Reference

Formula Sheet

Every key formula of the course on one page, grouped by chapter, with a link back to the section that develops it.

Chapter 1 · Calculus of Several Variables

Directional derivative and gradient ($f$ differentiable, $\|u\|=1$)§ 1.1 →
$D_uf(a)=\nabla f(a)\cdot u$, $\ \max_uD_uf(a)=\|\nabla f(a)\|$ along $\nabla f(a)$; tangent plane to $F=c$ at $a$: $\nabla F(a)\cdot(x-a)=0$
Implications at a point§ 1.1 →
$C^1\Rightarrow$ differentiable $\Rightarrow$ continuous and all $D_uf=\nabla f\cdot u$; none of these arrows reverses, and "partials exist" implies none of them
Chain rule and Jacobians§ 1.1 →
$J_{f\circ g}(a)=J_f(g(a))\,J_g(a)$; $\ \dfrac{\partial(u,v)}{\partial(x,y)}\cdot\dfrac{\partial(x,y)}{\partial(u,v)}=1$
Euler's theorem ($f$ homogeneous of degree $k$)§ 1.1 →
$xf_x+yf_y=kf$; $\ x^2f_{xx}+2xyf_{xy}+y^2f_{yy}=k(k-1)f$ (for $f\in C^2$)
Implicit derivatives (nonzero denominators)§ 1.1 →
$\dfrac{dy}{dx}=-\dfrac{F_x}{F_y}$; $\ z_x=-\dfrac{F_x}{F_z}$, $z_y=-\dfrac{F_y}{F_z}$; $\ u_x=-\dfrac{\partial(F,G)/\partial(x,v)}{\partial(F,G)/\partial(u,v)}$
Second derivative test, two variables ($\nabla f(a)=0$, $f\in C^2$)§ 1.2 →
$D=f_{xx}f_{yy}-f_{xy}^2$: $\ D\gt0,f_{xx}\gt0$ min; $\ D\gt0,f_{xx}\lt0$ max; $\ D\lt0$ saddle; $\ D=0$ no conclusion
Three variables: leading minors of $H_f(a)$§ 1.2 →
$\Delta_1,\Delta_2,\Delta_3$ all $\gt0$: min; signs $-,+,-$: max; $\Delta_3\ne0$ otherwise: saddle; $\Delta_3=0$: test fails
Lagrange multipliers (independent constraint gradients)§ 1.2 →
$\nabla f=\lambda\nabla g,\ g=0$; $\ \nabla f=\lambda\nabla g_1+\mu\nabla g_2,\ g_1=g_2=0$; add points where $\nabla g=0$
Bordered Hessian ($n=2$, one constraint)§ 1.2 →
$\det\begin{pmatrix}0&g_x&g_y\\g_x&\mathcal{L}_{xx}&\mathcal{L}_{xy}\\g_y&\mathcal{L}_{xy}&\mathcal{L}_{yy}\end{pmatrix}$: $\gt0$ max, $\lt0$ min
Quadratic form on the unit sphere ($A$ symmetric)§ 1.2 →
$\lambda_{\min}\le x^{\mathsf T}Ax\le\lambda_{\max}$ for $\|x\|=1$, with equality at eigenvectors
Fubini on regions ($f$ continuous)§ 1.3 →
$\iint_Df\,dA=\int_a^b\!\int_{g_1(x)}^{g_2(x)}f\,dy\,dx=\int_c^d\!\int_{h_1(y)}^{h_2(y)}f\,dx\,dy$ when $D$ is described both ways
Change of variables ($T$ injective, $C^1$, $J\ne0$ inside)§ 1.3 →
$\iint_Df\,dx\,dy=\iint_{D^*}f(x(u,v),y(u,v))\left|\dfrac{\partial(x,y)}{\partial(u,v)}\right|du\,dv$; polar: $dA=r\,dr\,d\theta$
Volume elements§ 1.3 →
$dV=r\,dr\,d\theta\,dz=\rho^2\sin\varphi\,d\rho\,d\varphi\,d\theta$ ($0\le\varphi\le\pi$ from the $z$-axis)
Mass, centroid, Pappus§ 1.3 →
$M=\iint_D\delta\,dA$, $\ \bar x=\frac1M\iint_Dx\,\delta\,dA$; solid of revolution: $V=2\pi\bar x\operatorname{Area}(D)$
Surface area§ 1.3 →
$A=\iint_D|\mathbf{r}_u\times\mathbf{r}_v|\,du\,dv$; graph: $\iint_D\sqrt{1+g_x^2+g_y^2}\,dA$; revolution: $2\pi\int_a^bf\sqrt{1+f'^2}\,dx$; sphere zone of height $h$: $2\pi ah$
Gaussian and improper integrals§ 1.3 →
$\int_{\mathbb{R}}e^{-x^2}dx=\sqrt\pi$; $\int_{\mathbb{R}^n}e^{-x^{\mathsf T}Ax}dx=\pi^{n/2}/\sqrt{\det A}$ ($A\gt0$); $\iint_{\mathbb{R}^2}\frac{dA}{(1+r^2)^p}=\frac\pi{p-1}$ iff $p\gt1$

Chapter 2 · Vector Calculus

Gradient and directional derivative§ 2.1 →
$D_{\mathbf{u}}f=\nabla f\cdot\mathbf{u}$ ($|\mathbf{u}|=1$); max rate $|\nabla f|$ along $\nabla f$; tangent plane to $f=c$ at $\mathbf{p}$: $\nabla f(\mathbf{p})\cdot(\mathbf{x}-\mathbf{p})=0$ (if $\nabla f(\mathbf{p})\ne\mathbf{0}$)
Divergence, curl, Laplacian§ 2.1 →
$\nabla\cdot\mathbf{F}=P_x+Q_y+R_z$, $\ \nabla\times\mathbf{F}=(R_y-Q_z,\ P_z-R_x,\ Q_x-P_y)$, $\ \Delta f=f_{xx}+f_{yy}+f_{zz}$; planar scalar curl $Q_x-P_y$
Identities and product rules ($C^2$ fields)§ 2.1 →
$\nabla\times\nabla f=\mathbf{0}$, $\ \nabla\cdot(\nabla\times\mathbf{F})=0$, $\ \nabla\times(\nabla\times\mathbf{F})=\nabla(\nabla\cdot\mathbf{F})-\Delta\mathbf{F}$; $\ \nabla\cdot(f\mathbf{F})=f\nabla\cdot\mathbf{F}+\nabla f\cdot\mathbf{F}$, $\ \nabla\times(f\mathbf{F})=f\nabla\times\mathbf{F}+\nabla f\times\mathbf{F}$, $\ \nabla\cdot(\mathbf{F}\times\mathbf{G})=\mathbf{G}\cdot\nabla\times\mathbf{F}-\mathbf{F}\cdot\nabla\times\mathbf{G}$
Radial fields in $\mathbb{R}^3$ ($r=|\mathbf{r}|\ne0$)§ 2.1 →
$\nabla r^n=nr^{n-2}\mathbf{r}$, $\ \nabla\cdot(r^n\mathbf{r})=(n+3)r^n$, $\ \nabla\times(r^n\mathbf{r})=\mathbf{0}$, $\ \Delta r^n=n(n+1)r^{n-2}$; $\ \nabla\cdot\mathbf{r}=3$
Line integrals and the fundamental theorem§ 2.1 →
$\int_Cf\,ds=\int_a^bf(\mathbf{r})|\mathbf{r}'|\,dt$, $\ \int_C\mathbf{F}\cdot d\mathbf{r}=\int_a^b\mathbf{F}(\mathbf{r})\cdot\mathbf{r}'\,dt$; $\ \int_C\nabla f\cdot d\mathbf{r}=f(B)-f(A)$
Conservative fields§ 2.1 →
On a domain: $\mathbf{F}=\nabla f$ $\iff$ $\oint\mathbf{F}\cdot d\mathbf{r}=0$ for all closed paths $\iff$ path independence. $C^1$ conservative $\Rightarrow\nabla\times\mathbf{F}=\mathbf{0}$; converse on simply connected domains only — $(-y,x)/(x^2+y^2)$ on $\mathbb{R}^2\setminus\{0\}$ has curl $0$ and circulation $2\pi$.
Green's theorem ($P,Q\in C^1$ near $D$, $\partial D$ positively oriented)§ 2.2 →
$\oint_{\partial D}P\,dx+Q\,dy=\iint_D(Q_x-P_y)\,dA$; $\ \oint_{\partial D}\mathbf{F}\cdot\mathbf{n}\,ds=\iint_D(P_x+Q_y)\,dA$; $\ \operatorname{Area}(D)=\tfrac12\oint(x\,dy-y\,dx)$
Surface area element§ 2.2 →
$dS=|\mathbf{r}_u\times\mathbf{r}_v|\,du\,dv$; graph $z=g$: $\sqrt{1+g_x^2+g_y^2}\,dx\,dy$; sphere radius $a$: $a^2\sin\varphi\,d\varphi\,d\theta$; cylinder radius $a$: $a\,d\theta\,dz$
Flux through an oriented surface§ 2.2 →
$\iint_S\mathbf{F}\cdot\mathbf{n}\,dS=\iint_D\mathbf{F}(\mathbf{r})\cdot(\mathbf{r}_u\times\mathbf{r}_v)\,du\,dv$; graph $z=g$, upwards: $\iint_D(-Pg_x-Qg_y+R)\,dx\,dy$
Stokes' theorem§ 2.2 →
$\oint_{\partial S}\mathbf{F}\cdot d\mathbf{r}=\iint_S(\nabla\times\mathbf{F})\cdot\mathbf{n}\,dS$ (right-hand rule; $\mathbf{F}\in C^1$ near $S$)
Divergence theorem§ 2.2 →
$\iint_{\partial E}\mathbf{F}\cdot\mathbf{n}\,dS=\iiint_E\nabla\cdot\mathbf{F}\,dV$ (outward $\mathbf{n}$, $\mathbf{F}\in C^1$ near $E$); $\ \operatorname{vol}(E)=\tfrac13\iint_{\partial E}\mathbf{r}\cdot\mathbf{n}\,dS$
Gauss's law and Green's identities§ 2.2 →
$\iint_{\partial E}\frac{\mathbf{r}\cdot\mathbf{n}}{r^3}\,dS=4\pi$ if $\mathbf{0}$ is inside $E$, $0$ if outside; $\ \iiint_E(f\Delta g+\nabla f\cdot\nabla g)\,dV=\iint_{\partial E}f\,\partial_ng\,dS$

