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 · Sets, Real Numbers and Countability

De Morgan's laws (any family)§ 1.1 →
$X\setminus\bigcup_i A_i=\bigcap_i(X\setminus A_i)$, $\ X\setminus\bigcap_i A_i=\bigcup_i(X\setminus A_i)$
Images and preimages§ 1.1 →
$f^{-1}$ preserves $\cup$, $\cap$, complements. $f(\bigcup A_i)=\bigcup f(A_i)$; $f(A\cap B)\subseteq f(A)\cap f(B)$ (equality for all $A,B$ iff $f$ injective); $f(f^{-1}(B))=B\cap f(X)$; $A\subseteq f^{-1}(f(A))$ (equality for all $A$ iff $f$ injective).
Bijections $\mathbb{N}\times\mathbb{N}\to\mathbb{N}$§ 1.1 →
$\pi(m,n)=\tfrac{(m+n-1)(m+n-2)}{2}+m$, $\quad\varphi(m,n)=2^{m-1}(2n-1)$
Countable sets§ 1.1 →
Countable: subsets and images of countable sets, finite products, countable unions of countable sets; $\mathbb{Z}$, $\mathbb{Q}$, $\mathbb{Q}^k$, $\mathbb{Z}[x]$, algebraic numbers, finite subsets / finite sequences of $\mathbb{N}$. Uncountable: $\mathbb{R}$, intervals, $\mathbb{R}\setminus\mathbb{Q}$, $\{0,1\}^{\mathbb{N}}$, $\mathcal{P}(\mathbb{N})$, $\mathbb{N}^{\mathbb{N}}$.
Cantor and Schröder–Bernstein§ 1.1 →
No surjection $A\to\mathcal{P}(A)$, so $|A|\lt|\mathcal{P}(A)|$. If injections $A\to B$ and $B\to A$ exist, then $|A|=|B|$.
Cardinal arithmetic§ 1.1 →
$\aleph_0+\aleph_0=\aleph_0\cdot\aleph_0=\aleph_0$; $\ 2^{\aleph_0}=\aleph_0^{\aleph_0}=\mathfrak{c}=\mathfrak{c}^2=\mathfrak{c}^{\aleph_0}$; $\ |C(\mathbb{R})|=\mathfrak{c}\lt2^{\mathfrak{c}}=|\mathbb{R}^{\mathbb{R}}|$
Triangle inequalities§ 1.2 →
$|x+y|\le|x|+|y|$ (equality iff $xy\ge0$); $\ \big||x|-|y|\big|\le|x-y|$
$\varepsilon$-characterisation of sup and inf§ 1.2 →
$u=\sup S\iff$ $s\le u\ \forall s\in S$ and $\forall\varepsilon\gt0\ \exists s\in S:\ s\gt u-\varepsilon$. $\quad\inf S=-\sup(-S)$.
Algebra of suprema (nonempty bounded sets)§ 1.2 →
$\sup(A+B)=\sup A+\sup B$; $\sup(A-B)=\sup A-\inf B$; $\sup(cA)=c\sup A$ $(c\gt0)$, $=c\inf A$ $(c\lt0)$; $\sup(AB)=\sup A\sup B$ only for $A,B\subseteq[0,\infty)$.
Archimedean property and density§ 1.2 →
$x\gt0\Rightarrow\exists n:\ nx\gt y$. For $a\lt b$: choose $\tfrac1n\lt b-a$, $m=\lfloor na\rfloor+1$; then $a\lt\tfrac mn\lt b$, and $r\sqrt2$ gives an irrational.
Nested interval property§ 1.2 →
$[a_n,b_n]\supseteq[a_{n+1},b_{n+1}]$ closed and bounded $\Rightarrow\bigcap_n[a_n,b_n]=[\sup a_n,\inf b_n]\ne\varnothing$; a single point if $b_n-a_n\to0$.
Expansions in base $b$§ 1.2 →
$x=\sum_{k\ge1}d_kb^{-k}$, $d_k=\lfloor b^kx\rfloor-b\lfloor b^{k-1}x\rfloor$. Two expansions iff $x=m/b^n$; rational iff eventually periodic.

