When a system gets bumped away from its balance point at [00], we want to predict where it goes over time and whether it settles back down or grows without bound.
1
Predicting future steps with eigenvectors
In a stepping system, multiplying the current state [xy] by a matrix gives the state at the next step:
[xnextynext]=[2112][xy]
We want to find where a starting point like [31] lands at Step 2, and at any future Step k, without multiplying by the matrix over and over.
Predicting future steps with eigenvectors
Text solution
Find the eigenvector lines and multipliersThe matrix [2112] keeps the line through v1=[11] and the line through v2=[1−1] pointing in the same direction, scaling them by their eigenvalues each step:
v1=[11]v2=[1−1] scales by λ1=3, scales by λ2=1
Split the starting point into eigenvector piecesWrite Step 0=[31] as a combination of the two eigenvectors so we can track each piece along its own line:
[31]=2[11]+1[1−1]
Multiply each piece by λ2 to get Step 2Each step multiplies the first piece by 3 and the second piece by 1:
Write the formula for any Step kAfter k steps, each piece has been multiplied by its eigenvalue k times:
Step k=c1λ1kv1+c2λ2kv2
So starting from [31], the system is at 2⋅3k[11]+1⋅1k[1−1] at Step k, which gives [1917] at Step 2.
2
Stability and saddle points in stepping systems
A balance point at [00] is stable if any point bumped away from [00] eventually returns to [00]. We want to see which values of λ1 and λ2 pull points into the origin and what happens when one eigenvector line shrinks while the other grows.
Stability and saddle points in stepping systems
Stable
λ1=0.5
λ2=0.8
Unstable
λ1=3
λ2=2
Saddle Point
λ1=0.5
λ2=2
In the Step k formula, only λ1k and λ2k change as k grows. A negative multiplier like λ=−0.5 sends a point on its line back and forth across [00] (−0.5,+0.25,−0.125,… for a point starting at 1), so what controls whether λk shrinks to zero is its size ∣λ∣:
Stable stepping system: both pieces shrink to [00]∣λ1∣<1and∣λ2∣<1
When one multiplier has ∣λ1∣<1 and the other has ∣λ2∣>1, the origin is an unstable Saddle Point.
For a saddle point with λ1=0.5 and λ2=2, starting closer to the λ1=0.5 line makes the stepped path move closer to [00] at first, but any non-zero piece on the λ2=2 line doubles every step and pushes the system outward along that growing line.
At a saddle point, even a tiny piece on the growing line eventually pulls the path outward, so only points right on the shrinking line reach the center.
3
Complex eigenvalues and spirals in stepping systems
Now suppose a stepping matrix has complex eigenvalues λ=a±bi, so there are no real eigenvector lines. For a matrix of the form [ab−ba], which has these eigenvalues, each step rotates the point and scales its distance from [00], and we want to know whether that spiral steps inward, steps outward, or stays on a circle.
Complex eigenvalues and spirals in stepping systems
[ab−ba]
λ=a±bi
∣λ∣=?
1
Text solution
Find the distance multiplier ∣λ∣Multiplying by λ=a±bi rotates the point and scales its radius by the magnitude of the complex number:
∣λ∣=a2+b2
Case 1: [0.50.5−0.50.5] with λ=0.5±0.5iCompute the magnitude from a=0.5 and b=0.5:
∣λ∣=0.52+0.52=0.5≈0.71<1
Each step rotates the point by 45∘ and shrinks its distance from [00] to 71%, so points step along an inward spiral into [00] (Stable).
Case 2: [11−11] with λ=1±1iCompute the magnitude from a=1 and b=1:
∣λ∣=12+12=2≈1.41>1
Each step rotates the point and stretches its distance by 41%, so points step along an outward spiral to infinity (Unstable).
Case 3: [0.60.8−0.80.6] with λ=0.6±0.8iEven though 0.6 and 0.8 are individually less than 1, find the total magnitude:
∣λ∣=0.62+0.82=0.36+0.64=1=1
Because the multiplier has size 1, points step around a closed circle forever without shrinking to [00] (Not stable, though they never blow up either, so this is often called neutrally stable).