Chapter 3 · Linear Algebra

Rank–nullity and the dimension formula ($\dim V\lt\infty$)§ 3.1 →
$\operatorname{rank}T+\operatorname{nullity}T=\dim V$; $\ \dim(U+W)=\dim U+\dim W-\dim(U\cap W)$
Counting over $\mathbb{F}_q$ ($\dim V=n$)§ 3.1 →
$|GL_n(\mathbb{F}_q)|=\prod_{i=0}^{n-1}(q^n-q^i)$; $\ \#\{k\text{-dim subspaces}\}=\prod_{i=0}^{k-1}\dfrac{q^n-q^i}{q^k-q^i}$
Change of basis ($P$: new basis in old coordinates)§ 3.1 →
$[v]_{\mathcal{B}}=P[v]_{\mathcal{B}'}$, $\ [T]_{\mathcal{B}'}=P^{-1}[T]_{\mathcal{B}}P$
Rouché–Capelli ($A$ is $m\times n$)§ 3.1 →
$Ax=b$ solvable $\iff\operatorname{rank}A=\operatorname{rank}[A\,|\,b]$; then solutions $=x_p+N(A)$ with $n-\operatorname{rank}A$ free parameters; unique iff $\operatorname{rank}A=n$
Rank rules ($A$: $m\times n$, $B$: $n\times p$)§ 3.1 →
$\operatorname{rank}A+\operatorname{rank}B-n\le\operatorname{rank}(AB)\le\min(\operatorname{rank}A,\operatorname{rank}B)$; $\ \operatorname{rank}(A^{\mathsf T}A)=\operatorname{rank}A$ (real $A$)
Characteristic polynomial ($A\in M_n$, eigenvalues $\lambda_i$ over $\mathbb{C}$)§ 3.2 →
$p_A(x)=\det(xI-A)=x^n-(\operatorname{tr}A)x^{n-1}+\cdots+(-1)^n\det A$; $\ \sum\lambda_i=\operatorname{tr}A$, $\prod\lambda_i=\det A$
Eigenvalue shortcuts§ 3.2 →
$\sigma(f(A))=\{f(\lambda_i)\}$; $\ p_{uv^{\mathsf T}}=x^{n-1}(x-v^{\mathsf T}u)$; $\ aI+bJ$: $a+nb$, $a$ ($n-1$ times); $\ p_{AB}=p_{BA}$ (square)
Cayley–Hamilton and the minimal polynomial§ 3.2 →
$p_A(A)=0$; $\ f(A)=0\iff m_A\mid f$; $\ m_A\mid p_A$ with the same roots; $\ m_{A\oplus B}=\operatorname{lcm}(m_A,m_B)$
Diagonalisable over $\mathbb{F}$§ 3.2 →
$\iff\sum_\lambda g(\lambda)=n\iff$ ($p_A$ splits and $g(\lambda)=a(\lambda)$ for all $\lambda$) $\iff m_A=\prod(x-\lambda_i)$, $\lambda_i$ distinct
Jordan form from ranks ($r_k=\operatorname{rank}(A-\lambda I)^k$, $r_0=n$)§ 3.2 →
#blocks $=g(\lambda)=n-r_1$; #blocks of size $\ge k$ $=r_{k-1}-r_k$; largest block = exponent of $x-\lambda$ in $m_A$
Counting similarity classes in $M_n(\mathbb{C})$§ 3.2 →
$p_A=\prod(x-\lambda_i)^{a_i}$, $m_A=\prod(x-\lambda_i)^{m_i}$: $\prod_iP(a_i,m_i)$ classes; nilpotent: $p(n)$; with $N^k=0$: partitions of $n$ into parts $\le k$
Gram–Schmidt and projection (independent $v_1,\dots,v_k$)§ 3.3 →
$u_j=v_j-\sum_{i\lt j}\frac{\langle v_j,u_i\rangle}{\|u_i\|^2}u_i$; $\ P_Wx=\sum\langle x,e_i\rangle e_i$, $\ \operatorname{dist}(x,W)^2=\|x\|^2-\sum|\langle x,e_i\rangle|^2$
Eigenvalues of the special classes§ 3.3 →
$A^*=A$: $\lambda\in\mathbb{R}$; $\ A^*=-A$: $\lambda\in i\mathbb{R}$; $\ U^*U=I$: $|\lambda|=1$; $\ 3\times3$ rotation: $\operatorname{tr}Q=1+2\cos\theta$
Spectral theorems§ 3.3 →
$A^*A=AA^*\iff A=UDU^*$ ($U$ unitary); $\ A=A^{\mathsf T}$ real $\iff A=QDQ^{\mathsf T}$ ($Q$ real orthogonal)
Schur's inequality (any square $A$)§ 3.3 →
$\sum_j|\lambda_j|^2\le\sum_{i,j}|a_{ij}|^2$, equality iff $A$ normal
Definiteness ($A$ Hermitian or real symmetric)§ 3.3 →
PD $\iff$ all $\lambda\gt0\iff\Delta_1,\dots,\Delta_n\gt0$; $\ $ PSD $\iff$ all $\lambda\ge0\iff$ all principal minors $\ge0$
Inertia, signature, Rayleigh, ellipse area§ 3.3 →
$r=n_++n_-$, $\ s=n_+-n_-$; $\ \lambda_{\min}\le\frac{x^*Ax}{x^*x}\le\lambda_{\max}$; $\ x^{\mathsf T}Ax=c$: area $\frac{\pi c}{\sqrt{\det A}}$