Chapter 2 · Sequences and Series of Real Numbers

Limit of a sequence§ 2.1 →
$a_n\to L\iff\forall\varepsilon\gt0\ \exists N\ \forall n\ge N:\ |a_n-L|\lt\varepsilon$. Convergent $\Rightarrow$ bounded; limits are unique; $a_n\le b_n\Rightarrow\lim a_n\le\lim b_n$.
Standard limits§ 2.1 →
$a^n\to0\ (|a|\lt1)$; $\ a^{1/n}\to1\ (a\gt0)$; $\ n^{1/n}\to1$; $\ \dfrac{n^k}{a^n}\to0\ (a\gt1)$; $\ \dfrac{a^n}{n!}\to0$; $\ \dfrac{n!}{n^n}\to0$.
Ratio criterion for sequences§ 2.1 →
$\left|\dfrac{a_{n+1}}{a_n}\right|\to\rho\lt1\Rightarrow a_n\to0$; $\ \rho\gt1\Rightarrow|a_n|\to\infty$; $\rho=1$: no conclusion.
The number $e$§ 2.1 →
$\left(1+\tfrac1n\right)^n\uparrow e$, $\ \left(1+\tfrac1n\right)^{n+1}\downarrow e$, $\ e=\sum_{k\ge0}\tfrac1{k!}$, $\ 0\lt e-\sum_{k=0}^n\tfrac1{k!}\lt\tfrac1{n!\,n}$.
Contractive sequences: error bounds§ 2.1 →
$|a_{n+2}-a_{n+1}|\le c|a_{n+1}-a_n|$, $0\lt c\lt1$ $\Rightarrow$ $|a-a_n|\le\dfrac{c^{\,n-1}}{1-c}|a_2-a_1|$ and $|a-a_n|\le\dfrac{c}{1-c}|a_n-a_{n-1}|$.
lim sup and lim inf§ 2.2 →
$\limsup x_n=\inf_n\sup_{k\ge n}x_k=\max\mathcal{S}$, $\ \liminf x_n=\sup_n\inf_{k\ge n}x_k=\min\mathcal{S}$; $\ x_n\to L\iff\liminf x_n=\limsup x_n=L$.
lim sup of sums and products (bounded sequences)§ 2.2 →
$\liminf x+\liminf y\le\liminf(x+y)\le\liminf x+\limsup y\le\limsup(x+y)\le\limsup x+\limsup y$; $\ \limsup(xy)\le\limsup x\,\limsup y$ if $x_n,y_n\ge0$.
Ratios control roots ($x_n\gt0$)§ 2.2 →
$\liminf\tfrac{x_{n+1}}{x_n}\le\liminf x_n^{1/n}\le\limsup x_n^{1/n}\le\limsup\tfrac{x_{n+1}}{x_n}$; e.g. $\tfrac{(n!)^{1/n}}{n}\to\tfrac1e$, $\binom{2n}n^{1/n}\to4$.
Cesàro means§ 2.2 →
$x_n\to L\Rightarrow\dfrac{x_1+\cdots+x_n}{n}\to L$; $\ x_n\gt0,\ x_n\to L\Rightarrow(x_1\cdots x_n)^{1/n}\to L$. Converse false: $(-1)^n$.
Stolz–Cesàro theorem§ 2.2 →
$b_n\uparrow+\infty$ strictly, $\dfrac{a_{n+1}-a_n}{b_{n+1}-b_n}\to L\ \Rightarrow\ \dfrac{a_n}{b_n}\to L$; e.g. $\dfrac{1^k+\cdots+n^k}{n^{k+1}}\to\dfrac1{k+1}$.
Model series§ 2.3 →
$\sum_{n\ge0}r^n=\dfrac1{1-r}\ (|r|\lt1)$; $\ \sum\dfrac1{n^p}\lt\infty\iff p\gt1$; $\ \sum\dfrac1{n(\log n)^p}\lt\infty\iff p\gt1$; $\ \sum(b_n-b_{n+1})=b_1-\lim b_n$.
Condensation and integral tests ($a_n$, $f$ decreasing, $\ge0$)§ 2.3 →
$\sum a_n\lt\infty\iff\sum2^ka_{2^k}\lt\infty$; $\ \int_{N+1}^\infty f\le\sum_{n\gt N}f(n)\le\int_N^\infty f$; $\ H_n-\log n\downarrow\gamma\approx0.5772$.
