We describe vectors with coordinates on a grid, and we need a way to translate those coordinates when we switch to a different grid.
1
Translating Between a Custom Basis and the Standard Grid
To translate a vector's coordinates between our standard (x,y) grid and a custom Basis B built from b1=[21] and b2=[−12], we put the new basis vectors into the columns of a matrix B.
Translating Between a Custom Basis and the Standard Grid
Text solution
Multiply by B to convert from Basis B to the standard gridPut the basis vectors b1 and b2 into the columns of B and multiply by [31]B to find where 3 steps of b1 and 1 step of b2 land in our standard coordinates.
Row-reduce or multiply by B−1 to convert from the standard grid to Basis BTo find the B-coordinates of the standard vector [34]ours, solve B[xy]B=[34]ours by row-reducing the augmented matrix or multiplying by B−1.
So [34]ours and [21]B describe the exact same arrow on the board, just counted on two different grids.
2
Connecting Two Custom Bases Through the Standard Grid
Multiplying by B or B−1 connects a custom basis to the standard grid, but now suppose we have a second custom grid, Basis C with c1=[11] and c2=[12], and we want a single matrix that translates a vector's coordinates directly from Basis B to Basis C.
Connecting Two Custom Bases Through the Standard Grid
Instead of inventing a new rule, we go through the standard grid in the middle: first multiply [v]B by B to get standard coordinates, and then multiply that result by C−1 to get the coordinates in Basis C.
[v]C=PC←Bchange-of-basis matrixC−1B[v]B
The subscript arrow in PC←B points from right to left to match the order of matrix multiplication: the input vector [v]B goes in on the right, and the translated output [v]C comes out on the left.
For any arrow in the plane, the arrow itself stays in the same place while its coordinate numbers change across Basis B, the standard grid, and Basis C. Multiplying [v]B by B and then C−1 gives the exact C-coordinates of that same arrow.
The same arrow in space has a different step count on each grid, and the two matrix multiplications translate its coordinates from B's grid through our standard grid into C's grid.
3
Finding the Change-of-Basis Matrix by Row Reduction
To translate [v]B=[32]B into Basis C, we need the entries of PC←B=C−1B. Rather than finding C−1 and multiplying two matrices, remember that multiplying C−1 by B=[b1b2] means multiplying C−1 by one column of B at a time:
Because the columns of PC←B are just b1 and b2 written in C-coordinates, we can find both columns at once with a single row reduction.
Finding the Change-of-Basis Matrix by Row Reduction
[C∣B]PC←B=[111221−12]=?
Text solution
Set up the augmented matrix [C∣B]Put the target basis C on the left and the starting basis B on the right to solve for [b1]C and [b2]C at the same time.
[C∣B]=[111221−12]
Row-reduce the left side to the identity matrix ISubtract Row 1 from Row 2 to turn the bottom-left 1 into 0, then subtract the new Row 2 from Row 1 to turn the top-right 1 into 0, turning the right side into PC←B.
So [13]C in Basis C and [32]B in Basis B both describe the exact same arrow [47]ours in our standard grid.
If a problem already gives b1 and b2 as combinations of c1 and c2 (such as b1=3c1−c2 and b2=−4c1+3c2), you can skip the row reduction completely and put those numbers straight into the first and second columns of PC←B.
4
Going Back the Other Way with the Inverse Matrix
Now suppose we want to translate in the opposite direction, from Basis C back to Basis B, using PB←C. Starting from scratch, we could row-reduce [B∣C]→[I∣PB←C], but since we already have PC←B, we can just invert it.
[acbd]−1=ad−bc1[d−c−ba]
Going Back the Other Way with the Inverse Matrix
Text solution
Invert PC←B to reverse the translation directionUse the 2×2 inverse rule on PC←B=[3−1−43] to find PB←C.
This brings us back to the original coordinate vector [32]B we started with.
5
Simplifying Transformations by Switching Bases
So far we have only changed the coordinates of a fixed vector, but switching grids also helps us do geometric transformations that are hard to calculate in our standard (x,y) grid, such as reflecting the vector [34]ours across the tilted line through [21].
Simplifying Transformations by Switching Bases
Text solution
Translate the input vector from the standard grid to Basis BMultiply [34]ours by B−1 to split it into components along the line (b1) and perpendicular to the line (b2).
B−1[34]ours=[21]B
Reflect the vector in Basis BMultiply by the reflection matrix D=[100−1] in Basis B to keep the b1 component along the line unchanged and flip the sign of the perpendicular b2 component.
D[21]B=[100−1][21]B=[2−1]B
Translate the reflected vector back to the standard gridMultiply by B to convert the reflected B-coordinates back into standard coordinates.
So reflecting [34]ours across the line through [21] lands on [50]ours. Putting those three steps together from right to left connects the standard transformation matrix A and the Basis-B transformation matrix D:
Reflecting across the tilted line through b1=[21] mixes a vector's standard x and y coordinates together, but in Basis B every vector's b1 component along the line never changes and its b2 component perpendicular to the line simply flips sign.
Reflecting across a tilted line changes both regular coordinates at once, but in the tilted grid aligned with the line it just keeps the step along the line and flips the step across it.
Key concepts
Every change-of-basis problem comes down to one question: which grid are you starting in, and which grid are you going to?
