To see what repeated multiplication by a matrix does to a vector, we look for its eigenvalues and eigenvectors, though sometimes a matrix of real numbers gives us complex eigenvalues.
1
Why Complex Eigenvalues Mean Rotation
When a matrix has real eigenvalues, its eigenvectors stay on their own lines, and diagonalizing it as A=PDP−1 turns the matrix into simple scaling along those lines. We want to find the eigenvalues of
A=[11−2−1]
to see where the vector [10] lands after multiplying by A a hundred times.
Why Complex Eigenvalues Mean Rotation
A=[11−2−1]
A100[10]=?
Text solution
Subtract λ from the diagonal and set the determinant to zeroWe set up the characteristic equation det(A−λI)=0 to look for vectors that stay on their own lines:
(1−λ)(−1−λ)−(−2)(1)=0
Expand the determinant and solve for λMultiplying out the terms simplifies the equation to λ2+1=0:
−1+λ2+2λ2+1λ2λ=0=0=−1=±i
So the eigenvalues are λ=±i, the exact same eigenvalues as the pure 90∘ counterclockwise rotation matrix [01−10], under which no real vector stays on its own line.
2
Rotation-Scaling Matrices
A pure ±90∘ rotation gives λ=±i, where the eigenvalue has no real part. Now we want to see what matrix comes from a full complex eigenvalue λ=a+bi, and how to find the scale factor and rotation angle of
[11−11]
A pure counterclockwise rotation by an angle θ is given by the standard rotation matrix
[cosθsinθ−sinθcosθ]
Rotation-Scaling Matrices
A matrix with eigenvalues λ=a±bi that stretches by a along each axis and by b rotated 90∘ counterclockwise factors into a scaling and a pure rotation:
Match the matrix with [ab−ba] to read off the legs a and bThe diagonal entries give the horizontal leg a and the bottom-left entry gives the vertical leg b:
a=1,b=1
Use the right triangle with legs 1 and 1 to find the hypotenuse r and angle θThe Pythagorean theorem gives the scale factor r, and the ratios ra and rb give the angle θ:
rcosθsinθ=12+12=2=21,=21⟹θ=4π (or 45∘)
Factor the scale factor r out of the matrixPulling out 2 leaves the pure 45∘ rotation matrix:
[11−11]=2[cos(4π)sin(4π)−sin(4π)cos(4π)]
So each multiplication by [11−11] rotates a vector 45∘ counterclockwise and stretches its length by 2.
When you multiply by C=[ab−ba] over and over, the path traces a circle when r=1, spirals outward when r>1, spirals inward when r<1, and stays on a straight line when b=0.
Where the eigenvalue sits sets both the rotation angle and the scale factor of the matrix, turning a vector’s repeated path into a circle, a spiral, or a straight line.
3
Rotation on a Tilted Ellipse
A matrix of the form [ab−ba] has equal diagonal entries and rotates and scales vectors on the standard grid, so its paths are circles or spirals around the origin. Our starting matrix A=[11−2−1] has different diagonal entries (1 and −1), yet its eigenvalues are still λ=±i (r=1,θ=90∘). Let's find what A actually does to [10] over repeated steps and compute
A100[10]forA=[11−2−1]
Rotation on a Tilted Ellipse
[ab−ba]
A=[11−2−1]
[10]
Text solution
Multiply [10] by A four times until it returns to the startMultiplying the vector by A one step at a time shows a four-step loop:
Use the four-step loop to evaluate the 100th powerBecause 100 is 25 full cycles of 4, the vector goes around the loop 25 times:
A100[10]=(A4)25[10]=[10]
So after 100 multiplications, the vector lands right back at [10], having gone around a tilted ellipse 25 times.
[01−10]
[10]
90∘
A=[11−2−1]
[10]
[11]
[−10]
[−1−1]
A
4
Complex Eigenvectors and the Tilted Basis
Instead of multiplying vectors out step by step to see the tilted ellipse, we want a direct algebraic way to find the basis vectors of the tilted grid where
A=[11−2−1]
rotates space.
Complex Eigenvectors and the Tilted Basis
A=[11−2−1],λ=i:v=?
Text solution
Substitute λ=i into Av=λv and read off the simpler rowOnce an eigenvalue of a 2×2 matrix is plugged in, one row is always a multiple of the other, so the bottom row of [11−2−1][xy]=i[xy] gives the whole relationship:
1x−1yxx=iy=y+iy=(1+i)y
Choose a nonzero value for the free variable y and split v into real and imaginary partsSetting y=1 gives x=1+i, which splits into two real vectors u and w:
v=[1+i1]=real part u[11]+iimaginary part w[10]
Flip the sign of i to get the conjugate eigenvector for λˉ=−iBecause A is real, taking the complex conjugate of Av=λv gives Avˉ=λˉvˉ without any extra algebra:
Notice that both u=[11] and w=[10] are completely real vectors (w is the vector multiplied by i, without the i inside it). Expanding A(u+iw)=i(u+iw)=−w+iu and matching the real and imaginary parts on each side shows what A does to them:
Au=−w,Aw=u
w
u
45∘
A
−w
A
So u and w are the basis vectors of the skewed grid for the tilted ellipse, sitting one quarter-turn apart under A.