Root, ratio and Raabe tests§ 2.3 →
$\limsup|a_n|^{1/n}\lt1\Rightarrow$ abs. conv., $\gt1\Rightarrow$ div.; $\ \limsup\left|\tfrac{a_{n+1}}{a_n}\right|\lt1\Rightarrow$ abs. conv.; $\ n\big(\tfrac{a_n}{a_{n+1}}-1\big)\to R$: $R\gt1$ conv., $R\lt1$ div. ($a_n\gt0$).
Leibniz, Dirichlet and Abel tests§ 2.3 →
$a_n\downarrow0\Rightarrow\sum(-1)^{n+1}a_n=s$ with $|s-s_n|\le a_{n+1}$; $\ |A_n|\le M,\ b_n\text{ monotone}\to0\Rightarrow\sum a_nb_n$ conv.; $\ \sum a_n$ conv., $b_n$ monotone bounded $\Rightarrow\sum a_nb_n$ conv.
Rearrangements and Cauchy products§ 2.3 →
Abs. conv. $\Rightarrow$ every rearrangement has the same sum; cond. conv. $\Rightarrow$ any sum (Riemann). $\ c_n=\sum_{k=0}^na_kb_{n-k}$: $\sum c_n=AB$ if one factor converges absolutely (Mertens).

Chapter 3 · Continuity and Differentiability

Limit of a function§ 3.1 →
$\lim_{x\to c}f(x)=L\iff\forall\varepsilon\gt0\ \exists\delta\gt0:\ 0\lt|x-c|\lt\delta\Rightarrow|f(x)-L|\lt\varepsilon$ $\iff f(x_n)\to L$ for every $x_n\to c$, $x_n\ne c$.
Mixed definitions on $\mathbb{Q}$ and $\mathbb{R}\setminus\mathbb{Q}$§ 3.1 →
$f=g$ on $\mathbb{Q}$, $f=h$ off $\mathbb{Q}$ ($g,h$ continuous): $f$ is continuous exactly on $\{g=h\}$. Thomae $T$: continuous exactly at the irrationals.
Heine–Borel in $\mathbb{R}$§ 3.1 →
$K$ compact $\iff$ $K$ closed and bounded $\iff$ every sequence in $K$ has a subsequence converging in $K$.
EVT and IVT§ 3.1 →
$f$ continuous on $[a,b]$ $\Rightarrow f([a,b])=[m,M]$ with $m=\min f$, $M=\max f$ attained. Continuous $f:[a,b]\to[a,b]$ has a fixed point.
Monotone functions and inverses§ 3.1 →
On an interval: continuous $+$ injective $\Rightarrow$ strictly monotone; monotone $+$ interval image $\Rightarrow$ continuous; continuous strictly monotone $\Rightarrow f^{-1}$ continuous.
Uniform continuity§ 3.2 →
$\forall\varepsilon\ \exists\delta:\ |x-y|\lt\delta\Rightarrow|f(x)-f(y)|\lt\varepsilon$. Not UC $\iff\exists x_n,y_n$: $|x_n-y_n|\to0$, $|f(x_n)-f(y_n)|\ge\varepsilon_0$.
Lipschitz and Hölder bounds§ 3.2 →
$|f'|\le K$ on an interval $\Rightarrow|f(x)-f(y)|\le K|x-y|$. $\ |x^\alpha-y^\alpha|\le|x-y|^\alpha$ for $x,y\ge0$, $0\lt\alpha\le1$.
When continuity is uniform§ 3.2 →
Continuous on compact $\Rightarrow$ UC. Continuous on bounded $(a,b)$: UC $\iff f(a+),f(b-)$ exist. Continuous on $[a,\infty)$ with finite $\lim_{\infty}f$ $\Rightarrow$ UC.
Jumps of monotone functions§ 3.2 →
$f\uparrow$ on $[a,b]$: $f(c-)\le f(c)\le f(c+)$; $\sum(f(c_i+)-f(c_i-))\le f(b)-f(a)$; discontinuities countable. Froda: first-kind discontinuities of any $f$ are countable.
