Lagrangian Mechanics: Setting Up the Double Pendulum Problem

Chaotic systems exhibit exponential sensitivity to initial conditions. While phase space trajectories for different systems never cross (maintaining determinism), very similar initial conditions lead to trajectories that diverge exponentially over time. In the double pendulum example (two degrees of freedom, four-dimensional phase space), initially similar trajectories track together briefly but then diverge dramatically, becoming completely unrecognizable from each other after sufficient time. This divergence characterizes chaotic behavior despite the underlying deterministic nature of classical mechanics.

The double pendulum demonstrates extreme sensitivity to initial conditions, where even the tiniest differences in starting positions lead to vastly different outcomes over time. This phenomenon is called the butterfly effect, famously illustrated by the idea that a butterfly flapping its wings in Brazil could cause a tornado in Texas. When double pendulums are released from nearly identical positions with minute angular differences (as small as 0.00000001 degrees), they quickly diverge onto completely different paths, showing how small perturbations can amplify dramatically in chaotic systems.

This concluding section presents simulation results demonstrating the double pendulum's complex behavior and chaotic nature. Multiple visualization techniques are employed: trajectory plots showing mass positions over time, angular displacement vs. time graphs for θ₁ and θ₂, angular velocity graphs for θ̇₁ and θ̇₂, and phase space plots of θ₁ vs. θ₂. Examples with various initial angles (45°, 90°, 120°) reveal non-periodic motion with emerging patterns. The critical demonstration of chaos theory shows three simulations with initial angles differing by only 0.001 degrees. Initially, trajectories appear nearly identical, but they diverge dramatically over time, illustrating extreme sensitivity to initial conditions. This exemplifies the butterfly effect: deterministic systems can produce unpredictable long-term behavior due to nonlinear interactions, making practical prediction impossible despite complete theoretical determinism.

The double pendulum system demonstrates chaos, where trajectories starting from nearly identical initial conditions (differing by only 0.1%) eventually diverge dramatically. This sensitive dependence on initial conditions means that even tiny measurement errors lead to completely different predictions of future behavior, representing a fundamental limit on predictability. The system's evolution is governed by differential equations relating the accelerations of both angles, creating a four-dimensional dynamics that can be visualized in two dimensions by plotting the positions of the pendulum ends.

Real pendulums have extended mass distributions, requiring the moment of inertia (I) for accurate period prediction. For spherical bobs, T = 2π√(L/g) × √(1 + 2r²/(5L²)). Phase space representation (displacement vs. velocity) reveals behaviors invisible in time-domain plots: ideal pendulums trace ellipses, while damped pendulums trace spirals converging toward equilibrium. When pendulums receive sufficient velocity to complete full rotations, trajectories become non-closed. The double pendulum exhibits chaotic behavior—small initial condition changes produce dramatically different trajectories. This sensitive dependence on initial conditions is the butterfly effect. Linearizing equations (replacing sin(θ) with θ) eliminates chaos, demonstrating that chaos arises from non-linear dynamics.

Near equilibrium, nonlinear systems can be linearized using small angle approximations: cos(φ) ≈ 1 - φ²/2 and sin(φ) ≈ φ. Keeping only linear terms in displacements and their first/second time derivatives reduces the equations to linear form. The normal mode ansatz assumes all coordinates vibrate at the same frequency: X(t) = X₀e^(iωt) and φ(t) = φ₀e^(iωt). Substituting into the linearized equations yields algebraic relations between amplitudes and frequency. This transforms the differential equations into an eigenvalue problem, enabling analytical solution of the otherwise unsolvable coupled system.

The instructor analyzes small oscillations around the equilibrium position θ₁ = 0 and θ₂ = 0. For small angles, sin(θ) ≈ θ and cos(θ) ≈ 1, allowing linearization of the equations. The linearized equations are written in matrix form Mθ̈ + Kθ = 0, where M is the mass matrix [[m₁l₁² + m₂l₁², m₂l₁l₂], [m₂l₁l₂, m₂l₂²]] and K is the stiffness matrix [[(m₁+m₂)gl₁, 0], [0, m₂gl₂]]. The normal modes are found by solving the characteristic equation det(K - ω²M) = 0, yielding two solutions for ω² corresponding to the two normal mode frequencies. The general solution is a superposition of normal modes: θ₁(t) = A₁cos(ω₁t + φ₁) + A₂cos(ω₂t + φ₂) and θ₂(t) = B₁cos(ω₁t + φ₁) + B₂cos(ω₂t + φ₂). The constants are determined by initial conditions. The instructor recommends exercises 7, 12, 13, 14, 17, and 20 for practice.

This segment covers finding the natural frequency and applying small angle approximations. The instructor explains: (1) Rearranging the equation of motion into standard form θ̈ + ω²θ = 0; (2) Taking the square root of the coefficient to find ω; (3) Small angle approximations - sin(θ) ≈ θ and cos(θ) ≈ 1 for θ << 1 radian; (4) These approximations linearize the equations, allowing straightforward solution; (5) The natural frequency represents the rate of oscillation when displaced from equilibrium; (6) The same result can be obtained using either Newtonian or energy methods.

