Chapter 8 — Vector Spaces and Linear Transformations
The second half of the course begins: vector spaces, bases and dimension, linear maps and the matrices that represent them, rank, determinants and systems of linear equations.
In 1844 Hermann Grassmann, a secondary-school teacher in Stettin, published a book about "extensive magnitudes" that could be added and stretched in any number of dimensions. Almost nobody read it. Forty-four years later Giuseppe Peano extracted from it the short list of axioms we still use, and gave the first definition of a vector space — including spaces of functions, which have no finite dimension at all. The idea turned out to be everywhere: the solutions of a homogeneous linear system, of a linear differential equation and of a linear recurrence; the polynomials of degree at most $n$; the $m\times n$ matrices; the continuous functions on an interval; the signals and images stored by every computer. Whatever the objects are, the same theorems about subspaces, bases and dimension apply to all of them at once.
The second half of the story is about maps between vector spaces that respect addition and scaling. Arthur Cayley introduced matrix multiplication in 1858 precisely so that the product of two matrices would describe the composite of two linear substitutions. You have already met the most important linear map of analysis: in §6.1 the derivative of $f:\mathbb{R}^n\to\mathbb{R}^m$ at a point is a linear map, written as the Jacobian matrix, and the chain rule is the product of matrices. This chapter builds the algebra behind those facts: the dimension of a space and the rank of a map are the two numbers that control almost everything, from the solvability of $Ax=b$ to the number of planes in a finite space, and changing the basis changes a matrix $A$ into $P^{-1}AP$ without changing the map it describes — the starting point for the eigenvalues and canonical forms of Chapter 9.
- decide whether a subset of $\mathbb{F}^n$, of a polynomial, matrix or function space is a subspace, and whether given vectors are linearly independent — over the right field;
- prove the Steinitz exchange theorem and use it to compute the dimension of spaces such as the symmetric, skew-symmetric or trace-zero matrices and solution spaces of linear conditions;
- work with sums, direct sums, complements and quotient spaces, and apply the dimension formula $\dim(U+W)+\dim(U\cap W)$ $=\dim U+\dim W$;
- count vectors, bases and subspaces of a vector space over a finite field with the Gaussian binomial coefficients;
- find kernels and images, apply the rank–nullity theorem and decide when a linear map is injective, surjective or an isomorphism — and give the infinite-dimensional counterexamples;
- write the matrix of a linear map in any pair of bases, change bases with $P^{-1}AP$, and use the invariants of similar matrices; work with dual bases, projections and nilpotent maps;
- row-reduce, compute ranks and determinants, apply the rank inequalities, and decide consistency and the structure of the solution set of $Ax=b$.
How this is examined
This chapter feeds a large share of the linear-algebra questions in CSIR NET. Part B typically asks for the dimension of a described space (matrices with symmetry, trace or row-sum conditions; polynomials vanishing at given points; solutions of a recurrence), for the rank or nullity of a concrete map such as $X\mapsto AX-XA$ or $p\mapsto p''+p'$, for the matrix of a map in a stated basis, for a count over a finite field (subspaces, bases, invertible or surjective maps), or for the rank or determinant of a patterned matrix and the number of solutions of a system. Part C, where several options may be correct, prefers statements about all subspaces or all maps at once — "the union of two subspaces is a subspace", "two subspaces of these dimensions must intersect", "an injective operator is surjective", "$\ker T\cap\operatorname{im}T=\{0\}$", "matrices with the same trace and determinant are similar", "$\operatorname{rank}(AB)=\operatorname{rank}(BA)$" — and each false option is one of the standard counterexamples collected in the ⚠️ boxes of the three sections. GATE MA asks very similar questions.
Sections in this chapter
Vector Spaces, Bases and Dimension
Axioms and examples, subspaces, span, linear independence, the Steinitz exchange theorem, bases and dimension, sums and direct sums, the dimension formula, quotient spaces, counting over finite fields.
Linear Transformations and Their Matrices
Kernel and image, rank–nullity, isomorphisms, the algebra $\mathcal{L}(V,W)$, the matrix of a linear map, change of basis and similarity, dual spaces, projections and nilpotent maps.
Rank, Determinants and Linear Equations
Row reduction, row rank equals column rank, the Sylvester and Frobenius inequalities, determinants and the product rule, the adjugate and Cramer's rule, consistency of $Ax=b$, special matrices.
You should be at ease with matrices at the level of a first BSc course: matrix multiplication, Gaussian elimination for small systems, and $2\times2$ and $3\times3$ determinants (all of which are developed properly in §8.3). From this course you need the language of sets, functions, equivalence relations and countability (§1.1) — quotient spaces are built from an equivalence relation, and a countability argument shows that $\mathbb{R}$ is infinite-dimensional over $\mathbb{Q}$. The derivative as a linear map (§6.1) and normed linear spaces (§7.4) are good motivation but are not required. Fields are used only through their axioms; finite fields $\mathbb{F}_p=\mathbb{Z}_p$ are treated in Unit 2 (Algebra), but nothing beyond arithmetic modulo a prime is needed here.