Mean value theorems§ 3.3 →
$f,g$ continuous on $[a,b]$, differentiable on $(a,b)$: $\ f(b)-f(a)=f'(c)(b-a)$; $\ (f(b)-f(a))g'(c)=(g(b)-g(a))f'(c)$ for some $c\in(a,b)$. Rolle: $f(a)=f(b)\Rightarrow f'(c)=0$.
L'Hôpital's rule§ 3.3 →
$f,g$ differentiable near $a$, $g'\ne0$, $f,g\to0$ or $|g|\to\infty$, and $\dfrac{f'}{g'}\to L\in[-\infty,\infty]$ $\Rightarrow$ $\dfrac fg\to L$. Only sufficient.
Taylor's theorem with remainder§ 3.3 →
$f^{(n)}$ continuous, $f^{(n+1)}$ exists inside: $R_n(x)=\dfrac{f^{(n+1)}(\xi)}{(n+1)!}(x-c)^{n+1}$ (Lagrange) $=\dfrac{f^{(n+1)}(\eta)}{n!}(x-\eta)^n(x-c)$ (Cauchy); $R_n=o((x-c)^n)$ (Peano).
Derivative facts§ 3.3 →
Darboux: $f'(I)$ is an interval; $f'$ has no jumps. $f'(x)\to L$ as $x\to c$ and $f$ continuous at $c$ $\Rightarrow f'(c)=L$. First nonzero $f^{(n)}(c)$: $n$ even $\Rightarrow$ extremum, $n$ odd $\Rightarrow$ none.
Convexity criteria§ 3.3 →
$f$ differentiable on open $I$: convex $\iff f'\uparrow\iff f(y)\ge f(x)+f'(x)(y-x)$; twice differentiable: $\iff f''\ge0$. Jensen: $f(\sum\lambda_ix_i)\le\sum\lambda_if(x_i)$.

Chapter 4 · Integration: Riemann to Lebesgue

Darboux sums and Riemann's criterion§ 4.1 →
$U(f,P)=\sum M_k\Delta x_k$, $L(f,P)=\sum m_k\Delta x_k$; $f$ bounded is integrable $\iff\forall\varepsilon\gt0\ \exists P:\ U(f,P)-L(f,P)\lt\varepsilon$.
Lebesgue's criterion§ 4.1 →
$f\in\mathcal{R}[a,b]\iff f$ bounded and $\{x: f\text{ discontinuous at }x\}$ has measure zero.
Limits of sums§ 4.1 →
$f\in\mathcal{R}[0,1]\Rightarrow\displaystyle\lim_{n\to\infty}\frac1n\sum_{k=1}^nf\!\left(\tfrac kn\right)=\int_0^1f(x)\,dx$
FTC and Leibniz's rule ($f$ continuous)§ 4.1 →
$\dfrac{d}{dx}\displaystyle\int_{u(x)}^{v(x)}f(t)\,dt=f(v(x))\,v'(x)-f(u(x))\,u'(x)$; $\ \int_a^bF'=F(b)-F(a)$ if $F'\in\mathcal{R}[a,b]$.
Mean value theorem for integrals§ 4.1 →
$f$ continuous, $g\ge0$ integrable $\Rightarrow\exists c\in[a,b]:\ \int_a^bfg=f(c)\int_a^bg$.
The $p$-tests§ 4.2 →
$\int_1^\infty\frac{dx}{x^p}\lt\infty\iff p\gt1$; $\ \int_0^1\frac{dx}{x^p}\lt\infty\iff p\lt1$; $\ \int_2^\infty\frac{dx}{x(\log x)^q}\lt\infty\iff q\gt1$.
Limit comparison ($f\ge0$, $g\gt0$)§ 4.2 →
$f/g\to L\in(0,\infty)$ at the bad point $\Rightarrow\int f$ and $\int g$ converge or diverge together.
Dirichlet integral and family§ 4.2 →
$\int_0^\infty\frac{\sin x}{x}dx=\frac\pi2$; $\ \int_0^\infty\frac{\sin x}{x^p}dx$ converges iff $0\lt p\lt2$, absolutely iff $1\lt p\lt2$.