Chapter 4 · Real Analysis

Contraction principle ($X$ complete, $d(Tx,Ty)\le k\,d(x,y)$, $k\lt1$)§ 4.1 →
unique $x^*=Tx^*$; $\ d(x_n,x^*)\le\dfrac{k^n}{1-k}d(x_1,x_0)$, $\ d(x_n,x^*)\le\dfrac{k}{1-k}d(x_n,x_{n-1})$
Derivative test ($g:[a,b]\to[a,b]$)§ 4.1 →
$|g'|\le k\lt1$ on $[a,b]\Rightarrow x_{n+1}=g(x_n)\to$ the unique fixed point
Metrics on $\mathbb{R}^n$§ 4.1 →
$d_\infty\le d_2\le d_1\le n\,d_\infty$
Compactness in a metric space§ 4.1 →
compact $\iff$ sequentially compact $\iff$ complete + totally bounded; $\ $ in $\mathbb{R}^n$: $\iff$ closed + bounded
Lipschitz constant of $v\mapsto Av$§ 4.1 →
$d_\infty$: $\max_i\sum_j|a_{ij}|$; $\ d_1$: $\max_j\sum_i|a_{ij}|$
Uniform convergence§ 4.2 →
$f_n\rightrightarrows f$ on $D\iff M_n=\sup_D|f_n-f|\to0$
Weierstrass M-test§ 4.2 →
$|u_n|\le M_n$ on $D$, $\sum M_n\lt\infty\Rightarrow\sum u_n$ uniformly convergent
Interchange theorems (on $[a,b]$)§ 4.2 →
$f_n\rightrightarrows f\Rightarrow\int_a^bf_n\to\int_a^bf$; $\ f_n'\rightrightarrows g$, $f_n(x_0)$ convergent $\Rightarrow(\lim f_n)'=g$
Radius of convergence of $\sum a_n(x-c)^n$§ 4.2 →
$\dfrac1R=\limsup|a_n|^{1/n}$; $\ \dfrac1R=\lim\left|\dfrac{a_{n+1}}{a_n}\right|$ if this limit exists
Bernstein polynomials (Weierstrass)§ 4.2 →
$B_nf(x)=\sum_{k=0}^nf(\tfrac kn)\binom nkx^k(1-x)^{n-k}\rightrightarrows f$; $\ B_n(x^2)=x^2+\dfrac{x(1-x)}n$
Sums from $\sum x^n$ ($|x|\lt1$)§ 4.2 →
$\sum nx^n=\dfrac x{(1-x)^2}$, $\ \sum n^2x^n=\dfrac{x(1+x)}{(1-x)^3}$, $\ \sum\dfrac{x^n}n=-\log(1-x)$
Outer measure; null sets§ 4.3 →
$m^*(A)=\inf\{\sum_k\ell(I_k):A\subseteq\bigcup_kI_k\}$; $\ m^*(I)=\ell(I)$; countable sets are null; $m(A\cup N)=m(A)$ if $N$ null
Continuity of measure; Cantor sets§ 4.3 →
$E_k\uparrow$: $m(\bigcup E_k)=\lim m(E_k)$; $\ E_k\downarrow$, $m(E_1)\lt\infty$: $m(\bigcap E_k)=\lim m(E_k)$; $\ m(C)=0$, $m(K_\alpha)=1-\alpha$
Markov's inequality ($f\ge0$ measurable, $t\gt0$)§ 4.3 →
$m(\{f\ge t\})\le\frac1t\int f$
MCT and Fatou ($f_n\ge0$ measurable)§ 4.3 →
$f_n\uparrow f\Rightarrow\int f_n\to\int f$; $\quad\int\liminf f_n\le\liminf\int f_n$
DCT and term-by-term integration§ 4.3 →
$f_n\to f$ a.e., $|f_n|\le g\in\mathcal L^1\Rightarrow\int f_n\to\int f$; $\ \sum\int|g_k|\lt\infty\Rightarrow\int\sum g_k=\sum\int g_k$
Riemann versus Lebesgue§ 4.3 →
Bounded $f$ is Riemann integrable iff its discontinuities form a null set; then both integrals agree. Improper $\int_a^\infty f$ is a Lebesgue integral iff $\int_a^\infty|f|\lt\infty$

Chapter 5 · Complex Analysis

Cauchy–Riemann equations ($f=u+iv$ differentiable at the point)§ 5.1 →
$u_x=v_y,\ u_y=-v_x$; $\ f'=u_x+iv_x$; polar: $u_r=\tfrac1rv_\theta,\ v_r=-\tfrac1ru_\theta$
Milne–Thomson (given $u$)§ 5.1 →
$f'(z)=u_x(z,0)-i\,u_y(z,0)$
Basic integral ($n\in\mathbb{Z}$)§ 5.1 →
$\oint_{|z-a|=r}(z-a)^n\,dz=2\pi i$ if $n=-1$, else $0$
Cauchy's integral formula ($f$ analytic on and inside $\gamma$, $a$ inside)§ 5.1 →
$f^{(n)}(a)=\dfrac{n!}{2\pi i}\displaystyle\oint_\gamma\frac{f(z)}{(z-a)^{n+1}}\,dz$
Cauchy's estimates ($|f|\le M$ on $|z-a|=R$)§ 5.1 →
$|f^{(n)}(a)|\le n!\,M/R^n$
ML-inequality§ 5.1 →
$\left|\int_\gamma f\,dz\right|\le\max_\gamma|f|\cdot L(\gamma)$
Radius of convergence§ 5.2 →
$\dfrac1R=\limsup|c_n|^{1/n}$; $\ R=\lim\left|\dfrac{c_n}{c_{n+1}}\right|$ if it exists
Residues at poles§ 5.2 →
simple: $\operatorname{Res}_a\dfrac pq=\dfrac{p(a)}{q'(a)}$; $\ $order $\le m$: $\dfrac{1}{(m-1)!}\lim\limits_{z\to a}\dfrac{d^{m-1}}{dz^{m-1}}\big[(z-a)^mf\big]$
Residue theorem; residue at $\infty$§ 5.2 →
$\oint_\gamma f\,dz=2\pi i\sum\operatorname{Res}_{a_k}f$; $\ \operatorname{Res}_\infty f=-\operatorname{Res}_0\big[w^{-2}f(1/w)\big]$; $\ \sum\operatorname{Res}+\operatorname{Res}_\infty=0$
Trigonometric integral ($a\gt|b|$)§ 5.2 →
$\displaystyle\int_0^{2\pi}\frac{d\theta}{a+b\cos\theta}=\frac{2\pi}{\sqrt{a^2-b^2}}$
Real-line integrals ($Q$ has no real zeros; $\deg Q\ge\deg P+2$; with $e^{ikx}$, $k\gt0$: $\deg Q\ge\deg P+1$)§ 5.2 →
$\displaystyle\int_{-\infty}^\infty\frac{P}{Q}e^{ikx}dx=2\pi i\sum_{\operatorname{Im}a\gt0}\operatorname{Res}_a\Big[\frac PQe^{ikz}\Big]$
Argument principle; Rouché ($|g|\lt|f|$ on $\gamma$)§ 5.2 →
$\dfrac1{2\pi i}\displaystyle\oint_\gamma\frac{f'}{f}dz=Z-P$; $\ Z(f+g)=Z(f)$ inside $\gamma$
Schwarz and Schwarz–Pick ($f$ analytic, $f(\mathbb{D})\subseteq\mathbb{D}$)§ 5.3 →
$f(0)=0\Rightarrow|f(z)|\le|z|,\ |f'(0)|\le1$; $\ \ |f'(z)|\le\dfrac{1-|f(z)|^2}{1-|z|^2}$
Automorphisms of the disc ($|a|\lt1$)§ 5.3 →
$e^{i\theta}\dfrac{z-a}{1-\bar az}$; $\ \varphi_a'(0)=1-|a|^2,\ \varphi_a'(a)=\dfrac1{1-|a|^2}$
Möbius map through three points§ 5.3 →
$(w,w_1,w_2,w_3)=(z,z_1,z_2,z_3)$, $\ (z,z_1,z_2,z_3)=\dfrac{(z-z_1)(z_2-z_3)}{(z-z_3)(z_2-z_1)}$
Fixed points of $\frac{az+b}{cz+d}$ ($c\ne0$)§ 5.3 →
$cz^2+(d-a)z-b=0$; one root iff $(a+d)^2=4(ad-bc)$ (parabolic)
Half-plane and disc§ 5.3 →
$\mathbb{H}\to\mathbb{D}$: $e^{i\theta}\dfrac{z-a}{z-\bar a}$ ($a\in\mathbb{H}$); $\ \operatorname{Aut}(\mathbb{H})$: $\dfrac{az+b}{cz+d}$, $a,b,c,d\in\mathbb{R}$, $ad-bc\gt0$
Standard maps§ 5.3 →
$z^{\pi/\theta_0}$: sector $\to\mathbb{H}$; $\ e^z$: $\{0\lt\operatorname{Im}z\lt\pi\}\to\mathbb{H}$; $\ \frac{1+z}{1-z}$: $\mathbb{D}\to\{\operatorname{Re}w\gt0\}$; $\ \frac12(z+\frac1z)$: $\mathbb{D}\setminus\{0\}\to\mathbb{C}\setminus[-1,1]$

