Skip to content

Linear Algebra Formula Sheet and Short Notes

One-page CSIR NET linear algebra revision: dimensions, rank and det rules, eigenvalues, Cayley–Hamilton, Jordan form, spectral theorem, quadratic forms.

  • 5 min read
  • Updated
On this page · 10 sections
  1. The paper in one table
  2. Vector spaces and linear maps
  3. Rank, determinants, systems
  4. Eigenvalues
  5. Canonical forms
  6. Inner product spaces
  7. Quadratic forms
  8. Counter-examples table
  9. Formula box
  10. Where to go next

Every definition, formula and counter-example from Vector Spaces, Bases and Linear Transformations, Rank, Determinants and Linear Equations, Eigenvalues and the Cayley–Hamilton Theorem, Change of Basis and Canonical Forms, Inner Product Spaces and Orthonormal Bases and Quadratic Forms, on one page.

The paper in one table

Per the NTA information bulletin for December 2026 (Mathematical Sciences, code 704):

PartGivenAttemptCorrectWrongMaximum
B4025+3−0.7575
C6020+4.75095

Part C: one or more options may be correct; marks only if you mark all correct options and no wrong one. No partial credit, no negative marking. Details on the Mathematical Sciences overview.

Vector spaces and linear maps

SpaceDimension
\( M_{m \times n}(F) \)\( mn \)
\( P_n \) (degree \( \le n \))\( n + 1 \)
symmetric / upper triangular \( n \times n \)\( \frac{n(n+1)}{2} \)
skew-symmetric\( \frac{n(n-1)}{2} \)
trace zero\( n^2 - 1 \)
Hermitian (over \( \mathbb{R} \))\( n^2 \)
\( \mathbb{C}^n \) over \( \mathbb{R} \)\( 2n \)
\( \mathcal{L}(V, W) \)\( \dim V \cdot \dim W \)
\( V/W \)\( \dim V - \dim W \)
$$\boxed{\dim(U + W) = \dim U + \dim W - \dim(U \cap W)} \qquad \boxed{\operatorname{rank} T + \operatorname{nullity} T = \dim(\text{domain})}$$
  • \( U \cup W \) is a subspace iff one contains the other.
  • Equal finite dimensions: injective \( \iff \) surjective \( \iff \) bijective (fails in infinite dimensions: \( D \) on \( \mathbb{R}[x] \)).
  • \( T^2 = 0 \Rightarrow \operatorname{im} T \subseteq \ker T \), rank \( \le n/2 \). Nilpotent \( N \Rightarrow I + N \) invertible.
  • Over \( \mathbb{F}_q \): \( \lvert GL_n \rvert = \prod_{i=0}^{n-1}(q^n - q^i) \); lines in \( \mathbb{F}_q^n \): \( \frac{q^n - 1}{q - 1} \). \( \lvert GL_2(\mathbb{F}_2) \rvert = 6 \), \( \lvert GL_2(\mathbb{F}_3) \rvert = 48 \), \( \lvert GL_3(\mathbb{F}_2) \rvert = 168 \).
  • \( X \mapsto X^T \) on \( M_n \): eigenvalues \( \pm 1 \), trace \( n \). \( X \mapsto AX - XA \), \( A = \operatorname{diag}(a_i) \): \( E_{ij} \mapsto (a_i - a_j)E_{ij} \).

Rank, determinants, systems

RuleFormula
Sylvester\( \operatorname{rank} A + \operatorname{rank} B - n \le \operatorname{rank}(AB) \le \min(\operatorname{rank} A, \operatorname{rank} B) \)
Sum\( \operatorname{rank}(A + B) \le \operatorname{rank} A + \operatorname{rank} B \)
Real Gram\( \operatorname{rank}(A^TA) = \operatorname{rank} A \) (fails over \( \mathbb{C} \) with \( ^T \), and over \( \mathbb{Z}_p \))
Scalar\( \det(cA) = c^n\det A \)
\( aI + bJ \)\( \det = a^{n-1}(a + nb) \)
Rank-one update\( \det(I + uv^T) = 1 + v^Tu \)
Vandermonde\( \prod_{i < j}(x_j - x_i) \)
Adjugate\( \det \operatorname{adj} A = (\det A)^{n-1} \); \( \operatorname{adj}\operatorname{adj} A = (\det A)^{n-2}A \)
Rank of adj\( n \), 1, 0 for rank \( n \), \( n - 1 \), \( \le n - 2 \)
Odd skew-symmetric real\( \det = 0 \)
Integer inverseiff \( \det = \pm 1 \)

Systems. Consistent \( \iff \operatorname{rank} A = \operatorname{rank}[A \mid b] \). Unique \( \iff \) also \( = n \). Free parameters \( n - r \). Over \( \mathbb{F}_q \): \( q^{n - r} \) solutions. Fewer equations than unknowns: homogeneous system has non-trivial solutions.