Now let's see how this looks on the complex plane, where the unit circle ∣λ∣=1 forms the boundary between stable and unstable stepping systems.
Whether the steps spiral inward, stay on a circle, or spiral outward depends only on whether the eigenvalue lies inside, on, or outside the circle of radius 1.
4
Continuous flow and why stability needs λ<0
In continuous systems like a hovering drone or a swinging pendulum, time flows smoothly and the matrix gives the instantaneous velocity [x′y′]=A[xy] at every point (the prime ′ means rate of change with respect to time t). We want to find the flowing path [x(t)y(t)] and see why the stability rule switches from ∣λ∣<1 to λ<0.
Continuous flow and why stability needs λ < 0
[x′y′]=A[xy]
Stepping: λ=+0.5
0
1
2
3
Flowing: λ=+0.5
+0.5×2=+1
0
1
2
3
Flowing: λ<0
0
1
2
3
On an eigenvector line, Velocity=λ×Position. For the velocity arrow to point opposite to the position and pull the point back toward [00],λ must be negative. Along each eigenvector line, the function with rate of change λ times itself is eλt, which turns our stepping formula into the continuous flowing formula:
Stepping (xnext=Ax):Step k=c1λ1kv1+c2λ2kv2(Stable if ∣λ1∣<1,∣λ2∣<1)Flowing (x′=Ax):[x(t)y(t)]=c1decays if λ1<0grows if λ1>0eλ1tv1+c2eλ2tv2(Stable if λ1<0,λ2<0)
5
Solving a continuous system with real eigenvalues
Now let's find the paths [x(t)y(t)] for the continuous system [x′y′]=[−211−2][xy] and see which line paths follow as they approach [00].
To get the eigenvalues of a 2×2 matrix A=[acbd] quickly, remember how the Trace and Determinant relate to λ1 and λ2:
λ1+λ2Trace=a+dandλ1λ2Determinant=ad−bc
Solving a continuous system with real eigenvalues
[x′y′]=[−211−2][xy]
[x(t)y(t)]=?
Text solution
Find the eigenvalues using Trace and DeterminantAdd the diagonal entries for the sum and compute ad−bc for the product:
Write the flowing solution [x(t)y(t)]Multiply each eigenvector by its eλt factor:
[x(t)y(t)]=c1e−t[11]+c2e−3t[1−1]
So both pieces decay to zero, and because e−3t shrinks much faster than e−t, the [1−1] piece disappears first and the path moves into [00] along the slower [11] line.
Here is how that faster decay shapes the paths in the velocity field for [x′y′]=[−211−2][xy].
Because the fast component shrinks to zero much sooner than the slow one, almost every path bends onto the slow line before reaching the origin.
6
Complex eigenvalues and spirals in continuous flow
Next, let's see what happens when a continuous system has complex eigenvalues λ=a±bi, starting with [x′y′]=[−12−2−1][xy]. We want to know what controls whether the flowing spiral goes in, goes out, or travels in a closed loop.
Complex eigenvalues and spirals in continuous flow
[x′y′]=[−12−2−1][xy]
λ=?
Text solution
Find the eigenvalues from Trace and DeterminantCompute the sum and product for [−12−2−1]:
Split eλt into its radius and rotation factorsBecause λ=a±bi is in the exponent, addition separates into a product of two exponentials:
e(a±bi)t=radiuseat⋅rotatione±ibt
Check the sign of the real part aFor λ=−1±2i, the real part is a=−1<0:
e(−1±2i)t=e−te±2it
Because e−t→0 as t grows while e±2it rotates the point, the paths form a Stable Spiral into [00] (whereas a>0 like +1±2i gives an Unstable Spiral via e+t, and a=0 like 0±2i gives e0t=1, a closed loop called a Center).
Now let's compare continuous and stepping paths for the same complex eigenvalue λ=a+bi and see how their stability regions differ on the complex plane.
For the same complex eigenvalue, a flowing system switches between spiraling in and out across the vertical axis, while a stepping system switches across the unit circle.
7
Quick stability test using trace and determinant
When we only want to know whether a continuous 2×2 system x′=Ax is stable, unstable, a saddle, or a center, we can tell straight from Trace=λ1+λ2 and Determinant=λ1λ2 without solving for the eigenvalues.