Gamma function ($s\gt0$)§ 4.2 →
$\Gamma(s)=\int_0^\infty x^{s-1}e^{-x}dx$; $\Gamma(s+1)=s\Gamma(s)$; $\Gamma(n+1)=n!$; $\Gamma(\tfrac12)=\sqrt\pi$; $\int_0^\infty x^me^{-x^n}dx=\tfrac1n\Gamma\big(\tfrac{m+1}n\big)$.
Beta function ($p,q\gt0$)§ 4.2 →
$B(p,q)=\frac{\Gamma(p)\Gamma(q)}{\Gamma(p+q)}=2\int_0^{\pi/2}\sin^{2p-1}\theta\cos^{2q-1}\theta\,d\theta$; $\ \Gamma(p)\Gamma(1-p)=\frac{\pi}{\sin p\pi}$.
Total variation§ 4.3 →
$V_a^b(f)=\sup_P\sum|f(x_k)-f(x_{k-1})|$; $\ =|f(b)-f(a)|$ if $f$ is monotone; $\ \le K(b-a)$ if $f$ is $K$-Lipschitz; $\ =\int_a^b|f'|$ if $f'$ exists everywhere and is Riemann integrable.
Jordan decomposition ($f\in\mathrm{BV}[a,b]$)§ 4.3 →
$v(x)=V_a^x(f)$, $\ v^\pm=\tfrac12\big(v\pm(f-f(a))\big)$ increasing, $\ f=f(a)+v^+-v^-$; $\ v(c^+)-v(c)=|f(c^+)-f(c)|$.
The family $x^a\sin(x^{-b})$ on $[0,1]$ ($a,b\gt0$)§ 4.3 →
Continuous for all $a,b\gt0$; of bounded variation $\iff a\gt b$, with $V_0^1\le1+\frac{a}{a-b}$. So $x\sin\frac1x\notin\mathrm{BV}$, $x^2\sin\frac1x\in\mathrm{BV}$, $x^2\sin\frac1{x^2}\notin\mathrm{BV}$.
Length of curves and graphs§ 4.3 →
$\gamma\in C^1$: $L=\int_a^b|\gamma'(t)|\,dt$; graph of $f$: rectifiable $\iff f\in\mathrm{BV}$, and $\sqrt{(b-a)^2+V^2}\le L\le(b-a)+V$.
Riemann–Stieltjes integrals ($f$ continuous)§ 4.3 →
$\int f\,d\alpha=\int f\alpha'\,dx$ ($\alpha\in C^1$); step $\alpha$: $\sum_if(c_i)\big(\alpha(c_i^+)-\alpha(c_i^-)\big)$, end points one-sided; $\int f\,d\alpha+\int\alpha\,df=f\alpha\big|_a^b$; $\big|\int f\,d\alpha\big|\le\|f\|_\infty V(\alpha)$.
Lebesgue outer measure§ 4.4 →
$m^*(A)=\inf\big\{\sum_k\ell(I_k):A\subseteq\bigcup_kI_k\big\}$; $\ m^*(I)=\ell(I)$; $\ m^*\big(\bigcup A_k\big)\le\sum m^*(A_k)$; countable sets have $m^*=0$.
Carathéodory's criterion§ 4.4 →
$E\in\mathcal M\iff m^*(A)=m^*(A\cap E)+m^*(A\cap E^c)$ for all $A$; on $\mathcal M$, $m\big(\bigsqcup E_k\big)=\sum m(E_k)$.
Continuity of measure§ 4.4 →
$E_n\uparrow E\Rightarrow m(E_n)\to m(E)$; $\ E_n\downarrow E$ and $m(E_1)\lt\infty\Rightarrow m(E_n)\to m(E)$.
Cantor sets§ 4.4 →
$m(C)=\lim(\tfrac23)^n=0$, $|C|=\mathfrak c$; removing $2^{n-1}$ middle intervals of length $\alpha3^{-n}$ at stage $n$ leaves $m(K_\alpha)=1-\alpha$.
Simple approximation ($f\ge0$ measurable)§ 4.4 →
$\varphi_n=\min\big(n,2^{-n}\lfloor2^nf\rfloor\big)\uparrow f$ pointwise; uniformly where $f$ is bounded.