Eigenvalues

  • \( \chi_A(x) = x^n - (\operatorname{tr} A)x^{n-1} + \dots + (-1)^n\det A \); middle coefficients = sums of principal minors.
  • \( p(A) \) has eigenvalues \( p(\lambda_i) \); \( A^{-1} \): \( \frac{1}{\lambda_i} \); \( \operatorname{adj} A \): \( \frac{\det A}{\lambda_i} \).
  • \( \chi_{BA} = x^{n-m}\chi_{AB} \); \( AB \), \( BA \) need not be similar.
  • \( 1 \le g(\lambda) \le a(\lambda) \); \( g(\lambda) = n - \operatorname{rank}(A - \lambda I) \).
  • Odd-order real matrix: has a real eigenvalue. Gershgorin: eigenvalues in \( \lvert z - a_{ii} \rvert \le \sum_{j \ne i}\lvert a_{ij} \rvert \).
MatrixEigenvalues
\( J_n \)\( n, 0, \dots, 0 \)
\( uv^T \)\( v^Tu, 0, \dots, 0 \)
\( aI + bJ \)\( a + nb \) once, \( a \) otherwise
symmetric / Hermitianreal
skewimaginary
orthogonal / unitarymodulus 1
nilpotent / idempotent / involution\( 0 \) / \( 0, 1 \) / \( \pm 1 \)

Cayley–Hamilton: \( \chi_A(A) = 0 \). For \( 2 \times 2 \): \( A^2 = (\operatorname{tr} A)A - (\det A)I \), so \( A^{-1} = \frac{(\operatorname{tr} A)I - A}{\det A} \). Minimal polynomial: divides every annihilating polynomial; same roots as \( \chi_A \); exponent of \( (x - \lambda) \) = largest Jordan block.

Canonical forms

$$\boxed{[T]_{\text{new}} = P^{-1}[T]_{\text{old}}P}, \quad \text{columns of } P = \text{new basis in old coordinates}$$
  • Diagonalisable \( \iff g = a \) for all eigenvalues \( \iff m_A \) has distinct roots in \( F \). Sufficient: distinct eigenvalues; \( A^2 = A \); \( A^k = I \) over \( \mathbb{C} \); normal.
  • Triangularisable over \( F \) \( \iff \chi_A \) splits; Schur: unitarily over \( \mathbb{C} \).
  • Jordan: number of \( \lambda \)-blocks \( = n - r_1 \); blocks of size \( \ge k \): \( r_{k-1} - r_k \), with \( r_k = \operatorname{rank}(A - \lambda I)^k \), \( r_0 = n \).
  • Counting: product over eigenvalues of partitions of \( a(\lambda) \) with largest part fixed by \( m_A \). Nilpotent classes: \( p(n) = 1, 2, 3, 5, 7, 11, 15 \) for \( n = 1, \dots, 7 \).
  • \( \chi \) and \( m \) decide similarity only for \( n \le 3 \). \( A \sim A^T \) always.
  • \( J_2(\lambda)^m = \begin{pmatrix} \lambda^m & m\lambda^{m-1} \\ 0 & \lambda^m \end{pmatrix} \).