Quick stability test using trace and determinant
x′=Ax
Determinant=λ1λ2
Trace=λ1+λ2
?
?
?
Check the sign of the Determinant first, and if it is positive, check the sign of the Trace:
Determinant<0⟹λ1 and λ2 have opposite signs⟶Saddle (Unstable)Determinant>0⟹⎩⎨⎧Trace<0:Trace>0:Trace=0:Real parts<0⟶StableReal parts>0⟶UnstableReal parts=0⟶Center
Checking the Determinant and Trace classifies stability in seconds, though if you need to know which eigenvector line paths follow or whether they spiral vs. move along straight lines, you still need to solve for λ1 and λ2.
Key concepts
Stepping (xnext=Ax) vs. Flowing (x′=Ax)
Always check how time moves before testing λ.
Stepping multiplies position by λ, so compare the size∣λ∣=a2+b2 to 1:
Flowing sets velocity to λ×Position, so compare the real part of λ to 0:
Which line paths follow
One shrinking and one growing eigenvalue (∣λ1∣<1<∣λ2∣ in stepping, or λ1<0<λ2 in flowing) creates an unstable Saddle Point where paths that start close to the shrinking line move toward [00] at first before heading out along the growing eigenvector line.
Two negative real eigenvalues in continuous flow (like λ1=−1,λ2=−3) pull every path into [00], and every path that doesn't start on the faster line comes in along the slower eigenvector line (λ1=−1), because the faster exponential (e−3t) disappears first.
When not to solve for eigenvalues
If you only need to classify the stability of a continuous 2×2 system, use Determinant=ad−bc and Trace=a+d directly: Det<0 is a Saddle, while Det>0 is Stable if Trace<0,Unstable if Trace>0, and a Center if Trace=0.
8
Practice
1
The stepping system [xnextynext]=[3113][xy] has eigenvector v1=[11] with λ1=4 and eigenvector v2=[1−1] with λ2=2. If the system starts at Step 0=[51], write the formula for Step k and find the exact position at Step 3.
First split [51]=c1[11]+c2[1−1] by checking what c1+c2 and c1−c2 must equal.
Split the starting point [51] into its two eigenvector pieces (watch out: don't use the coordinates 5 and 1 directly as c1 and c2, which would give [328312]):
[51]=3[11]+2[1−1]
Multiply each piece by λk to get the formula for Step k:
Step k=3⋅4k[11]+2⋅2k[1−1]
Plug in k=3 (raising 4 and 2 to the 3rd power 43=64 and 23=8, rather than multiplying by 3, which would give [4824]):
A stepping system has eigenvector v1=[12] with multiplier λ1=−0.5 and eigenvector v2=[2−1] with multiplier λ2=3. Is this system stable or unstable? Describe what happens step by step if you start at [50]=1[12]+2[2−1], and if you start at [36]=3[12].
Compare ∣λ1∣ and ∣λ2∣ to 1, and notice what multiplying by −0.5 does to the sign and size of the [12] piece each step.
Check the sizes of the multipliers: ∣λ1∣=∣−0.5∣=0.5<1 (watch out: in a stepping system, λ1=−0.5 shrinks in size rather than growing negatively, because ∣−0.5∣<1), while ∣λ2∣=3>1. Because ∣λ2∣>1, the system is Unstable (a Saddle Point).
Starting at [50]=1[12]+2[2−1]:
Step k=1(−0.5)k[12]+2(3)k[2−1]
The [12] piece alternates sides across the origin while cutting its distance in half each step, while the [2−1] piece triples every step and pushes the point out to infinity along the line through [2−1].
Starting at [36]=3[12] (purely on the shrinking eigenvector line, with 0 piece on [2−1]):
The point alternates sides across the origin while halving its distance each step and settles into [00].
3
Three stepping systems have complex eigenvalues: System A has λ=0.3±0.4i, System B has λ=0.8±0.6i, and System C has λ=0.8±0.8i. For each system, find ∣λ∣, describe the path of a point, and state whether the system is stable.
Compute the magnitude ∣λ∣=a2+b2 for each pair and compare it to 1.