This segment demonstrates the complete process of analyzing small oscillations in a coupled system. The instructor derives the Lagrangian for a particle attached to two identical springs, then applies the small oscillations approximation (x << l) to linearize the equations. The instructor shows how to set up the matrix equation (V - ω²M)q = 0 and solve for normal mode frequencies by finding the determinant. The solution yields three frequencies: ω = 0, ω = √(3k/m), and ω = √(k/m). This demonstrates the systematic approach to analyzing coupled oscillatory systems: derive Lagrangian, apply small oscillations approximation, set up the characteristic equation, and solve for normal mode frequencies.

For small oscillations, the equations of motion can be linearized by approximating trigonometric functions: sin(θ) ≈ θ and cos(θ) ≈ 1. This simplifies the equations to linear differential equations. After linearization, the equations take the form: I1θ1'' + kL²(θ1 - θ2) + mgLθ1 = 0 and I2θ2'' + kL²(θ1 - θ2) + mgLθ2 = 0. These linear equations can then be solved to find the natural frequencies and mode shapes of the system.

This comprehensive section covers the complete derivation of Hamiltonian mechanics from the Legendre transform applied to the Lagrangian. The Lagrangian L(q, q̇, t) transforms to the Hamiltonian H(q, p, t) by changing independent variables from velocities q̇ to momenta p. Key relations emerge: pᵢ = ∂L/∂q̇ᵢ (canonical momentum definition), q̇ᵢ = ∂H/∂pᵢ, and ∂L/∂qᵢ = -∂H/∂qᵢ. Combining these with the Euler-Lagrange equation yields Hamilton's equations: ṗᵢ = -∂H/∂qᵢ and q̇ᵢ = ∂H/∂pᵢ. Unlike Lagrangian mechanics producing one second-order equation per degree of freedom, Hamiltonian mechanics produces two first-order equations, offering computational advantages for numerical solutions.

The Legendre transform provides the mathematical bridge between two major formulations of classical mechanics: Lagrangian mechanics and Hamiltonian mechanics. Lagrangian mechanics uses the Lagrangian L(q, q̇), a function of generalized coordinates q and velocities q̇, while Hamiltonian mechanics uses the Hamiltonian H(q, p), a function of coordinates q and momenta p. The Hamiltonian is obtained by performing a Legendre transform on the Lagrangian: H = pq̇ - L, where p = ∂L/∂q̇ is the generalized momentum. This transformation swaps the independent variables from (q, q̇) to (q, p). By equating differentials and using the definition of generalized momentum, we derive Hamilton's equations: dq/dt = ∂H/∂p and dp/dt = -∂H/∂q. These equations contain exactly the same physical information as the Euler-Lagrange equations from Lagrangian mechanics. The Legendre transform thus provides a powerful tool for converting between different mathematical representations of physical systems while preserving all dynamical information.

The Legendre transformation is a mathematical technique that converts a scalar function of one set of variables into a new function of its partial derivatives, serving as the bridge between Lagrangian and Hamiltonian mechanics; this transformation enables the derivation of Hamilton's canonical equations (q̇ = ∂H/∂p and ṗ = -∂H/∂q) by changing the independent variables from generalized coordinates and velocities to generalized coordinates and momenta, with the Hamiltonian H being defined as H = Σ(q̇ᵢpᵢ) - L, where L is the Lagrangian.

The Hamiltonian is obtained from the Lagrangian through a Legendre transformation that exchanges generalized velocities q̇_j with generalized momenta p_j. The transformation is: H(q_j, p_j, t) = Σ_j p_j q̇_j - L(q_j, q̇_j, t). The procedure involves: (1) Write H = Σ p_j q̇_j - L, (2) Express p_j = ∂L/∂q̇_j, (3) Solve for q̇_j in terms of p_j, (4) Substitute back to express H in terms of q_j, p_j, and t. For the harmonic oscillator: L = (1/2)mẋ² - (1/2)kx² gives H = p²/(2m) + (1/2)kx². For planetary motion: L = (1/2)m(ṙ² + r²θ̇²) - GMm/r gives H = p_r²/(2m) + p_θ²/(2mr²) - GMm/r. In most cases, H equals the total energy of the system.

This lecture demonstrates how Hamiltonian mechanics solves the simple harmonic oscillator problem by performing a Legendre transform on the Lagrangian to obtain the Hamiltonian H = p²/(2m) + (1/2)kx², then applying Hamilton's equations (ẋ = ∂H/∂p and ṗ = -∂H/∂x) to derive the equations of motion, ultimately revealing that phase space trajectories for conservative systems are closed ellipses that never cross each other due to energy conservation.