Inner product spaces

  • Cauchy–Schwarz \( \lvert \langle u, v \rangle \rvert \le \lVert u \rVert\lVert v \rVert \); norm from an inner product \( \iff \) parallelogram law.
  • Gram–Schmidt: \( u_j = v_j - \sum_{i
  • \( V = W \oplus W^\perp \). Projection onto \( a^\perp \): \( v - \frac{\langle v, a \rangle}{\lVert a \rVert^2}a \); distance \( \frac{\lvert \langle v, a \rangle \rvert}{\lVert a \rVert} \).
  • Orthogonal projection: \( P^2 = P = P^T \), \( \operatorname{tr} P = \operatorname{rank} P \); \( P = A(A^TA)^{-1}A^T \).
  • Least squares: \( A^TA\hat{x} = A^Tb \).
  • Orthogonal \( 3 \times 3 \): \( \det Q \) is an eigenvalue; rotation trace \( 1 + 2\cos\theta \). Householder \( I - 2uu^T \): \( \det = -1 \).
  • Spectral theorem: normal \( \iff \) unitarily diagonalisable; real symmetric \( \iff \) orthogonally diagonalisable.

Quadratic forms

  • Matrix: squares on the diagonal, half the cross coefficient off it. Change of variables: \( P^TAP \).
  • Inertia \( (p, q, z) \) = signs of eigenvalues; rank \( p + q \), signature \( p - q \) (some books: \( (p, q) \) or \( p \)).
  • Real symmetric congruent \( \iff \) same \( p, q \). Classes: \( \frac{(n+1)(n+2)}{2} \) (real), \( n + 1 \) (complex symmetric).
  • Positive definite \( \iff \) all eigenvalues \( > 0 \) \( \iff \) all leading minors \( > 0 \). Semidefinite: all principal minors \( \ge 0 \).
  • \( \lambda_{\min}\lVert x \rVert^2 \le x^TAx \le \lambda_{\max}\lVert x \rVert^2 \). Ellipse \( x^TAx = 1 \): area \( \frac{\pi}{\sqrt{\det A}} \).

Counter-examples table

Tempting claimBreaker
all eigenvalues 0 \( \Rightarrow A = 0 \)\( \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix} \)
all eigenvalues 1 \( \Rightarrow A = I \)\( \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} \)
same \( \chi \) \( \Rightarrow \) similar\( I_2 \) vs \( \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} \)
same \( \chi \) and \( m \) \( \Rightarrow \) similar\( J_2(0) \oplus J_2(0) \) vs \( J_2(0) \oplus 0 \oplus 0 \) (\( n = 4 \))
\( AB \sim BA \)\( A = E_{11} \), \( B = E_{12} \)
real matrix has a real eigenvaluerotation \( \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix} \) (even order)
orthogonal \( \Rightarrow \) diagonalisable over \( \mathbb{R} \)same rotation
real eigenvalues \( \Rightarrow \) symmetric\( \begin{pmatrix} 1 & 1 \\ 0 & 2 \end{pmatrix} \)
diagonalisable \( \Rightarrow \) unitarily diagonalisable\( \begin{pmatrix} 1 & 1 \\ 0 & 2 \end{pmatrix} \) (not normal)
distinct eigenvalues necessary for diagonalisability\( I \)
\( \det(A + B) = \det A + \det B \)\( A = B = I_2 \)
eigenvalues of \( A + B \) are sums\( E_{12} + E_{21} \): eigenvalues \( \pm 1 \) from \( 0 + 0 \)
\( \operatorname{rank}(AA^T) = \operatorname{rank} A \) always\( A = (1 \ \ i) \) over \( \mathbb{C} \): \( AA^T = 0 \)
leading minors \( \ge 0 \Rightarrow \) PSD\( \operatorname{diag}(0, -1) \)
\( \det A > 0 \Rightarrow \) positive definite\( -I_2 \)
same rank \( \Rightarrow \) congruent (over \( \mathbb{R} \))\( I_2 \) vs \( \operatorname{diag}(1, -1) \)
injective \( \iff \) surjectivedifferentiation on \( \mathbb{R}[x] \)
union of subspaces is a subspacethe two axes in \( \mathbb{R}^2 \)

Formula box

$$\det(\operatorname{adj} A) = (\det A)^{n-1} \qquad \det(aI + bJ) = a^{n-1}(a + nb) \qquad \det(I + uv^T) = 1 + v^Tu$$$$[T]_{B'} = P^{-1}[T]_BP \qquad A' = P^TAP \qquad A^TA\hat{x} = A^Tb$$$$\#\{\lambda\text{-blocks of size} \ge k\} = \operatorname{rank}(A - \lambda I)^{k-1} - \operatorname{rank}(A - \lambda I)^k$$

Where to go next

Advertisement
Advertisement