System A (λ=0.3±0.4i)
∣λ∣=0.32+0.42=0.09+0.16=0.25=0.5<1
Points step along an inward spiral to [00] (Stable).
System B (λ=0.8±0.6i)Watch out for the temptation to call Systems B and C stable just because the individual numbers 0.8 and 0.6 are less than 1:
∣λ∣=0.82+0.62=0.64+0.36=1=1
Points step around a closed circle forever without shrinking or growing, so the system is not stable, though it never blows up either (often called neutrally stable).
System C (λ=0.8±0.8i)
∣λ∣=0.82+0.82=0.64+0.64=1.28≈1.13>1
Points step along an outward spiral to infinity (Unstable).
4
Suppose a matrix A has eigenvalues λ1=0.4 and λ2=0.8. Compare the stability of the stepping system xnext=Ax with the continuous flowing system x′=Ax, and explain why the same eigenvalues give opposite results.
Check what multiplying position by 0.4 and 0.8 does at each step in xnext=Ax, versus which way the velocity arrow Velocity=λ×Position points in x′=Ax.
Stepping system (xnext=Ax)Both multipliers have size less than 1 (∣0.4∣<1 and ∣0.8∣<1), so multiplying position by 0.4 and 0.8 at each step shrinks both pieces to [00], since 0.4k and 0.8k go to 0 (Stable).
Flowing system (x′=Ax)Don't carry the ∣λ∣<1 rule over to continuous flow: here both eigenvalues are positive (0.4>0 and 0.8>0). Along both eigenvector lines, Velocity=λ×Position points in the same direction as position, pushing points outward away from [00] like e0.4t and e0.8t (Unstable).
5
For the continuous system [x′y′]=[−311−3][xy], find the eigenvalues and eigenvectors, write the solution [x(t)y(t)] starting at [x(0)y(0)]=[42], and state which line the path follows into [00] as t grows.
Find the eigenvalues from the Trace and Determinant, split [42] along the two eigenvectors, and check which eλt term disappears first as t increases.
Split the starting point [42]=3[11]+1[1−1] to write the solution:
[x(t)y(t)]=3e−2t[11]+1e−4t[1−1]
Watch out: don't pick [1−1] just because ∣−4∣>∣−2∣. Because e−4t decays much faster than e−2t, the [1−1] piece disappears first and the path moves into [00] along the slower [11] line (y=x).
6
Find the eigenvalues of the continuous system [x′y′]=[−23−3−2][xy], rewrite eλt to separate the radius and rotation factors, and classify the stability and shape of the paths.
Watch the negative sign on −3 when calculating Determinant=ad−bc, then write e(a±bi)t as a product of two exponentials.
Compute the Trace and Determinant (be careful with the signs in ad−bc: subtracting (−3)(3) adds 9, whereas dropping the minus sign would give 4−9=−5 and falsely suggest a Saddle):
Here e−2t shrinks the radius to zero while e±3it rotates the point around the origin. (Don't use the stepping test ∣λ∣=(−2)2+32=13>1, which would falsely call this unstable.) Because the real part is negative (−2<0), the system is a Stable Spiral.
7
Consider the continuous system [x′y′]=[k2−21][xy], where k is a real constant. Using only the Trace and Determinant, find the values of k for which the origin is (a) a Saddle, (b) Stable, (c) a Center, and (d) Unstable with outward paths (not a saddle).
Write Trace and Determinant in terms of k first; for parts (b)–(d), make sure both Determinant>0 and the Trace condition hold at the same time.
First find the Trace and Determinant in terms of k:
TraceDeterminant=k+1=k(1)−(−2)(2)=k+4
(a) SaddleRequires Determinant<0, so k+4<0, which gives k<−4.
(b) StableRequires both Determinant>0 and Trace<0 (watch out: checking only Trace<0 gives k<−1, which wrongly includes k≤−4, where the Determinant is not positive). Combining k+4>0 and k+1<0 gives −4<k<−1.
(c) CenterRequires Determinant>0 and Trace=0, which gives k=−1.
(d) Unstable with outward pathsRequires Determinant>0 and Trace>0, which gives k>−1.