Chapter 6 · Ordinary Differential Equations

Exact equations and integrating factors ($M,N\in C^1$ on a rectangle)§ 6.1 →
Exact $\iff M_y=N_x$; $\ \mu=e^{\int\frac{M_y-N_x}{N}dx}$ if that ratio depends on $x$ only; $\ \mu=e^{\int\frac{N_x-M_y}{M}dy}$ if it depends on $y$ only
Linear and Bernoulli equations ($P,Q$ continuous)§ 6.1 →
$y'+Py=Q$: $\ y=e^{-\int P}\left(\int e^{\int P}Q\,dx+C\right)$; $\ y'+Py=Qy^n$: put $v=y^{1-n}$
Picard–Lindelöf ($f$ continuous, Lipschitz in $y$ on $|x-x_0|\le a$, $|y-y_0|\le b$)§ 6.1 →
Unique solution on $|x-x_0|\le h=\min\left(a,\tfrac bM\right)$, $M=\max|f|$; $\ \varphi_{n+1}=y_0+\int_{x_0}^xf(t,\varphi_n)\,dt$
Abel's formula (normal form)§ 6.1 →
$W(x)=W(x_0)\,e^{-\int_{x_0}^{x}p_{n-1}(t)\,dt}$: never zero or identically zero
Variation of parameters for $y''+py'+qy=f$ (normal form!)§ 6.1 →
$y_p=-y_1\displaystyle\int\frac{y_2f}{W}\,dx+y_2\int\frac{y_1f}{W}\,dx$, $\ W=y_1y_2'-y_1'y_2$
Cauchy–Euler $x^2y''+axy'+by=0$, $x\gt0$§ 6.1 →
$r(r-1)+ar+b=0$: $x^{r_1},x^{r_2}$; $\ x^r,\ x^r\log x$; $\ x^\alpha\cos(\beta\log x),\ x^\alpha\sin(\beta\log x)$
Basic Laplace transforms§ 6.2 →
$t^n\leftrightarrow\dfrac{n!}{s^{n+1}}$, $\ t^\nu\leftrightarrow\dfrac{\Gamma(\nu+1)}{s^{\nu+1}}$, $\ e^{at}\leftrightarrow\dfrac1{s-a}$, $\ \cos bt\leftrightarrow\dfrac{s}{s^2+b^2}$, $\ \sin bt\leftrightarrow\dfrac{b}{s^2+b^2}$
Shifting theorems ($a\ge0$ in the second)§ 6.2 →
$e^{at}f(t)\leftrightarrow F(s-a)$; $\ f(t-a)u(t-a)\leftrightarrow e^{-as}F(s)$; $\ u(t-a)\leftrightarrow\dfrac{e^{-as}}s$; $\ \delta(t-a)\leftrightarrow e^{-as}$
Derivatives, integrals, powers of $t$ ($f$ continuous for the derivative rule)§ 6.2 →
$y''\leftrightarrow s^2Y-sy(0)-y'(0)$; $\ \int_0^tf\leftrightarrow\dfrac Fs$; $\ t^nf\leftrightarrow(-1)^nF^{(n)}$; $\ \dfrac{f}{t}\leftrightarrow\int_s^\infty F$
Convolution theorem§ 6.2 →
$(f*g)(t)=\int_0^tf(\tau)g(t-\tau)\,d\tau\leftrightarrow F(s)G(s)$
Periodic $f$ with period $T$§ 6.2 →
$F(s)=\dfrac{1}{1-e^{-sT}}\displaystyle\int_0^Te^{-st}f(t)\,dt$
Initial and final value theorems (final: only if $\lim_{t\to\infty}f$ exists, e.g. all poles of $sF$ in $\operatorname{Re}s\lt0$)§ 6.2 →
$f(0)=\lim_{s\to\infty}sF(s)$; $\ \lim_{t\to\infty}f(t)=\lim_{s\to0^+}sF(s)$
Frobenius at a regular singular point $0$ ($p_0=\lim xP$, $q_0=\lim x^2Q$)§ 6.3 →
$r(r-1)+p_0r+q_0=0$; $\ r_1-r_2\notin\mathbb{Z}$: two series $x^{r_i}\sum a_kx^k$; $\ r_1=r_2$: $y_2=y_1\log x+\cdots$; $\ r_1-r_2\in\mathbb{N}$: $\log$ possible
Series solution at an ordinary point§ 6.3 →
radius $\ge$ distance from $x_0$ to the nearest complex zero of $P_0$ in $P_0y''+P_1y'+P_2y=0$
Legendre polynomials§ 6.3 →
$P_n=\frac{1}{2^nn!}\frac{d^n}{dx^n}(x^2-1)^n$; $\ \sum P_nt^n=(1-2xt+t^2)^{-1/2}$; $\ (n+1)P_{n+1}=(2n+1)xP_n-nP_{n-1}$; $\ \int_{-1}^1P_mP_n=\frac{2\delta_{mn}}{2n+1}$
Bessel functions $J_\nu$§ 6.3 →
$(x^\nu J_\nu)'=x^\nu J_{\nu-1}$; $\ (x^{-\nu}J_\nu)'=-x^{-\nu}J_{\nu+1}$; $\ J_{\nu-1}+J_{\nu+1}=\frac{2\nu}xJ_\nu$; $\ J_0'=-J_1$; $\ J_{\pm1/2}=\sqrt{\frac2{\pi x}}\{\sin x,\cos x\}$
Bessel orthogonality ($J_\nu(\alpha)=J_\nu(\beta)=0$, $\alpha\ne\beta$)§ 6.3 →
$\int_0^1xJ_\nu(\alpha x)J_\nu(\beta x)\,dx=0$; $\ \int_0^1xJ_\nu(\alpha x)^2dx=\tfrac12J_{\nu+1}(\alpha)^2$
Sturm theory for $y''+q(x)y=0$§ 6.3 →
$m^2\le q\le M^2$ $\Rightarrow$ zero gaps in $\left[\frac\pi M,\frac\pi m\right]$; $\ q\le0$ $\Rightarrow$ at most one zero; SL: $\lambda_n\uparrow\infty$ real, simple, $\varphi_n$ has $n-1$ zeros
Matrix exponential and $x'=Ax$§ 6.4 →
$x(t)=e^{tA}x_0$; $\ \frac{d}{dt}e^{tA}=Ae^{tA}$; $\ e^{A+B}=e^Ae^B$ if $AB=BA$; $\ \det e^{tA}=e^{t\operatorname{tr}A}$; $\ x'=Ax+b$: $x=e^{tA}x_0+\int_0^te^{(t-s)A}b(s)\,ds$
$e^{tA}$ for $2\times2$ $A$§ 6.4 →
double $\lambda$: $e^{\lambda t}\big(I+t(A-\lambda I)\big)$; $\ \alpha\pm i\beta$: $e^{\alpha t}\big(\cos\beta t\,I+\frac{\sin\beta t}{\beta}(A-\alpha I)\big)$; $\ \lambda_1\ne\lambda_2$: $\frac{e^{\lambda_1t}(A-\lambda_2I)-e^{\lambda_2t}(A-\lambda_1I)}{\lambda_1-\lambda_2}$
Planar classification ($\tau=\operatorname{tr}A$, $\Delta=\det A$, $D=\tau^2-4\Delta$)§ 6.4 →
$\Delta\lt0$ saddle; $\Delta\gt0$: $D\gt0$ node, $D=0$ degenerate/star node, $D\lt0$ spiral ($\tau\ne0$) or centre ($\tau=0$); asymptotically stable $\iff\tau\lt0\lt\Delta$
Stability of $x'=Ax$§ 6.4 →
asympt. stable $\iff$ all $\operatorname{Re}\lambda\lt0$; stable $\iff$ all $\operatorname{Re}\lambda\le0$ and those with $\operatorname{Re}\lambda=0$ semisimple
Lyapunov's direct method ($V$ positive definite, $\dot V=\nabla V\cdot f$)§ 6.4 →
$\dot V\le0$ ⇒ stable; $\ \dot V\lt0$ (definite) ⇒ asympt. stable; $\ \dot V\gt0$ definite and $V\gt0$ somewhere near $0$ ⇒ unstable
Linearisation and periodic orbits§ 6.4 →
all $\operatorname{Re}\lambda(Df(x^*))\lt0$ ⇒ asympt. stable; some $\gt0$ ⇒ unstable; $\ \operatorname{div}(\varphi f)$ of one sign on simply connected $D$ ⇒ no periodic orbit in $D$

