Chapter 37 — Further Reading
This chapter sits exactly on the border between linear algebra and differential equations, so the best companions come from both shelves. Strang is the standout for the linear-algebra-first treatment of the matrix exponential and stability — he presents $e^{At}$ and $\mathbf{x}'=A\mathbf{x}$ as the natural sequel to eigenvalues, which is precisely our framing.
Core linear-algebra textbooks
-
Gilbert Strang, Introduction to Linear Algebra (5th/6th ed.), the sections on "Differential Equations and $e^{At}$" — the primary companion to this chapter. Strang derives the matrix exponential, solves $\mathbf{x}'=A\mathbf{x}$ via eigenvectors and via $e^{At}$, and reads stability off the eigenvalues in exactly the spirit we adopted — eigenvalues as the destiny of the system. His treatment of the connection between continuous systems ($\operatorname{Re}\lambda<0$) and discrete powers ($|\lambda|<1$) is the natural next read after our §37.5 Common Pitfall. If you read one source on this chapter, read this.
-
Sheldon Axler, Linear Algebra Done Right (4th ed.). Axler's coordinate-free treatment of operators, the minimal polynomial, and generalized eigenvectors is the rigorous backbone behind our §37.6 (the defective case). His development of why $e^{At}$ and functions of operators are well-defined — and his clean account of nilpotent operators — is the math-major's complement to our Jordan-splitting computation. Best read alongside Chapter 36 of our book.
-
Stephen Boyd & Lieven Vandenberghe, Introduction to Applied Linear Algebra (VMLS). The application-forward view; its material on linear dynamical systems, state-space models, and discrete-time recurrences is the engineering face of this chapter. Freely available online, and the natural bridge to the control-systems applications of Case Study 2.
On the matrix exponential specifically
-
Cleve Moler & Charles Van Loan, "Nineteen Dubious Ways to Compute the Exponential of a Matrix" (SIAM Review, 1978; revisited 2003). The famous, readable, and slightly mischievous survey of how (not) to compute $e^{At}$ numerically. It is the authoritative explanation of why the diagonalization route $Pe^{Dt}P^{-1}$ fails for defective and near-defective matrices, and why
scipy.linalg.expm's scaling-and-squaring with Padé approximation is the trustworthy default. Essential reading before Chapter 38, and the source for our §37.3/§37.6 Computational Notes and Warnings. -
Nicholas Higham, Functions of Matrices: Theory and Computation. The definitive modern reference on $f(A)$ for $f=\exp$ and beyond — the deep version of "functions of a matrix" that Chapter 25 introduced and this chapter exploited. Graduate level; consult for the rigorous numerical-stability story.
On differential equations and dynamical systems
-
Morris Hirsch, Stephen Smale & Robert Devaney, Differential Equations, Dynamical Systems, and an Introduction to Chaos. The canonical text fusing linear algebra with ODEs. Its early chapters develop exactly our material — $e^{At}$, the classification of planar phase portraits (node/saddle/spiral/center) by eigenvalues, and the trace-determinant plane — and then push onward to nonlinear systems, where linearization (the Jacobian) reduces local behavior to the linear case of this chapter. The direct next step after §37.8.
-
Steven Strogatz, Nonlinear Dynamics and Chaos. The most readable introduction to phase portraits and stability ever written; Chapter 5 ("Linear Systems") is essentially a longer, picture-rich version of our §37.8, and the rest of the book shows how the eigenvalue classification governs the local behavior of nonlinear systems too. Highly recommended for building the geometric intuition for trajectories.
-
William Boyce & Richard DiPrima, Elementary Differential Equations and Boundary Value Problems. The standard sophomore ODE text; its chapter on systems of first-order linear equations covers the eigenvalue method, complex and repeated eigenvalues, and the fundamental matrix in conventional notation, providing extra worked examples for our §37.4–37.6. The relationship to single higher-order equations and the foundations are in any calculus sequence — see differential equations.
On control and stability (Case Study 2)
- Karl Åström & Richard Murray, Feedback Systems: An Introduction for Scientists and Engineers (free online). The modern, accessible control-theory text. Its treatment of state-space models $\mathbf{x}'=A\mathbf{x}$, eigenvalue (pole) placement, and the Routh–Hurwitz stability criterion is the full development of the "keep $\operatorname{Re}\lambda<0$" idea behind Case Study 2 — and shows how engineers design $A$ rather than merely analyze it.
Free online resources
- MIT OpenCourseWare 18.03 (Differential Equations) and 18.06 (Strang, Linear Algebra). 18.03 has dedicated lectures on the matrix exponential, phase portraits, and stability with beautiful applets; 18.06 connects $e^{At}$ to the eigenvalue machinery. Together they cover this chapter from both sides.
- 3Blue1Brown, Differential Equations series (esp. "Phase space" and the matrix-exponential video). The single best visual intuition for $e^{At}$, the flow on phase space, and why the eigenvalues govern it — watch the matrix-exponential episode after reading §37.2–37.4.
- For the quantum connection: the time-evolution operator $e^{-iHt}$ is developed in any quantum-mechanics course; see time evolution in quantum mechanics for how the matrix exponential of a Hermitian Hamiltonian drives the dynamics of states, with the eigenvalues of $H$ as energies.
Where to go next in this book
Chapter 38 (Numerical Linear Algebra) is the indispensable sequel: it explains, via the condition number and the distinction between mathematical correctness and algorithmic stability, precisely why the diagonalization route we used can fail on a finite machine and why scipy.linalg.expm is the robust choice — closing the loop on the Computational Notes that ran through this chapter. Then Chapter 39 (Capstone) puts the whole toolkit, including this chapter's matrix_exponential and solve_linear_ode, to work on a chosen application.