathcal{B}}\) for \(\mathbf{x} = (4, 9)\). 1. Write \(c_1(2, 1) + c_2(-1, 3) = (4, 9)\), which is the system \(2c_1 - c_2 = 4\) and \(c_1 + 3c_2 = 9\). 2. From the first equation \(c_2 = 2c_1 - 4\); substitute: \(c_1 + 6c_1 - 12 = 9\), so \(c_1 = 3\) and \(c_2 = 2\). 3. So \([\mathbf{x}]_{\mathcal{B}} = (3, 2)\). 4. Check: \(3(2, 1) + 2(-1, 3)\) is \((6 - 2, 3 + 6) = (4, 9)\). On a grid drawn along \(\mathbf{b}_1\) and \(\mathbf{b}_2\), the point sits 3 steps along one and 2 along the other.

The change-of-coordinates matrix

In \(\mathbb{R}^n\), put the basis vectors into the columns of a matrix \(P_{\mathcal{B}} = [\mathbf{b}_1 \; \cdots \; \mathbf{b}_n]\). The equation \(\mathbf{x} = c_1\mathbf{b}_1 + \cdots + c_n\mathbf{b}_n\) is the same as

\mathbf{x} = P_{\mathcal{B}}[\mathbf{x}]_{\mathcal{B}}

so \(P_{\mathcal{B}}\) converts \(\mathcal{B}\)-coordinates into standard coordinates. Its columns form a basis, so it is invertible, and \([\mathbf{x}]_{\mathcal{B}} = P_{\mathcal{B}}^{-1}\mathbf{x}\) goes the other way. In practice you rarely compute the inverse: you solve \(P_{\mathcal{B}}\mathbf{c} = \mathbf{x}\) by row reduction, as in the last example, where \(P_{\mathcal{B}}\) has rows (2, -1) and (1, 3).

The coordinate mapping

The rule \(\mathbf{x} \mapsto [\mathbf{x}]_{\mathcal{B}}\) is called the coordinate mapping from \(V\) to \(\mathbb{R}^n\). It is linear: the coordinates of \(\mathbf{u} + \mathbf{w}\) are the coordinates of \(\mathbf{u}\) plus those of \(\mathbf{w}\), and the coordinates of \(c\mathbf{u}\) are \(c\) times those of \(\mathbf{u}\). It is one-to-one, by uniqueness, and onto, since every list of weights builds some vector. So it is an isomorphism in the sense of "Null spaces, column spaces, and kernels". Every vector-space calculation in \(V\) can be done on coordinate vectors in \(\mathbb{R}^n\), and translated back. In particular, a set in \(V\) is independent exactly when its coordinate vectors are.

Worked example: coordinates of a polynomial

In \(\mathbb{P}_2\), take the basis \(\mathcal{B} = \{1, \; 1 + t, \; 1 + t + t^2\}\). Find the coordinates of \(p(t) = 3 - 2t + 5t^2\). 1. Write

c_1 \cdot 1 + c_2(1 + t) + c_3(1 + t + t^2) = 3 - 2t + 5t^2

2. Match coefficients: \(t^2\) gives \(c_3 = 5\); \(t\) gives \(c_2 + c_3 = -2\), so \(c_2 = -7\); the constants give \(c_1 + c_2 + c_3 = 3\), so \(c_1 = 5\). 3. So \([p]_{\mathcal{B}} = (5, -7, 5)\). 4. Check:

5 - 7(1 + t) + 5(1 + t + t^2) = 3 - 2t + 5t^2

In the standard basis \(\{1, t, t^2\}\), the coordinates would simply be the coefficients \((3, -2, 5)\).

Testing polynomials with coordinates

Are \(p_1 = 1 + 2t\), \(p_2 = t - t^2\) and \(p_3 = 2 + 5t - t^2\) independent in \(\mathbb{P}_2\)? Translate them with the standard basis: their coordinate vectors are \((1, 2, 0)\), \((0, 1, -1)\) and \((2, 5, -1)\). Row reducing the matrix with these columns gives only two pivots, so the vectors are dependent, and so are the polynomials. The reduction even supplies the relation: \(p_3 = 2p_1 + p_2\). Check: \(2(1 + 2t) + (t - t^2)\) is \(2 + 5t - t^2\). A question about polynomials became a routine matrix calculation.

Coordinates on a plane in space

A subspace with a basis gets its own coordinates too. Let \(H\) be the plane in \(\mathbb{R}^3\) spanned by \(\mathbf{v}_1 = (1, 0, 2)\) and \(\mathbf{v}_2 = (0, 1, -1)\), which are independent, so they form a basis for \(H\). The vector \(\mathbf{x} = (3, -2, 8)\) lies in \(H\), because \(3\mathbf{v}_1 - 2\mathbf{v}_2\) is \((3, -2, 6 + 2) = (3, -2, 8)\). So its coordinate vector relative to \(\{\mathbf{v}_1, \mathbf{v}_2\}\) is \((3, -2)\), with only two entries. The coordinate mapping identifies the plane \(H\) with \(\mathbb{R}^2\), even though \(H\) is not \(\mathbb{R}^2\) itself.

Common mistakes to avoid

Confusing \(\mathbf{x}\) with \([\mathbf{x}]_{\mathcal{B}}\) is the main trap: \((3, 2)\) and \((4, 9)\) above describe the same vector in two coordinate systems. Using \(P_{\mathcal{B}}\) in the wrong direction is the next: \(P_{\mathcal{B}}\) turns \(\mathcal{B}\)-coordinates into \(\mathbf{x}\), and solving \(P_{\mathcal{B}}\mathbf{c} = \mathbf{x}\) goes back. Reordering the basis silently reorders the coordinates. For polynomials, keep the coefficients in the same order as the basis, and include zeros for missing powers.

Key takeaways

A basis gives each vector a unique list of weights, its coordinate vector \([\mathbf{x}]_{\mathcal{B}}\) in \(\mathbb{R}^n\). In \(\mathbb{R}^n\), \(\mathbf{x} = P_{\mathcal{B}}[\mathbf{x}]_{\mathcal{B}}\), where \(P_{\mathcal{B}}\) has the basis vectors as columns; find coordinates by solving \(P_{\mathcal{B}}\mathbf{c} = \mathbf{x}\). The coordinate mapping is a one-to-one linear map onto \(\mathbb{R}^n\), an isomorphism, so questions about polynomials or subspaces become matrix questions. A plane in \(\mathbb{R}^3\) with a basis of two vectors has coordinates in \(\mathbb{R}^2\).

Flashcards

Flashcard

Unique representation theorem

Each x is c₁b₁ + … + cₙbₙ in exactly one way

Spanning gives existence, independence gives uniqueness.

Test your understanding

Question 1 of 5

B = {(1, 1), (1, -1)}. What is [x]_B for x = (5, 1)?

Question 2 of 5

P_B has the vectors of a basis B of Rⁿ as its columns. Which equation is correct?

Question 3 of 5

In P₂ with the standard basis {1, t, t²}, what is the coordinate vector of 4 - t²?

Question 4 of 5

Why is the coordinate mapping x ↦ [x]_B one-to-one?

Question 5 of 5

Polynomials p₁, p₂, p₃ in P₂ have standard coordinate vectors (1, 2, 0), (0, 1, -1) and (2, 5, -1). What follows?

← Previous topicNext topic →