Picking any other nonzero complex number for the free variable y in x=(1+i)y just gives another valid pair of basis vectors u and w on the same family of tilted ellipses, still one quarter-turn apart (Aw=u and Au=−w). Because the ellipse is tilted, u and w aren't perpendicular unless they happen to line up with its main axes.
Every nonzero choice of the complex free variable y gives you another valid pair of basis vectors u and w on the same family of tilted ellipses, always one quarter-turn apart when multiplied by A.
5
Factoring A=PCP−1
Now we can put the rotation-scaling matrix C and the tilted basis vectors u and w together to factor any real 2×2 matrix with complex eigenvalues as A=PCP−1, the complex version of A=PDP−1. Let's factor
A=[31−21]
Factoring A = PCP⁻¹
[31−21]=PCP−1
P=?C=?
For any eigenvalue λ=a+bi with eigenvector v=u+iw, expanding A(u+iw)=(a+bi)(u+iw)=(au−bw)+i(bu+aw) and matching real and imaginary parts tells us where A sends both basis vectors:
Au=au−bw,Aw=bu+aw
Reading the weights on u and w directly from those two equations gives the two valid ways to pair P and C:
u in column 1 puts −b in column 1P=∣u∣∣w∣,C=[a−bba]oru in column 2 puts −b in column 2P=∣w∣∣u∣,C=[ab−ba]
Text solution
Set det(A−λI)=0 and solve for the complex eigenvaluesSubtracting λ from the diagonal gives a quadratic equation that we can solve by completing the square:
(3−λ)(1−λ)−(−2)(1)λ2−4λ+5(λ−2)2+1λ=0=0=0=2±i
Plug λ=2+i into Av=λv and split the eigenvector into u and wUsing the bottom row of [31−21][xy]=(2+i)[xy] and picking y=1 gives u and w:
Build P,C, and P−1 with matching column orderPutting u in the first column and w in the second column of P puts −b=−1 in the first column (bottom-left) of C for a=2,b=1:
(Swapping the columns to P=[wu]=[1011] puts −1 in the top-right instead, giving C=[21−12] and P−1=[10−11].)
In the tilted basis P, each multiplication by A scales vectors by r=22+12=5 and rotates them by θ (where cosθ=52 and sinθ=51), spiraling them outward.
Changing the basis vectors u and w in P tilts and stretches the coordinate grid that the path follows, while C keeps the same scale factor r and rotation angle θ at each step.
Changing the basis vectors tilts and stretches the grid the spiral follows, without changing its scale factor or rotation angle at each step.
Key concepts
When you find the eigenvalues of a real 2×2 matrix from det(A−λI)=0, whether the eigenvalues are real or complex tells you whether the matrix scales along fixed lines or rotates, and whether to factor it as A=PDP−1 or A=PCP−1. If the eigenvalues are real and distinct, such as λ=2,3 for [2003], there is no rotation at all. Do not use C; diagonalize as A=PDP−1, where the real eigenvectors stay on their own lines and D scales along them.
When det(A−λI)=0 gives a complex conjugate pair λ=a±bi, the matrix is rotating space. In both cases below, the real part a and imaginary part b are the horizontal and vertical legs of a right triangle that set the scale factor and rotation angle at each step:
rcosθsinθ=a2+b2,=ra,=rb
If the matrix already has equal diagonal entries and opposite off-diagonal entries like C=[34−43] (λ=3±4i), it is already rotating on the standard grid, so no basis matrix P is needed; just factor out the hypotenuse r=5:
[ab−ba]=r[cosθsinθ−sinθcosθ]
If the matrix does not have the [ab−ba] pattern, like A=[31−21] (λ=2±i), it is rotating on a skewed grid, so factor it as A=PCP−1. Solve Av=λv for just one eigenvalue (λ=2+i) using the simpler row to get v=u+iw=[11]+i[10] (the conjugate eigenvalue λˉ=2−i automatically has eigenvector vˉ=u−iw). Put the real vectors u and w into the columns of P, and match the position of −b in C to the column of u using Au=au−bw and Aw=bu+aw:
P=[uw]P=[wu]⟹C=[a−bba],⟹C=[ab−ba]
6
Practice
1
Find the complex eigenvalues λ, the scale factor r, and the counterclockwise rotation angle θ of the matrix C=[31−13], and write C in the factored form r[cosθsinθ−sinθcosθ].
[31−13] has the form [ab−ba]: right triangle with horizontal leg a=3 and vertical leg b=1, hypotenuse r=?
Matching C=[31−13] with [ab−ba] gives the horizontal and vertical legs and the complex eigenvalues:
a=3,b=1,λ=3±i
(Watch out: the top-right entry of [ab−ba] is −b, so b=1, not −1. Reading b=−1 from the top-right gives sinθ=−21 and the wrong clockwise angle θ=−6π.)