Chapter 7 · Algebra

Cyclic groups ($|\langle a\rangle|=n$)§ 7.1 →
$o(a^k)=\dfrac{n}{\gcd(n,k)}$; one subgroup of each order $d\mid n$; $\varphi(d)$ elements of order $d$; $\#\{x\in\mathbb{Z}_n:dx=0\}=\gcd(d,n)$
$U(n)$, $\operatorname{Aut}(\mathbb{Z}_n)$ and homomorphisms§ 7.1 →
$U(n)$ cyclic $\iff n=1,2,4,p^k,2p^k$; $\ \operatorname{Aut}(\mathbb{Z}_n)\cong U(n)$; $\ \#\operatorname{Hom}(\mathbb{Z}_m,\mathbb{Z}_n)=\gcd(m,n)$; $\ \mathbb{Z}_m\times\mathbb{Z}_n$ cyclic $\iff\gcd(m,n)=1$
Permutations of cycle type $1^{a_1}2^{a_2}\cdots$ in $S_n$§ 7.1 →
number $=\dfrac{n!}{\prod_k k^{a_k}a_k!}$ (= class size); order = lcm of cycle lengths; even iff an even number of even-length cycles
Orbits, classes and $p$-groups§ 7.1 →
$|Gx|=[G:G_x]$; $\ |G|=|Z(G)|+\sum[G:C(x_i)]$; $\ \#\text{orbits}=\tfrac1{|G|}\sum_g|\operatorname{Fix}(g)|$; $\ |G|=p^k\ (k\ge1)\Rightarrow Z(G)\ne\{e\}$; order $p^2$ abelian
Sylow and abelian groups ($|G|=p^am$, $p\nmid m$)§ 7.1 →
$n_p\mid m$, $\ n_p\equiv1\pmod p$, $\ n_p=1\iff P\trianglelefteq G$; abelian groups of order $\prod p_i^{a_i}$: $\prod p(a_i)$
Counting in $\mathbb{Z}_n$, $n=p_1^{a_1}\cdots p_r^{a_r}$§ 7.2 →
units $\varphi(n)$; nilpotents $n/(p_1\cdots p_r)$; idempotents $2^r$; zero divisors $n-\varphi(n)-1$; ideals $\tau(n)$; prime = maximal ideals $r$
The quotient test§ 7.2 →
$P$ prime $\iff R/P$ integral domain; $\ M$ maximal $\iff R/M$ field; maximal $\Rightarrow$ prime; in a PID nonzero prime $\Rightarrow$ maximal
The hierarchy of domains§ 7.2 →
field $\subsetneq$ ED $\subsetneq$ PID $\subsetneq$ UFD $\subsetneq$ domain; $\ \mathbb{Z}[\tfrac{1+\sqrt{-19}}2]$ PID not ED; $\ \mathbb{Z}[x]$ UFD not PID; $\ \mathbb{Z}[\sqrt{-5}]$ not UFD
Eisenstein's criterion§ 7.2 →
$p\nmid a_n$, $\ p\mid a_0,\dots,a_{n-1}$, $\ p^2\nmid a_0\ \Rightarrow f$ irreducible over $\mathbb{Q}$; apply to $f(x+c)$ if needed
$F[x]/\langle f\rangle$ and counts over $\mathbb{Z}_p$§ 7.2 →
field $\iff f$ irreducible; $\ |F[x]/\langle f\rangle|=|F|^{\deg f}$; monic irreducibles over $\mathbb{Z}_p$: degree 2: $\tfrac{p^2-p}2$, degree 3: $\tfrac{p^3-p}3$
Degrees of extensions§ 7.3 →
$[L:F]=[L:K][K:F]$ ($F\subseteq K\subseteq L$); $\ [F(\alpha):F]=\deg m_{\alpha,F}$; $\ \deg_F\alpha\mid[K:F]$ for $\alpha\in K$; $\ [F(\alpha,\beta):F]\le mn$, $=mn$ if $\gcd(m,n)=1$
Quotient fields ($\deg f=n$)§ 7.3 →
$F[x]/\langle f\rangle$ is a field $\iff f$ irreducible; it has basis $1,\theta,\dots,\theta^{n-1}$ and $q^n$ elements if $F=\mathbb{F}_q$; for $n\le3$: irreducible $\iff$ no root in $F$
Splitting fields over $\mathbb{Q}$§ 7.3 →
$[K:F]\le n!$; $\ x^n-1$: $\varphi(n)$; $\ x^p-a$ ($p$ prime, $a$ not a $p$-th power): $p(p-1)$; $\ x^4-2$: $8$; irreducible cubic: $3$ if $D$ is a square, else $6$
Finite fields $\mathbb{F}_q$, $q=p^n$§ 7.3 →
$\mathbb{F}_q^\times\cong\mathbb{Z}_{q-1}$: $\varphi(q-1)$ generators; $\ \#\{x^k=1\}=\gcd(k,q-1)$; $\ a^q=a$; $\ (a+b)^p=a^p+b^p$; $\ \mathbb{F}_{p^m}\subseteq\mathbb{F}_{p^n}\iff m\mid n$
Irreducible polynomials over $\mathbb{F}_q$§ 7.3 →
$x^{q^n}-x=\prod_{d\mid n}\prod_{\deg f=d}f$; $\ \sum_{d\mid n}dN_q(d)=q^n$; $\ N_q(n)=\tfrac1n\sum_{d\mid n}\mu(d)q^{n/d}$; $\ N_q(\ell)=\tfrac{q^\ell-q}{\ell}$ ($\ell$ prime)
Algebraically closed fields§ 7.3 →
$\mathbb{C}$, $\overline{\mathbb{Q}}$ (countable, $[\overline{\mathbb{Q}}:\mathbb{Q}]=\infty$), $\overline{\mathbb{F}}_p$; always infinite; finite $\Rightarrow$ algebraic, not conversely

