Linear Algebra

Jordan Normal Form

Visualize continuous phase portraits and canonical forms of 2x2 Jordan blocks.

Jordan Normal Form

Concept Overview

The Jordan Normal Form (JNF) is a nearly-diagonal representation of a square matrix. While not every matrix can be fully diagonalized, every square matrix over the complex numbers is similar to a block-diagonal matrix known as its Jordan Canonical Form. This decomposition separates a matrix into independent subsystems corresponding to distinct eigenvalues. When dealing with systems of linear differential equations, dx/dt = Ax, analyzing the equivalent system dx/dt = Jx allows for deep understanding of the continuous vector field phase portraits and the stability of the dynamical system.

Mathematical Definition

Every square matrix A is similar to a block diagonal matrix J:

A = P J P-1
Where J consists of Jordan blocks on its diagonal:
J = diag(J1, J2, ..., Jk)
Each Jordan block Ji corresponds to an eigenvalue λ:
Ji = [ λ 1 0 ... 0 ]
    [ 0 λ 1 ... 0 ]
    [ . . . ... . ]
    [ 0 0 0 ... λ ]

Key Concepts

Diagonalizable Matrices

A matrix is diagonalizable if its Jordan normal form is strictly diagonal (all 1s above the diagonal are 0). This occurs when the algebraic multiplicity of every eigenvalue equals its geometric multiplicity—meaning there are enough linearly independent eigenvectors to span the space. In a 2×2 system with real distinct eigenvalues λ1 and λ2, the phase portrait shows a node (sink or source) or saddle point.

Defective Matrices (Single Jordan Block)

When an eigenvalue has a geometric multiplicity strictly less than its algebraic multiplicity, the matrix is "defective" and cannot be fully diagonalized. For a 2×2 system with a repeated eigenvalue λ and only one independent eigenvector, the Jordan block contains a 1 on the superdiagonal. This lack of a second eigenvector results in an "improper node" in the phase portrait, where all trajectories approach the origin parallel to the single eigenvector direction. We must find a "generalized eigenvector" to form the basis P.

Complex Eigenvalues (Real Canonical Form)

Real matrices can have complex eigenvalues, which always appear in conjugate pairs a ± bi. Over the complex numbers, these form diagonal blocks. However, to keep the phase space entirely real, we use the real canonical form. The 2×2 block is formed by the real part (a) on the diagonal and the imaginary part (b) and its negative (-b) on the off-diagonals. The phase portrait shows a spiral or a center, representing rotational dynamics governed by Euler's formula e(a+bi)t = eat(cos(bt) + i sin(bt)).

Historical Context

The Jordan Normal Form was developed by the French mathematician Camille Jordan in 1870. His pivotal work "Traité des substitutions et des équations algébriques" generalized the concept of matrix diagonalization to handle cases where the characteristic polynomial has multiple roots but not enough independent eigenvectors. This canonical representation unified the study of linear differential equations and laid foundational groundwork for modern abstract algebra and the classification of Lie algebras.

Real-world Applications

  • Control Theory: Analyzing the stability of dynamic systems. The eigenvalues (specifically their real parts) determine whether a system returns to equilibrium (stable) or diverges exponentially (unstable).
  • Mechanical Vibrations: Defective matrices arise in critically damped systems where the eigenvalues converge. The presence of generalized eigenvectors introduces polynomial time factors (t eλt) into the solution.
  • Quantum Mechanics: Finding canonical representations of Hamiltonian operators. Degenerate energy levels often require dealing with subspaces similar to Jordan blocks.
  • Markov Chains: Analyzing the long-term behavior of transition matrices that are not fully diagonalizable, particularly determining the rate of convergence to stationary distributions.

Related Concepts

  • Eigenvalues and Eigenvectors — the basis for the decomposition.
  • Change of Basis — computing P to transform coordinates.
  • Matrix Exponentiation — solving dx/dt = Ax using eAt.
  • Systems of Linear Equations — finding the generalized eigenvectors.

Experience it interactively

Adjust parameters, observe in real time, and build deep intuition with Riano’s interactive Jordan Normal Form module.

Try Jordan Normal Form on Riano →

More in Linear Algebra