Find the scale factor r from the hypotenuse of the right triangle:
r=(3)2+12=3+1=4=2
Use the legs over the hypotenuse to find the rotation angle θ:
cosθsinθ=23,=21⟹θ=6π radians (or 30∘)
Factor out r=2:
C=2[cos(6π)sin(6π)−sin(6π)cos(6π)]
2
A real 2×2 matrix A has eigenvalue λ=3+4i with complex eigenvector v=[1+2i1]=u+iw. Without finding A itself: (a) write the other eigenvalue of A and its eigenvector, (b) express Au and Aw in terms of u and w, and (c) write P and C in A=PCP−1 both when P=[uw] and when P=[wu].
Split v into two completely real vectors u and w first, then expand A(u+iw)=(3+4i)(u+iw) and match the real and imaginary parts.
Splitting v into its real and imaginary parts gives:
v=[1+2i1]=[11]+i[20]⟹u=[11],w=[20]
(a)Take the complex conjugate of λ and v by flipping the sign of i:
(Watch out: (4i)(iw)=4i2w=−4w; forgetting i2=−1 gives the wrong real part Au=3u+4w.)
Matching the real and imaginary parts on both sides gives:
Au=3u−4w,Aw=4u+3w
(c)Reading the columns of C from Au=3u−4w and Aw=4u+3w for each column order of P (putting C=[34−43] with P=[uw] puts −4 in the wrong column, matching Aw=−4u+3w instead):
When P=[uw]=[1120],C=[3−443]
When P=[wu]=[2011],C=[34−43]
3
Factor the matrix A=[31−5−1] as A=PCP−1, giving the matrices P,C, and P−1 explicitly.
After finding λ=a±bi from det(A−λI)=0, use the bottom row of Av=λv to express x in terms of y.
Set the determinant of A−λI to zero to find the eigenvalues:
So a=1 and b=1. For λ=1+i, write out the bottom row of [31−5−1][xy]=(1+i)[xy]:
x−yxx=(1+i)y=y+(1+i)y=(2+i)y
(Watch out: when adding y to (1+i)y, the real part increases from 1 to 2; dropping the −y term on the left without adding it on the right gives x=(1+i)y and the wrong basis vector u=[11].)
Setting y=1 gives x=2+i, which splits into u and w:
v=[2+i1]=[21]+i[10]⟹u=[21],w=[10]
Using P=[uw] (and remembering to divide by det(P)=2(0)−1(1)=−1 when inverting P, rather than forgetting to divide and leaving [0−1−12]):
PCP−1=[2110],=[1−111],=[011−2]
(Or, using P=[wu]:P=[1021],C=[11−11], and P−1=[10−21].) Starting from λ=1−i instead gives v=[2−i1], so u=[21] and w=[−10], and P=[uw]=[21−10] gives C=[11−11] and P−1=[0−112], an equally valid factorization.
4
When finding the eigenvector of A=[11−2−1] for λ=i from x=(1+i)y, one student picks y=1 to get v1=u1+iw1 and a second student picks y=i to get v2=u2+iw2. Find u2 and w2, and explain how the two pairs of real basis vectors are related on the tilted ellipse.
Substitute y=i into both entries of v2=[(1+i)yy] and expand (1+i)i using i2=−1 before splitting v2 into its real and imaginary parts.
For y=1, the first eigenvector splits into:
v1=[1+i1]=[11]+i[10]⟹u1=[11],w1=[10]
For y=i, substitute y=i into both entries of [xy]:
v2=[(1+i)ii]=[i+i2i]=[−1+ii]
(Watch out: the bottom entry of v2 is y=i, not 1. Leaving the bottom entry as 1 gives [−1+i1] and the wrong pair u2=[−11],w2=[10].)
Split v2 into its real and imaginary parts:
v2=[−10]+i[11]⟹u2=[−10],w2=[11]
Because v2=iv1, multiplying the eigenvector by i rotates both basis vectors forward by one 90∘ step along the exact same tilted ellipse: w2=u1 and u2=−w1.
5
For the matrix A=[21−20], find its complex eigenvalues, determine the scale factor r and rotation angle θ in the tilted basis, and use them to compute A8[35].
Once you have r and θ from λ=a+bi, find the total rotation angle 8θ and total scaling r8 in C8 before deciding whether you need to compute P.
Find the eigenvalues from det(A−λI)=0:
(2−λ)(0−λ)−(−2)(1)λ2−2λ+2(λ−1)2+1λ=0=0=0=1±i
Read off a=1 and b=1 to find the scale factor r and rotation angle θ in the tilted basis:
rcosθsinθ=12+12=2,=21,=21⟹θ=4π (or 45∘)
Over 8 steps, the total rotation angle and total scale factor are:
8θr8=8×45∘=360∘,=(2)8=24=16
(Watch out: r=12+12=2, not 2; forgetting the square root gives 28=256 and the wrong vector [7681280].)
Because a 360∘ rotation brings every vector all the way back to its starting direction, C8=16I, which makes P and P−1 cancel out: