Web Simulation 

Eigenvalues and Eigenvectors - Interactive Tutorial

Eigenvalues and Eigenvectors are fundamental concepts in linear algebra with far-reaching applications in physics, engineering, machine learning, and data science. This interactive tutorial provides a geometric understanding of these concepts through visualization.

1. Introduction: What Are Eigenvalues and Eigenvectors?

1.1 The Definition

Given a square matrix A, a non-zero vector v is called an eigenvector if:

Where:

• A is an n×n matrix (linear transformation)
• v is the eigenvector (non-zero)
• λ (lambda) is the eigenvalue (scalar)

1.2 Geometric Interpretation

When you apply a linear transformation to most vectors, they change both direction and magnitude. But eigenvectors are special:

2. Finding Eigenvalues and Eigenvectors

2.1 The Characteristic Equation

To find eigenvalues, we solve the characteristic equation:

For a 2×2 matrix:

A = | a  b |
    | c  d |

Characteristic equation:
det | a-λ   b  | = 0
    |  c   d-λ |

(a-λ)(d-λ) - bc = 0

λ² - (a+d)λ + (ad-bc) = 0

λ² - trace(A)·λ + det(A) = 0

2.2 Solving for Eigenvalues

Using the quadratic formula:

2.3 The Discriminant

The discriminant Δ = trace² - 4·det determines the nature of eigenvalues:

3. Geometric Visualization

3.1 How Linear Transformations Warp Space

A 2×2 matrix transforms the entire 2D plane. In the visualization below:

• Dark gray grid: Original coordinate space (background)
• Blue grid lines: Transformed coordinate system
• Green shaded square: Input unit square (area = 1)
• Red shaded parallelogram: Transformed unit square (area = |det(A)|)
• Yellow dashed lines: Eigenvector directions (if real eigenvalues exist)

3.2 The "Aha!" Moment

3.3 The Determinant and Area

The determinant of a matrix has a beautiful geometric meaning:

In the simulation, the Input Area (green square) is always 1. The Output Area (red parallelogram) equals |det(A)|. Watch how different matrices stretch, compress, or flip the unit square!

3.4 Complex Eigenvalues: Rotation and Spirals

4. Common Matrix Types and Their Eigenvectors

The simulation includes 13 preset matrices demonstrating different transformation types:

4.1 Scaling Matrix (Preset: "Scale")

A = | 2  0 |    λ₁ = 2, v₁ = (1, 0)  ← x-axis
    | 0  3 |    λ₂ = 3, v₂ = (0, 1)  ← y-axis

Every axis-aligned vector is an eigenvector! Scales x by 2, y by 3. det = 6 (area multiplied by 6).

4.2 Shear Matrix (Preset: "Shear")

A = | 1  1 |    λ₁ = λ₂ = 1 (repeated)
    | 0  1 |    Only one eigenvector direction: (1, 0)

Only horizontal vectors stay on their span. det = 1 (area preserved!). This is called an "area-preserving shear".

4.3 Rotation Matrices (Presets: "Rotate 45°", "Rotate 90°")

A = | cos(θ)  -sin(θ) |    λ = cos(θ) ± i·sin(θ)
    | sin(θ)   cos(θ) |    COMPLEX eigenvalues!

No real eigenvectors! Every vector gets rotated - none stay on their span. det = 1 (rotation preserves area). Try the "▶ Spiral" button to see a circular orbit!

4.4 Reflection Matrices (Presets: "Reflect (x-axis)", "Reflect (y=x)")

Reflect x-axis:         Reflect y=x:
A = | 1   0 |           A = | 0  1 |
    | 0  -1 |               | 1  0 |
    
λ = +1, -1              λ = +1, -1

Reflections always have det = -1 (area preserved but orientation flipped). Notice the "(flipped)" indicator in the simulation!

4.5 Projection Matrix (Preset: "Project (onto y=x)")

A = | 0.5  0.5 |    λ₁ = 1 (onto the line y=x)
    | 0.5  0.5 |    λ₂ = 0 (perpendicular direction)

Projection onto the line y=x. det = 0 (singular!) - the entire plane collapses to a line. The red parallelogram becomes a line segment with zero area.

4.6 Symmetric Matrix (Preset: "Symmetric")

A = | 2  1 |    Symmetric matrices (A = Aᵀ) always have:
    | 1  2 |    - REAL eigenvalues (λ = 3, 1)
                - ORTHOGONAL eigenvectors

Notice the eigenvector directions are perpendicular! This is a fundamental property of symmetric matrices.

4.7 Spiral Matrices (Presets: "Spiral In/Out Fast/Slow")

5. Applications of Eigenvalues and Eigenvectors

5.1 Principal Component Analysis (PCA)

In machine learning, PCA uses eigenvectors of the covariance matrix to find the directions of maximum variance in data:

• Eigenvector with largest eigenvalue = direction of most variance
• Used for dimensionality reduction
• The eigenvalue magnitude indicates how much variance is captured

5.2 Stability Analysis

In dynamical systems dx/dt = Ax, eigenvalues determine stability:

5.3 Google PageRank

The PageRank algorithm finds the dominant eigenvector of the web link matrix:

• Web pages = states in a Markov chain
• Links = transition probabilities
• PageRank scores = eigenvector with λ = 1

5.4 Quantum Mechanics

In quantum physics, observable quantities are eigenvalues of Hermitian operators:

• Energy levels = eigenvalues of Hamiltonian
• Quantum states = eigenvectors
• Measurement collapses system to an eigenstate

5.5 Vibration Analysis

Natural frequencies of mechanical systems are determined by eigenvalues:

• Eigenvalues = squared natural frequencies (ω²)
• Eigenvectors = mode shapes (how the system vibrates)
• Critical for avoiding resonance in bridges, buildings

6. Important Properties

6.1 Key Theorems

6.2 Diagonalization

If matrix A has n linearly independent eigenvectors, it can be diagonalized:

This makes computing powers of A easy: Aⁿ = PDⁿP⁻¹

6.3 When Are Eigenvectors Orthogonal?

Symmetric Matrices: The Spectral Theorem

A matrix is symmetric if A = Aᵀ (i.e., a[i][j] = a[j][i]). For symmetric matrices:

Try it: Select the "Symmetric" preset in the simulation. Notice the two dashed eigenvector lines are exactly perpendicular!

General Matrices: No Orthogonality Guarantee

For non-symmetric matrices, eigenvectors can be at any angle:

Example: A = | 2  1 |    (upper triangular, NOT symmetric)
             | 0  3 |

λ₁ = 2 → v₁ = (1, 0)     ← horizontal
λ₂ = 3 → v₂ = (1, 1)     ← 45° diagonal

Angle between eigenvectors: 45° (NOT orthogonal!)

Summary: Eigenvector Orthogonality

7. Using the Simulation

7.1 Simulation Controls

7.2 Visual Elements

8. Exercises to Try

8.1 Find the Eigenvectors

Use the simulation to find eigenvector directions for these matrices:

| 3  1 |
| 1  3 |

| 0  1 |
| 1  0 |

| 2  0 |
| 0  2 |

8.2 Predict Before You Check

1. Calculate eigenvalues by hand using the quadratic formula
2. Predict where the eigenvector directions should be
3. Predict the output area (det = λ₁ × λ₂)
4. Verify using the simulation

8.3 Explore Edge Cases

• Singular matrix (det = 0): Try the "Project (onto y=x)" preset. Notice the red parallelogram collapses to a line!
• Pure rotation (det = 1, complex λ): Try "Rotate 45°" or "Rotate 90°". The spiral is a perfect circle.
• Negative determinant: Try "Reflect (x-axis)". Notice the "(flipped)" label!
• Repeated eigenvalue: Try the "Shear" preset (λ₁ = λ₂ = 1).

8.4 Spiral Challenge

1. Select "Spiral In (Fast)" preset
2. Drag the purple marker to position (2, 2)
3. Click "▶ Spiral" and count how many iterations until it nearly reaches origin
4. Now try "Spiral In (Slow)" - how many more iterations does it take?
5. Try "Spiral Out" presets - the trajectory diverges! Can you predict when it will go off-screen?

9. Mathematical Summary