Chapter 8 · Functional Analysis

Hölder and Minkowski ($\tfrac1p+\tfrac1q=1$)§ 8.1 →
$\sum|x_ny_n|\le\|x\|_p\|y\|_q,\qquad\|x+y\|_p\le\|x\|_p+\|y\|_p$ (also for $\int_a^b$); $\ \ell^p\subseteq\ell^r$ and $L^r[0,1]\subseteq L^p[0,1]$ for $p\lt r$
Operator norm and matrix norms§ 8.1 →
$\|T\|=\sup_{\|x\|\le1}\|Tx\|$, $\ \|ST\|\le\|S\|\|T\|$; $\ \|A\|_1=\max_j\sum_i|a_{ij}|$, $\ \|A\|_\infty=\max_i\sum_j|a_{ij}|$, $\ \|A\|_2=\sqrt{\lambda_{\max}(A^*A)}$
Integral functionals and operators on $(C[a,b],\|\cdot\|_\infty)$, $g$, $K$ continuous§ 8.1 →
$\big\|f\mapsto\int_a^bgf\big\|=\int_a^b|g|$ (attained if $g$ has constant sign; not attained for $g(t)=2t-1$ on $[0,1]$); $\ \|T_K\|=\max_x\int_a^b|K(x,t)|\,dt$; $\ \|\sum c_i\delta_{t_i}\|=\sum|c_i|$
Duals of the sequence spaces§ 8.1 →
$(\ell^p)^*=\ell^q$ ($1\le p\lt\infty$), $\ c_0^*=\ell^1$, $\ \|x\mapsto\sum x_ny_n\|=\|y\|_q$; but $(\ell^\infty)^*\supsetneq\ell^1$
Hahn–Banach and consequences§ 8.1 →
$g\in M^*\Rightarrow\exists F\in X^*$, $F|_M=g$, $\|F\|=\|g\|$; $\ \|x\|=\max_{\|f\|\le1}|f(x)|$; $\ \operatorname{dist}(x_0,\ker f)=\dfrac{|f(x_0)|}{\|f\|}$
Baire category theorem ($X$ complete metric)§ 8.2 →
$U_n$ open dense $\Rightarrow\bigcap_nU_n$ dense; $\ X=\bigcup_nF_n$, $F_n$ closed $\Rightarrow$ some $F_n$ contains a ball
Uniform boundedness and Banach–Steinhaus ($X$ Banach)§ 8.2 →
$\sup_\alpha\|T_\alpha x\|\lt\infty\ \forall x\Rightarrow\sup_\alpha\|T_\alpha\|\lt\infty$; $\ T_nx\to Tx\ \forall x\Rightarrow T\in B(X,Y)$, $\|T\|\le\liminf\|T_n\|$
Open mapping and bounded inverse ($X$, $Y$ Banach, $T\in B(X,Y)$)§ 8.2 →
$T$ onto $\Rightarrow T(B_X)\supseteq\delta B_Y$; $\ T$ bijective $\Rightarrow\|Tx\|\ge c\|x\|$; $\ \|\cdot\|_a\le C\|\cdot\|_b$, both complete $\Rightarrow$ equivalent
Closed graph theorem ($T:X\to Y$ linear, everywhere defined, $X$, $Y$ Banach)§ 8.2 →
$\big(x_n\to x,\ Tx_n\to y\Rightarrow y=Tx\big)\ \Longrightarrow\ T$ bounded
Dirichlet kernel and Lebesgue constants§ 8.2 →
$D_n(t)=\dfrac{\sin((n+\frac12)t)}{\sin(t/2)}$, $\ L_n=\dfrac1{2\pi}\displaystyle\int_{-\pi}^{\pi}|D_n|\ge\dfrac4{\pi^2}\sum_{k=1}^n\dfrac1k$; $\ L_1=\tfrac13+\tfrac{2\sqrt3}{\pi}$
Cauchy–Schwarz, parallelogram law, polarization§ 8.3 →
$|\langle x,y\rangle|\le\|x\|\|y\|$; $\ \|x+y\|^2+\|x-y\|^2=2\|x\|^2+2\|y\|^2$; $\ \langle x,y\rangle=\tfrac14\big(\|x+y\|^2-\|x-y\|^2\big)$ (real)
Projection theorem ($M$ closed subspace of a Hilbert space)§ 8.3 →
$H=M\oplus M^\perp$, $\ x-P_Mx\perp M$, $\ \|P_M\|=1$, $\ M^{\perp\perp}=M$; $\ \operatorname{dist}(x,\{y\}^\perp)=\dfrac{|\langle x,y\rangle|}{\|y\|}$
Distance to $M=\operatorname{span}\{x_1,\dots,x_n\}$§ 8.3 →
$\operatorname{dist}(x,M)^2=\|x\|^2-\sum|\langle x,e_k\rangle|^2$ ($e_k$ orthonormal basis of $M$) $=\dfrac{G(x_1,\dots,x_n,x)}{G(x_1,\dots,x_n)}$
Bessel and Parseval§ 8.3 →
$\sum|\langle x,e_\alpha\rangle|^2\le\|x\|^2$, equality $\forall x$ iff ONB; $\ \frac1{2\pi}\int_{-\pi}^{\pi}|f|^2=\sum_k|\hat f(k)|^2$; $\ \sum\frac1{n^2}=\frac{\pi^2}6$, $\ \sum\frac1{n^4}=\frac{\pi^4}{90}$
Riesz representation and adjoint§ 8.3 →
$f(x)=\langle x,y_f\rangle$, $\|f\|=\|y_f\|$; $\ \langle Tx,y\rangle=\langle x,T^*y\rangle$, $\ \|T^*T\|=\|T\|^2$, $\ \ker T^*=(\operatorname{ran}T)^\perp$; $\ T_K^*$ has kernel $\overline{K(t,x)}$
Compact operators on Hilbert spaces§ 8.4 →
compact $\iff$ norm limit of finite rank; $\ D_\lambda$ compact $\iff\lambda_n\to0$; $\ \|T_K\|\le\|K\|_{L^2}$, $T_K$ compact for $K\in L^2$; $\ T$ compact $\Rightarrow\|Te_n\|\to0$
Spectrum of $T\in B(X)$§ 8.4 →
$\|A\|\lt1\Rightarrow(I-A)^{-1}=\sum A^n$; $\ \sigma(T)\subseteq\{|\lambda|\le\|T\|\}$ compact, nonempty; $\ r(T)=\lim\|T^n\|^{1/n}$
Self-adjoint $T$ ($m,M$ = inf, sup of $\langle Tx,x\rangle$ on $\|x\|=1$)§ 8.4 →
$\sigma(T)\subseteq[m,M]$, $\ m,M\in\sigma(T)$; $\ \|T\|=\sup_{\|x\|=1}|\langle Tx,x\rangle|=r(T)$
Spectral theorem (compact self-adjoint $T$)§ 8.4 →
$Tx=\sum_n\lambda_n\langle x,e_n\rangle e_n$, $\ \lambda_n\in\mathbb{R}\setminus\{0\}$, $\lambda_n\to0$; $\ \|T\|=|\lambda_1|$; $\ \|T-T_N\|=|\lambda_{N+1}|$; $\ H=\ker T\oplus\overline{\operatorname{span}}\{e_n\}$
Fredholm alternative ($T$ compact self-adjoint, $\mu\ne0$)§ 8.4 →
$\mu\notin\sigma_p$: $u=-\dfrac f\mu+\sum\dfrac{\lambda_n\langle f,e_n\rangle}{\mu(\lambda_n-\mu)}e_n$; $\ \mu\in\sigma_p$: solvable iff $f\perp\ker(T-\mu)$; $\ u=f+\lambda Tu$ fails iff $1/\lambda\in\sigma_p(T)$
Trace and Hilbert–Schmidt identities§ 8.4 →
$\sum\lambda_n^2=\iint|K|^2$ ($K\in L^2$ Hermitian); $\ \sum\lambda_n=\int K(x,x)\,dx$ ($K$ continuous, $T\ge0$); $\ \min(x,t)$: $\lambda_n=\dfrac4{(2n-1)^2\pi^2}$; $\ \min(x,t)-xt$: $\lambda_n=\dfrac1{n^2\pi^2}$

Chapter 9 · Numerical Analysis

Elimination, LU and Cholesky ($\Delta_k$ = leading principal minors)§ 9.1 →
pivot$_k=\Delta_k/\Delta_{k-1}$; $\ \det A=\prod u_{kk}$; $\ A=LL^{\mathsf T}$ (positive diagonal) iff $A$ symmetric positive definite; cost $\approx\tfrac23n^3$ flops
Jacobi and Gauss–Seidel ($A=L+D+U$)§ 9.1 →
$T_J=-D^{-1}(L+U)$, $\ T_{GS}=-(D+L)^{-1}U$; converge for every start iff $\rho(T)\lt1$; strictly diagonally dominant $\Rightarrow$ both converge; tridiagonal: $\rho(T_{GS})=\rho(T_J)^2$
Bisection on $[a,b]$§ 9.1 →
bracket length $\dfrac{b-a}{2^n}$; $\ |c_n-\alpha|\le\dfrac{b-a}{2^{n+1}}$
Fixed-point iteration $x_{n+1}=g(x_n)$§ 9.1 →
$|g'(\alpha)|\lt1$: converges, rate $|g'(\alpha)|$; first nonzero $g^{(p)}(\alpha)$ gives order $p$
Newton and secant methods (simple root, $f\in C^2$)§ 9.1 →
$x_{n+1}=x_n-\dfrac{f(x_n)}{f'(x_n)}$, $\ e_{n+1}\approx\dfrac{f''(\alpha)}{2f'(\alpha)}e_n^2$; secant order $\tfrac{1+\sqrt5}2$; root of multiplicity $m$: linear, rate $\tfrac{m-1}{m}$
Interpolation error ($f\in C^{n+1}$, distinct nodes)§ 9.2 →
$f(x)-p_n(x)=\dfrac{f^{(n+1)}(\xi)}{(n+1)!}(x-x_0)\cdots(x-x_n)$; linear: $|f-p_1|\le\tfrac18M_2h^2$
Newton's forms§ 9.2 →
$p_n=\sum_kf[x_0,\dots,x_k]\prod_{j\lt k}(x-x_j)$; equal steps: $f[x_0,\dots,x_k]=\dfrac{\Delta^kf_0}{k!h^k}$, $\ p_n(x_0+sh)=\sum_k\binom sk\Delta^kf_0$
Difference quotients ($\xi$ near $x_0$)§ 9.2 →
$\dfrac{f_1-f_{-1}}{2h}=f'+\dfrac{h^2}6f'''(\xi)$; $\ \dfrac{f_1-2f_0+f_{-1}}{h^2}=f''+\dfrac{h^2}{12}f^{(4)}(\xi)$
Basic Newton–Cotes errors ($I-Q$)§ 9.2 →
trapezoidal $-\tfrac{h^3}{12}f''$; midpoint $+\tfrac{h^3}{24}f''$; Simpson 1/3 $-\tfrac{h^5}{90}f^{(4)}$; Simpson 3/8 $-\tfrac{3h^5}{80}f^{(4)}$
Composite rules, $h=(b-a)/n$§ 9.2 →
$I-T_n=-\tfrac{(b-a)h^2}{12}f''(\xi)$; $\ I-S_n=-\tfrac{(b-a)h^4}{180}f^{(4)}(\xi)$ ($n$ even); $\ S_{2n}=\tfrac{4T_{2n}-T_n}3$
Euler and the second-order two-stage methods§ 9.3 →
Euler $y_{n+1}=y_n+hf_n$; $\ $Heun $y_{n+1}=y_n+\tfrac h2\left[f_n+f(x_{n+1},y_n+hf_n)\right]$; $\ $midpoint $y_{n+1}=y_n+hf\left(x_n+\tfrac h2,y_n+\tfrac h2f_n\right)$
Two-stage RK: $k_2=f(x_n+\alpha h,y_n+\beta hk_1)$, $y_{n+1}=y_n+h(ak_1+bk_2)$§ 9.3 →
order $\ge1\iff a+b=1$; $\ $order $2\iff a+b=1,\ b\alpha=b\beta=\tfrac12$; $\ $order $3$ never
Local and global error ($|T_{n+1}|\le Kh^{p+1}$, $\Phi$ Lipschitz with $\Lambda$, $e_0=0$)§ 9.3 →
$|e_n|\le\dfrac{Kh^p}{\Lambda}\left(e^{\Lambda(x_n-x_0)}-1\right)$; $\ $Euler: $T_{n+1}=\tfrac{h^2}2y''(\xi_n)$, $K=\tfrac{M_2}2$, $\Lambda=L$
Stability for $y'=\lambda y$, $z=h\lambda$ (stable iff $|R(z)|\lt1$)§ 9.3 →
Euler $1+z$; $\ $RK2 $1+z+\tfrac{z^2}2$ (both: interval $(-2,0)$, $h\lt2/|\lambda|$); $\ $backward Euler $\tfrac1{1-z}$; $\ $trapezoidal $\tfrac{1+z/2}{1-z/2}$ (A-stable)
Step halving (order $p$)§ 9.3 →
$y-y_{h/2}\approx\dfrac{y_{h/2}-y_h}{2^p-1}$, $\ \ y\approx\dfrac{2^py_{h/2}-y_h}{2^p-1}$

Chapter 10 · Partial Differential Equations

Constant coefficients ($a\ne0$); transport equation§ 10.1 →
$a\,z_x+b\,z_y=c\,z\ \Rightarrow\ z=e^{cx/a}f(bx-ay)$; $\quad z_t+c\,z_x=0\ \Rightarrow\ z=f(x-ct)$
Lagrange's method for $Pp+Qq=R$ ($u,v$ independent first integrals)§ 10.1 →
$\dfrac{dx}{P}=\dfrac{dy}{Q}=\dfrac{dz}{R}$; general integral $\varphi(u,v)=0$, $\varphi$ arbitrary
Method of multipliers§ 10.1 →
If $lP+mQ+nR=0$ and $l\,dx+m\,dy+n\,dz=du$, then $u$ is a first integral
Transversality for data $z(x_0(s),y_0(s))=z_0(s)$ ($P,Q,R\in C^1$)§ 10.1 →
$J=x_0'\,Q-y_0'\,P\ne0\Rightarrow$ unique local solution; $J\equiv0$: none, or infinitely many if the data curve is a characteristic
Burgers' equation $z_y+z\,z_x=0$, $z(x,0)=g(x)$§ 10.1 →
$z=g(x-zy)$, constant on $x=s+g(s)y$; breaking time $y_b=-\dfrac{1}{\inf g'}$ (none if $g'\ge0$)
Type of $Au_{xx}+Bu_{xy}+Cu_{yy}+\dots=0$ ($B$ = whole coefficient of $u_{xy}$)§ 10.2 →
$\delta=B^2-4AC$: $\gt0$ hyperbolic, $=0$ parabolic, $\lt0$ elliptic; characteristics $A\left(\frac{dy}{dx}\right)^2-B\frac{dy}{dx}+C=0$
Constant coefficients, no lower terms ($Am^2-Bm+C=0$)§ 10.2 →
$m_1\ne m_2$ real: $u=f(y-m_1x)+g(y-m_2x)$; $\ $double root $m$: $u=f(y-mx)+x\,g(y-mx)$
Laplace in $(0,a)\times(0,b)$, $u=f(x)$ on $y=b$, $0$ on the other sides§ 10.2 →
$u=\sum b_n\dfrac{\sinh(n\pi y/a)}{\sinh(n\pi b/a)}\sin\dfrac{n\pi x}a$, $\ b_n=\frac2a\int_0^af\sin\frac{n\pi x}a\,dx$
Laplace in the disc $r\lt\rho$, $u(\rho,\theta)=f(\theta)$§ 10.2 →
$u=\frac{a_0}2+\sum\left(\frac r\rho\right)^n(a_n\cos n\theta+b_n\sin n\theta)$; $u(0)=\frac1{2\pi}\int f$; Neumann needs $\int g=0$
Heat $u_t=ku_{xx}$ on $(0,L)$, ends at $0$ / insulated§ 10.2 →
$\sum b_ne^{-k(n\pi/L)^2t}\sin\frac{n\pi x}L$ $\ /\ $ $\frac{a_0}2+\sum a_ne^{-k(n\pi/L)^2t}\cos\frac{n\pi x}L$
String $u_{tt}=c^2u_{xx}$, fixed ends, $u=f$, $u_t=g$§ 10.2 →
$u=\sum\left(a_n\cos\frac{n\pi ct}L+\frac{L\beta_n}{n\pi c}\sin\frac{n\pi ct}L\right)\sin\frac{n\pi x}L$; period $\frac{2L}c$
d'Alembert's formula ($f\in C^2$, $g\in C^1$)§ 10.3 →
$u=\tfrac12[f(x+ct)+f(x-ct)]+\tfrac1{2c}\int_{x-ct}^{x+ct}g$
Duhamel: $u_{tt}-c^2u_{xx}=F$, zero data ($F,F_x$ continuous)§ 10.3 →
$u=\tfrac1{2c}\int_0^t\!\int_{x-c(t-\tau)}^{x+c(t-\tau)}F(y,\tau)\,dy\,d\tau$
Heat equation on $\mathbb{R}$ ($f$ bounded, continuous)§ 10.3 →
$u=\int\Phi(x-y,t)f(y)\,dy,\ \ \Phi=\dfrac{e^{-x^2/(4kt)}}{\sqrt{4\pi kt}}$
Fourier transforms ($a\gt0$)§ 10.3 →
$e^{-a|x|}\mapsto\dfrac{2a}{a^2+\omega^2}$, $\ e^{-ax^2}\mapsto\sqrt{\tfrac\pi a}\,e^{-\omega^2/(4a)}$
Fourier rules§ 10.3 →
$\widehat{f'}=i\omega\hat f$, $\ \widehat{f*g}=\hat f\hat g$, $\ \widehat{f(x-b)}=e^{-ib\omega}\hat f$
Laplace pairs for half-line heat problems ($a\gt0$)§ 10.3 →
$\mathcal{L}\{\operatorname{erfc}\tfrac{a}{2\sqrt t}\}=\dfrac{e^{-a\sqrt s}}{s}$; rod: $u=T_0\operatorname{erfc}\dfrac{x}{2\sqrt{kt}}$

