Change of Coordinates
Introduction
Abstract Given two ordered bases \(\mathcal{A}\) and \(\mathcal{B}\) for a subspace \(W\) of \(\RNrSpc{n}\), we show that there is a certain matrix which transforms coordinate vectors with respect to \(\mathcal{A}\) into coordinate vectors with respect to \(\mathcal{B}\).
Outline Let's begin with an example: We know that ‘kilometer’ and ‘mile’ are units of measurement of length. On a line L with a point designated as the origin, we may therefore use either of these two units to obtain a basis for \(L\). Converting then between kilometers and miles amounts to converting coordinate vectors with respect to the km-basis into coordinate vectors with respect to the miles basis; and we know that this conversion is handled by a single factor:
(number of km's) = 1.6 x (number of miles)
In essence this explains what happens when we want to convert between coordinate vectors with respect to two different bases in a subspace \(W\) of \(\RNrSpc{n}\). Say, we have two such bases
\[\VSpcBss{B}=(\Vect{b}_1,\dots ,\Vect{b}_r) \quad\text{and}\quad \VSpcBss{C}=(\Vect{c}_1,\dots ,\Vect{c}_r)\]
Then a vector \(\Vect{x}\) in \(W\)
can be expressed in exactly one way as a linear combination of \(\VSpcBss{B}\) and of \(\VSpcBss{C}\):
\[
\begin{array}{rcllrcl}
\Vect{x} & = & s_1 \Vect{b}_1 + \cdots + s_r\Vect{b}_r &\quad \text{so} & \CoordVect{x}{B} = (s_1,\dots ,s_r) \\
\Vect{x} & = & t_1 \Vect{c}_1 + \cdots + t_r\Vect{c}_r &\quad \text{so} & \CoordVect{x}{C} = (t_1,\dots ,t_r) \\
\end{array}
\]
To convert from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{A}\)-coordinates, use the coordinate vectors of \(\OrdVSpcBss{B}\) with respect to \(\OrdVSpcBss{C}\) to construct the matrix
\[
\CoordTrafoMtrx{C}{C}{B} \DefEq \left[\begin{array}{cccc}
\uparrow & \uparrow & & \uparrow \\
(\Vect{b}_1)_{\OrdVSpcBss{C}} & (\Vect{b}_2)_{\OrdVSpcBss{C}} & \cdots & (\Vect{b}_r)_{\OrdVSpcBss{C}} \\
\downarrow & \downarrow & & \downarrow
\end{array}\right]
\]
We then have the
coordinate conversion identity
| \(\Vect{x}_{\OrdVSpcBss{C}}\) | \(=\) | \(\CoordTrafoMtrx{C}{C}{B} \Vect{x}_{\OrdVSpcBss{B}}\) |
Such coordinate conversion matrix is always invertible, and it satisfies
\[\CoordTrafoMtrx{C}{B}{C} = \left( \CoordTrafoMtrx{C}{C}{B} \right)^{-1}\]
This follows from the
transitivity property of coordinate vector conversion:
| \(\CoordTrafoMtrx{C}{D}{B}\) | \(=\) | \(\CoordTrafoMtrx{C}{D}{C} \CoordTrafoMtrx{C}{C}{B}\) |
Here we are given three bases \(\OrdVSpcBss{B},\OrdVSpcBss{C},\OrdVSpcBss{D}\) of \(W\), and the above identity asserts that the conversion from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{D}\)-coordinates may just as well be accomplished by first converting from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{C}\)-coordinates and then further to \(\OrdVSpcBss{D}\)-coordinates.
PropositionChange coordinates matrix
Given ordered bases \(\OrdVSpcBss{B}=(\Vect{b}_1,\dots ,\Vect{b}_r)\) and \(\OrdVSpcBss{C}=(\Vect{c}_1,\dots ,\Vect{c}_r)\) of a vector space \(\VSpc{W}\), let
\[
\CoordTrafoMtrx{C}{C}{B}\ =\
\left[\begin{array}{cccc}
\uparrow & \uparrow & & \uparrow \\
(\Vect{b}_1)_{\OrdVSpcBss{C}} & (\Vect{b}_2)_{\OrdVSpcBss{C}} & \cdots & (\Vect{b}_r)_{\OrdVSpcBss{C}} \\
\downarrow & \downarrow & & \downarrow
\end{array}\right]
\]
Then, for every \(\Vect{x}\) in \(W\),
| \(\Vect{x}_{\EuScript{C}}\) | \(=\) | \(\CoordTrafoMtrx{C}{C}{B} \Vect{x}_{\OrdVSpcBss{B}}\) |
Moreover, the matrix \(\CoordTrafoMtrx{C}{C}{B}\) with this property is unique.
The matrix \(\CoordTrafoMtrx{C}{C}{B}\) is called the change of coordinates matrix or the coordinate conversion matrix from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{C}\)-coordinates.
Let us now turn to rules which apply to coordinate conversion matrices: Suppose we are given three ordered bases of a subspace \(W\) of \(\RNrSpc{n}\), say \(\OrdVSpcBss{B}\), \(\OrdVSpcBss{C}\), and \(\OrdVSpcBss{D}\). We then have associated coordinate conversion matrices
\(\CoordTrafoMtrx{C}{C}{B}\) | to convert from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{C}\)-coordinates |
\(\CoordTrafoMtrx{C}{D}{C}\) | to convert from \(\OrdVSpcBss{C}\)-coordinates to \(\OrdVSpcBss{D}\)-coordinates |
\(\CoordTrafoMtrx{C}{D}{B}\) | to convert from \(\OrdVSpcBss{B}\)-coordinates to \(\OrdVSpcBss{D}\)-coordinates |
The relationship between these matrices is given by the following:
CorollaryProperties of coordinate change
Given ordered bases \(\OrdVSpcBss{B}\), \(\OrdVSpcBss{C}\), and \(\OrdVSpcBss{D}\) of a subvector space \(\VSpc{W}\) of \(\RNrSpc{n}\), the associated coordinate conversion matrices are related by
| \(\CoordTrafoMtrx{C}{D}{B}\) | \(=\) | \(\CoordTrafoMtrx{C}{D}{C} \CoordTrafoMtrx{C}{C}{B}\) |
CorollaryReversing coordinate change
If \(\OrdVSpcBss{B}\) and \(\OrdVSpcBss{C}\) are ordered bases of a subvector space \(\VSpc{W}\) of \(\RNrSpc{n}\), then
| \(\CoordTrafoMtrx{C}{B}{C}\) | \(=\) | \(\left(\CoordTrafoMtrx{C}{C}{B}\right)^{-1}\) |
http://emath.ualberta.ca/LinrAlgbra/LinAlgInRn.xml/pages/Sec_GramSchmidtOrthonormalizationChange of Coordinateshttp://emath.ualberta.ca/LinrAlgbra/LinAlgInRn.xml/pages/Sec_OrientationGeneral