When one of the grids is already the standard (x,y) grid, you don't need a two-basis matrix PC←B. You only need the basis matrix B=[b1b2] whose columns are the basis vectors. Multiplying a B-vector by B gives the vector in the standard grid, like B[31]B=[21−12][31]B=[55]ours, and multiplying a standard vector by B−1 (or row-reducing [B∣v]) translates it back into Basis B, like B−1[34]ours=[21]B.
When translating between two custom bases B and C, go through the standard grid with PC←B=C−1B so that [v]C=PC←B[v]B. If B and C are given as standard vectors, put the target basis C on the left (matching the arrow in C←B) and row-reduce [C∣B]→[I∣PC←B], as in [111221−12]→[10013−1−43]. Skip row reduction whenever the B-vectors are already written as combinations of the C-vectors: if b1=3c1−c2 and b2=−4c1+3c2, those coefficients are already [b1]C and [b2]C, so they go straight into the columns of PC←B=[3−1−43]. And to go back the other way, invert the matrix you already have: PB←C=(PC←B)−1.
To switch bases for a transformation, match the rightmost matrix in the product to the basis of the input vector. To apply a simple Basis-B matrix like D=[100−1] to vectors in the standard grid, translate from standard to B first with A=BDB−1; to see what a standard matrix A looks like in Basis B, translate from B to standard first with D=B−1AB.
6
Practice
1
Basis B consists of b1=[12] and b2=[35]. Find the standard coordinates of [u]B=[2−1]B, and find the B-coordinates of the standard vector w=[59].
Put b1 and b2 into the columns of B=[1235] (be careful not to put them into the rows as [1325], which would give [01]).
Which vector is already a recipe of b1 and b2 steps (multiply by B), and which one starts in the standard grid and needs to be converted into Basis B (multiply by B−1)?
Put b1 and b2 into the columns of B=[1235] (be careful not to put them into the rows as [1325], which would give [01]) and multiply by [u]B:
u=B[u]B=[1235][2−1]B=[−1−1]
To find [w]B, invert B using the determinant 1(5)−3(2)=−1 (watch the negative sign on the determinant so you don't miss flipping the signs of B−1 and get [−2−1]B, and don't multiply w by B to get [3255]):
[w]B=B−1w=[−523−1][59]=[21]B
2
Basis B has vectors b1=[31],b2=[−12] and Basis C has vectors c1=[11],c2=[23]. Row-reduce [C∣B] to find the change-of-basis matrix PC←B, then use it to find [v]C for [v]B=[12]B.
Put the target basis C on the left and B on the right (setting up [B∣C] in the order the bases are mentioned in the problem instead of [C∣B] would find PB←C=71[3277] and give 71[1716]).When turning the top-right 2 into 0 with R1−2R2→R1, watch the signs in 3−2(−2)=7 and −1−2(3)=−7 (subtracting the wrong sign gives [−1−253] and [94]C).
Check the arrow in PC←B: which basis matrix goes on the left side of the augmented line so that row reduction turns it into I?
Put the target basis C on the left and B on the right (setting up [B∣C] in the order the bases are mentioned in the problem instead of [C∣B] would find PB←C=71[3277] and give 71[1716]). When turning the top-right 2 into 0 with R1−2R2→R1, watch the signs in 3−2(−2)=7 and −1−2(3)=−7 (subtracting the wrong sign gives [−1−253] and [94]C):
Two bases B={b1,b2} and C={c1,c2} satisfy b1=4c1−c2 and b2=−3c1+2c2. Write down PC←B directly from these equations, then find PB←C and use it to find [v]B when [v]C=[16]C.
The given equations already tell you [b1]C and [b2]C. Do those coordinate pairs go into the rows or the columns of PC←B?
Put the C-coordinates [b1]C=[4−1]C and [b2]C=[−32]C into the columns of PC←B (writing them across the rows as [4−3−12] transposes the matrix and gives 51[827]B):
PC←B=[4−1−32]
Invert PC←B to get PB←C before multiplying by [v]C (multiplying [16]C directly by PC←B without inverting first translates in the wrong direction and gives [−1411]):
PB←C=(PC←B)−1=51[2134]
[v]B=51[2134][16]C=[45]B
4
Using the basis b1=[31] (along the line) and b2=[−13] (perpendicular to the line), find the standard matrix A=BDB−1 that reflects vectors across the line through the origin and [31], and use it to reflect the standard vector [17].
In Basis B, which coordinate stays the same and which coordinate flips sign when you reflect across the line along b1?
In Basis B, reflection across b1 keeps the first coordinate and flips the second, so D=[100−1] (flipping the first coordinate with [−1001] reflects across the perpendicular vector b2 instead and gives [−55]). With B=[31−13] and B−1=101[3−113], multiply in the order BDB−1 (reversing the order to B−1DB flips the signs off the diagonal to 51[4−3−3−4] and gives 51[−17−31]):
A linear transformation in the standard grid is given by A=[43−2−1]. Find the matrix D=B−1AB that represents this transformation in Basis B with b1=[11] and b2=[23].
Put B on the right and B−1 on the left to form D=B−1AB (putting B−1 on the right as BAB−1 translates in the wrong direction and gives [3444−24−31]).
Follow an input vector starting in Basis B: which matrix on the right translates it into the standard grid first so that A can multiply it?
With B=[1123], the 2×2 inverse rule gives B−1=[3−1−21]. Put B on the right and B−1 on the left to form D=B−1AB (putting B−1 on the right as BAB−1 translates in the wrong direction and gives [3444−24−31]):