Chapter 11 · Topology

Basis $\mathcal{B}$ and the topology it generates§ 11.1 →
(B1) $\bigcup\mathcal{B}=X$; (B2) $x\in B_1\cap B_2\Rightarrow\exists B_3:\ x\in B_3\subseteq B_1\cap B_2$; $\ U$ open $\iff U$ is a union of basic sets
Closure and boundary§ 11.1 →
$x\in\overline{A}\iff$ every basic neighbourhood of $x$ meets $A$; $\ \overline{A}=A\cup A'$; $\ \partial A=\overline{A}\cap\overline{X\setminus A}$
Continuity§ 11.1 →
$f^{-1}(\text{open})$ open $\iff f(\overline{A})\subseteq\overline{f(A)}$ for all $A$; $\ f:\mathbb{R}_\ell\to\mathbb{R}$ continuous $\iff f$ right-continuous
Products§ 11.1 →
basis $U\times V$; $\ \overline{A\times B}=\overline{A}\times\overline{B}$; $\ f$ into $\prod X_\alpha$ (product topology) continuous $\iff$ each $\pi_\alpha\circ f$ is
Quotients§ 11.1 →
$V$ open in $X/{\sim}\iff q^{-1}(V)$ open; $\ g$ continuous $\iff g\circ q$ continuous; compact $\to$ Hausdorff continuous surjection induces $X/{\sim}\cong Y$
Separation ladder ($T_3,T_4$ include $T_1$)§ 11.2 →
metric $\Rightarrow T_4\Rightarrow T_3\Rightarrow T_2\Rightarrow T_1\Rightarrow T_0$; $\ $ compact $+\,T_2\Rightarrow T_4$; $\ T_1\iff$ points closed
Countability axioms§ 11.2 →
2nd countable $\Rightarrow$ 1st countable, separable, Lindelöf; $\ $ metric: 2nd countable $\iff$ separable $\iff$ Lindelöf
Urysohn function (metric space, $A\cap B=\varnothing$ closed)§ 11.2 →
$f(x)=\dfrac{d(x,A)}{d(x,A)+d(x,B)}$, $\ f|_A=0$, $f|_B=1$
Compact and Hausdorff§ 11.2 →
compact $\subseteq$ Hausdorff $\Rightarrow$ closed; $\ f$ continuous bijection, compact $\to$ Hausdorff $\Rightarrow$ homeomorphism
Components and clopen sets§ 11.2 →
$k$ components ($k$ finite) $\Rightarrow$ exactly $2^k$ clopen subsets

Chapter 12 · Linear Programming

Basic feasible solutions of $Ax=b$, $x\ge0$ ($A$ is $m\times n$ of rank $m$)§ 12.1 →
$x_{\mathcal B}=B^{-1}b\ge0$, $x_{\mathcal N}=0$; at most $\binom nm$ basic solutions; BFS $=$ extreme point; degenerate iff fewer than $m$ positive components
Simplex tableau (maximisation)§ 12.1 →
$\Delta_j=c_{\mathcal B}^{\mathsf T}B^{-1}a_j-c_j$; all $\Delta_j\ge0\Rightarrow$ optimal; $\Delta_k\lt0$, $y_k\le0\Rightarrow$ unbounded; $\theta=\min_{y_{ik}\gt0}\bar x_{B_i}/y_{ik}$, $\ z_{\text{new}}=z_0-\theta\Delta_k$
Two-phase and Big-M methods§ 12.1 →
Phase I: $\min\sum a_i$; feasible iff $w^*=0$; Big-M net evaluations $\alpha M+\beta$, $\alpha=$ Phase I row
Revised simplex update§ 12.1 →
$\pi^{\mathsf T}=c_{\mathcal B}^{\mathsf T}B^{-1}$; $\ B^{-1}_{\text{new}}=EB^{-1}$, $E$ = identity with column $r$ replaced by $\eta$, $\eta_r=\tfrac1{y_{rk}}$, $\eta_i=-\tfrac{y_{ik}}{y_{rk}}$
Duality: $\max c^{\mathsf T}x$, $Ax\le b$, $x\ge0$ $\leftrightarrow$ $\min b^{\mathsf T}y$, $A^{\mathsf T}y\ge c$, $y\ge0$§ 12.1 →
$c^{\mathsf T}x\le b^{\mathsf T}y$ (weak); equal at optima (strong); optimal iff $y_i(b-Ax)_i=0$ and $x_j(A^{\mathsf T}y-c)_j=0$; $y_i^*=\partial z^*/\partial b_i$ while the basis stays feasible
Transportation problem ($m$ sources, $n$ destinations)§ 12.2 →
balanced iff $\sum a_i=\sum b_j$ (else a dummy row/column, costs $0$ or penalties); a basis $=$ $m+n-1$ cells with no loop; integer data $\Rightarrow$ integral BFS
MODI (u–v) test§ 12.2 →
$u_i+v_j=c_{ij}$ on basic cells ($u_1=0$); $\ d_{ij}=c_{ij}-u_i-v_j$; all $d_{ij}\ge0\Rightarrow$ optimal (unique if all nonbasic $\gt0$); cost $=\sum a_iu_i+\sum b_jv_j$
Loop pivot§ 12.2 →
$\theta=\min$ of the “$-$” corners; cost changes by $\theta\,d_{pq}$; $\theta=0$: degenerate pivot
Hungarian method ($n\times n$, minimise)§ 12.2 →
reduce rows, then columns; fewest covering lines $r$; if $r\lt n$: $k=$ min uncovered, $-k$ uncovered, $+k$ doubly covered, $L\mathrel{+}=k(n-r)$; $r=n$: optimal cost $=L$
Assignment variants§ 12.2 →
max: costs $P-p_{ij}$; unbalanced: dummy rows/columns of $0$; forbidden: cost $M$