Matrix Algebra Rules

Why This Matters

Matrix algebra isn't just abstract symbol manipulation—it's the language that physical scientists use to describe everything from quantum mechanical states to coordinate transformations to systems of coupled differential equations. When you're tested on these rules, you're really being tested on whether you understand how linear transformations work and why certain operations preserve (or reveal) physical properties of systems.

The rules in this guide break down into a few core ideas: how matrices combine, what special matrices do, and how we extract meaningful information from matrices. Don't just memorize that matrix multiplication isn't commutative—understand why order matters when you're chaining transformations. Know which properties (like the trace) stay constant under similarity transformations, because that's how physicists identify quantities that don't depend on your choice of basis. Master these conceptual threads, and the individual rules will stick.

---

Basic Operations: How Matrices Combine

These operations form the foundation of everything else. Addition and scalar multiplication make matrices a vector space; matrix multiplication encodes composition of linear transformations.

Matrix Addition and Subtraction

• Same dimensions required—you can only add matrices that match in both rows and columns, just like adding vectors of the same length
• Element-wise operation means $(A + B)_{ij} = A_{ij} + B_{ij}$; no mixing of different positions occurs
• Commutative and associative—unlike multiplication, order doesn't matter: $A + B = B + A$

Scalar Multiplication

• Every element scales uniformly—multiplying matrix $A$ by scalar $c$ gives $(cA)_{ij} = c \cdot A_{ij}$
• Dimensions unchanged but the "magnitude" of the transformation changes; think of stretching or compressing
• Distributes over addition: $c(A + B) = cA + cB$, which is essential for linearity arguments

Matrix Multiplication

• Dimension compatibility rule: for $AB$ to exist, $A$ must be $m \times n$ and $B$ must be $n \times p$, giving result $m \times p$
• Row-column dot product defines each element: $(AB)_{ij} = \sum_{k} A_{ik}B_{kj}$
• Non-commutative—$AB \neq BA$ in general, because applying transformation $A$ then $B$ differs from $B$ then $A$

---

Structural Operations: Rearranging and Reshaping

These operations change how we view a matrix without changing its fundamental information content. Transposition swaps the role of rows and columns, which matters for inner products and adjoint operators.

Matrix Transposition

• Rows become columns—the transpose $A^T$ satisfies $(A^T)_{ij} = A_{ji}$
• Dimensions flip: an $m \times n$ matrix becomes $n \times m$ after transposition
• Key identity for products: $(AB)^T = B^T A^T$—the order reverses, which frequently appears in derivations

Trace of a Matrix

• Sum of diagonal elements: $\text{Tr}(A) = \sum_{i} A_{ii}$ for any square matrix
• Cyclic invariance—$\text{Tr}(ABC) = \text{Tr}(CAB) = \text{Tr}(BCA)$, a property physicists exploit constantly
• Basis-independent because trace equals the sum of eigenvalues; it's a similarity invariant

---

Special Matrices: Identity and Inverse

These matrices play the role of "1" and "1/x" from regular algebra. The identity does nothing; the inverse undoes a transformation.

Identity Matrix Properties

• Ones on diagonal, zeros elsewhere—denoted $I$ or $I_n$ for the $n \times n$ case
• Multiplicative neutral element: $AI = IA = A$ for any compatible matrix $A$
• Eigenvalues are all 1—every vector is an eigenvector, which is why $I$ represents "no transformation"

Inverse Matrix Properties

• Defining equation: $AA^{-1} = A^{-1}A = I$, meaning the inverse "undoes" the original transformation
• Exists only for square matrices with non-zero determinant—singular matrices ($\det(A) = 0$) have no inverse
• Product rule: $(AB)^{-1} = B^{-1}A^{-1}$—order reverses, just like with transposes

---

Matrix Invariants: Extracting Key Information

These quantities tell you about the matrix's behavior without requiring you to apply it to specific vectors. Determinants measure volume scaling; eigenvalues reveal stretching along special directions.

Determinant Calculation

• Invertibility test: $\det(A) \neq 0$ if and only if $A$ is invertible
• 2×2 formula: for $A = \begin{pmatrix} a & b \\ c & d \end{pmatrix}$, we have $\det(A) = ad - bc$
• Geometric meaning—the determinant measures how the transformation scales volumes (negative means orientation flip)

Eigenvalues and Eigenvectors

• Defining equation: $A\mathbf{v} = \lambda\mathbf{v}$, where $\lambda$ is the eigenvalue and $\mathbf{v}$ is the eigenvector
• Characteristic polynomial: eigenvalues are roots of $\det(A - \lambda I) = 0$
• Physical interpretation—eigenvectors are directions that only get stretched (by factor $\lambda$), not rotated

---

Advanced Structure: Diagonalization

Diagonalization is the payoff for understanding eigenvalues. When possible, it lets you express a matrix in its simplest form—pure stretching along eigenvector directions.

Matrix Diagonalization

• Decomposition form: $A = PDP^{-1}$, where $D$ is diagonal (eigenvalues on diagonal) and $P$ contains eigenvectors as columns
• Requires enough eigenvectors—$A$ is diagonalizable only if it has $n$ linearly independent eigenvectors for an $n \times n$ matrix
• Powers become trivial: $A^k = PD^kP^{-1}$, and raising a diagonal matrix to a power just raises each diagonal entry to that power

---

Quick Reference Table

---

Self-Check Questions

1. Which two operations follow an "order reversal" rule when applied to products, and why does this pattern occur?

2. You're given a $3 \times 3$ matrix and told its trace is 6 and its determinant is 8. What can you conclude about the sum and product of its eigenvalues?

3. Compare and contrast the conditions required for matrix addition versus matrix multiplication. Why are the requirements different?

4. A matrix has $\det(A) = 0$. What does this tell you about (a) its invertibility, (b) at least one of its eigenvalues, and (c) the geometric effect of the transformation?

5. If an FRQ asks you to compute $A^{100}$ for a given matrix, what technique should you reach for first, and what could prevent that technique from working?