Lebesgue integral§ 4.5 →
$\int\sum a_i\mathbf 1_{E_i}=\sum a_im(E_i)$; $\ \int f=\sup\{\int\varphi:0\le\varphi\le f\}$ ($f\ge0$); $\ \int f=\int f^+-\int f^-$ if $\int|f|\lt\infty$.
MCT and Fatou ($f_n\ge0$ measurable)§ 4.5 →
$f_n\uparrow f\Rightarrow\int f_n\uparrow\int f$; $\ \int\liminf f_n\le\liminf\int f_n$; $\ \int\sum g_k=\sum\int g_k$ ($g_k\ge0$).
Dominated convergence§ 4.5 →
$f_n\to f$ a.e., $|f_n|\le g\in L^1\Rightarrow\int|f_n-f|\to0$, $\int f_n\to\int f$; bounded case: $m(E)\lt\infty$, $|f_n|\le M$.
Markov's inequality§ 4.5 →
$m(\{|f|\ge t\})\le\frac1t\int|f|$ for $t\gt0$.
Differentiation under the integral§ 4.5 →
$|\partial_tf(x,t)|\le h(x)\in L^1\Rightarrow\frac{d}{dt}\int f(x,t)\,dx=\int\partial_tf(x,t)\,dx$.
Riemann vs Lebesgue§ 4.5 →
Riemann integrable on $[a,b]$ $\Rightarrow$ Lebesgue, same value; improper $\int_a^\infty f$ is a Lebesgue integral iff $\int_a^\infty|f|\lt\infty$.

Chapter 5 · Sequences and Series of Functions

Uniform convergence§ 5.1 →
$f_n\rightrightarrows f$ on $D\iff M_n=\sup_{x\in D}|f_n(x)-f(x)|\to0$; $\ $ not uniform if $f_n(x_n)-f(x_n)\not\to0$ for some $x_n\in D$.
Standard sups§ 5.1 →
$\sup_{[0,1)}x^n=1$; $\ \max_{x\ge0}nxe^{-nx}=e^{-1}$; $\ \max_{\mathbb{R}}\dfrac{|x|}{1+nx^2}=\dfrac1{2\sqrt n}$; $\ \max_{x\ge0}\dfrac{nx}{1+n^2x^2}=\dfrac12$; $\ \max_{[0,1]}x^n(1-x)\lt\dfrac1{n+1}$.
Cauchy criterion and M-test§ 5.1 →
$\sum u_n$ unif. conv. $\iff\sup_x\big|\sum_{k=m+1}^nu_k(x)\big|\to0$ $\Rightarrow\|u_n\|_D\to0$; $\ |u_n|\le M_n,\ \sum M_n\lt\infty\Rightarrow$ abs. and unif., $\|R_N\|\le\sum_{n\gt N}M_n$.
Dirichlet and Abel tests (uniform)§ 5.1 →
$b_n(x)$ monotone in $n$: $\ |\sum_{k\le n}a_k|\le M,\ b_n\rightrightarrows0\Rightarrow\sum a_nb_n$ unif.; $\ \sum a_n$ unif., $|b_n|\le K\Rightarrow\sum a_nb_n$ unif.
Dini's theorem§ 5.1 →
$K$ compact, $f_n,f\in C(K)$, $f_n(x)$ monotone $\to f(x)$ $\Rightarrow f_n\rightrightarrows f$ on $K$.
Interchange theorems (uniform convergence)§ 5.2 →
$f_n\rightrightarrows f$, $f_n$ continuous $\Rightarrow f$ continuous; $\ f_n\rightrightarrows f$ on $[a,b]$, $f_n$ integrable $\Rightarrow\int_a^bf_n\to\int_a^bf$ (bounded interval only).
Differentiation term by term§ 5.2 →
$f_n$ differentiable on $[a,b]$, $f_n(x_0)$ converges, $f_n'\rightrightarrows g$ $\Rightarrow f_n\rightrightarrows f$ and $f'=g$.
Radius of convergence§ 5.2 →
$\dfrac1R=\limsup|a_n|^{1/n}$; $\ R=\lim\left|\dfrac{a_n}{a_{n+1}}\right|$ if it exists; same $R$ for $\sum na_nx^{n-1}$ and $\sum\frac{a_n}{n+1}x^{n+1}$; uniform on $[c-r,c+r]$, $r\lt R$.
