Matrices

Precise definition

An $m\times n$ matrix $A=(a_{ij})$ is a rectangular array with $m$ rows and $n$ columns over a stated field. Its size is part of the object. A matrix can store data, coefficients of a system, or a linear map from $\mathbb F^n$ to $\mathbb F^m$.

Notation and mathematical language

The entry $a_{ij}$ lies in row $i$, column $j$. Columns may be written $A=[a_1\ \cdots\ a_n]$. Square, diagonal, triangular, symmetric ($A^T=A$), identity, and zero matrices describe structural classes, not individual operations.

Conceptual picture

A matrix's meaning comes from context and chosen bases. The same numerical array can represent a data table or a transformation, while the same transformation has different matrices in different bases.

Conditions and key results

Rows and columns are ordered. Equality requires the same dimensions and equal corresponding entries. Operations have shape conditions: addition needs equal shape, multiplication needs matching inner dimensions, and inverses require square invertibility.

A reliable strategy

1. State the matrix dimensions and what rows, columns, and entries represent.
2. Use $a_{ij}$ with row index first and keep units consistent within intended operations.
3. Choose operations only after checking shapes and mathematical meaning.
4. Verify the output dimensions and interpret rows or columns rather than treating the array as context-free numbers.

Fully worked example

Interpretation and application

Matrices organize systems, networks, images, transitions, and datasets. Algebraic operations are meaningful only when the representation matches the question; multiplying two data tables merely because dimensions fit may have no sensible interpretation.

Common mistakes

Verification and reasonableness

• Count output entries and confirm the announced shape.
• Translate a matrix-vector product back into row dot products.
• Check special structure entrywise, such as $a_{ij}=a_{ji}$ for symmetry.

Practice

1. What is the size of a matrix with 3 rows and 5 columns?
2. Where is $a_{24}$?
3. What space does a $4\times2$ real matrix map from and to?

Further deduction

A matrix can be partitioned into blocks when their dimensions support the intended operations. Block multiplication follows ordinary multiplication, but block order cannot be changed. This is useful for augmented systems, transformations on direct sums, and structured algorithms; it is a notation device, not permission to ignore noncommutativity.

Sparse matrices store only nonzero entries and support algorithms whose cost depends on sparsity pattern rather than the full $mn$ array. A million-by-million matrix can be computationally manageable if each row has few nonzeros. Mathematical dimensions alone therefore do not determine memory or computation cost; structure matters.

An adjacency matrix encodes a finite graph: entry $a_{ij}$ records whether, or with what weight, an edge runs from vertex $i$ to vertex $j$. Powers then have combinatorial meaning, with $(A^k)_{ij}$ counting length-$k$ walks in an unweighted graph. Symmetry signals an undirected graph, while diagonal entries represent loops. This example shows that matrix entries acquire meaning from an indexing convention and the operation being modelled, not from the rectangular array alone.

Related topics

• Matrix Operations
• Matrix Multiplication
• Systems and Matrices