Abel's theorem§ 5.2 →
$\sum a_n$ converges $\Rightarrow\lim_{x\to1^-}\sum a_nx^n=\sum a_n$; $\ \sum\frac{(-1)^{n+1}}n=\log2$, $\ \sum\frac{(-1)^n}{2n+1}=\frac\pi4$.
Bernstein polynomials§ 5.2 →
$B_nf=\sum_kf(\tfrac kn)\binom nkx^k(1-x)^{n-k}\rightrightarrows f$ for $f\in C[0,1]$; $\ \sum_k(\tfrac kn-x)^2p_{n,k}=\frac{x(1-x)}n$; $\ \int_a^bfx^n=0\ \forall n\Rightarrow f\equiv0$.

Chapter 6 · Functions of Several Variables

Differentiability in ℝⁿ§ 6.1 →
$f$ differentiable at $a$ $\iff$ $\dfrac{\|f(a+h)-f(a)-Df(a)h\|}{\|h\|}\to0$; then $J_f(a)=\left[\partial f_i/\partial x_j(a)\right]$ and $D_vf(a)=Df(a)v$.
Directional derivative§ 6.1 →
$f$ differentiable at $a$, $\|u\|=1$: $\ D_uf(a)=\nabla f(a)\cdot u$, $\ \max_uD_uf(a)=\|\nabla f(a)\|$ at $u=\nabla f(a)/\|\nabla f(a)\|$.
The chain of implications§ 6.1 →
$C^1\Rightarrow$ differentiable $\Rightarrow$ continuous and all $D_uf=\nabla f\cdot u$. None reverses; partials (or all $D_uf$) existing $\not\Rightarrow$ continuous.
Chain rule§ 6.1 →
$g$ diff. at $a$, $f$ diff. at $g(a)$: $\ J_{f\circ g}(a)=J_f(g(a))\,J_g(a)$.
Euler's theorem§ 6.1 →
$f(tx)=t^kf(x)$ $\iff$ $x\cdot\nabla f=kf$ ($f$ differentiable); for $C^2$: $\ \sum_{i,j}x_ix_jf_{x_ix_j}=k(k-1)f$.
Schwarz and Taylor§ 6.1 →
$f_{xy}$ continuous at $a$ (with $f_x,f_y,f_{xy}$ near $a$) $\Rightarrow f_{yx}(a)=f_{xy}(a)$. $f\in C^2$: $\ f(a+h)=f(a)+\nabla f(a)\cdot h+\tfrac12h^{\mathsf T}H_f(a)h+o(\|h\|^2)$.
Inverse function theorem§ 6.2 →
$f\in C^1$ near $a$, $\det J_f(a)\ne0$ $\Rightarrow$ $f$ is a local $C^1$-diffeomorphism at $a$ with $Dg(f(x))=Df(x)^{-1}$. Local only.
Implicit function theorem§ 6.2 →
$F\in C^1$, $F(a,b)=0$, $D_yF(a,b)$ invertible $\Rightarrow$ $y=g(x)$ near $(a,b)$, $Dg=-(D_yF)^{-1}D_xF$; e.g. $g'=-F_x/F_y$, $z_x=-F_x/F_z$.
Second derivative test (two variables)§ 6.2 →
$\nabla f(a)=0$, $f\in C^2$, $D=f_{xx}f_{yy}-f_{xy}^2$: $D\gt0,f_{xx}\gt0$ min; $D\gt0,f_{xx}\lt0$ max; $D\lt0$ saddle; $D=0$ no conclusion.
Lagrange multipliers§ 6.2 →
Extremum of $f$ on $\{g=0\}$ with $\nabla g_1,\dots,\nabla g_k$ independent at $a$ $\Rightarrow$ $\nabla f(a)=\sum_i\lambda_i\nabla g_i(a)$.
Rayleigh§ 6.2 →
$A=A^{\mathsf T}$: $\ \max_{\|x\|=1}x^{\mathsf T}Ax=\lambda_{\max}$, $\ \min_{\|x\|=1}x^{\mathsf T}Ax=\lambda_{\min}$.

Chapter 7 · Metric and Normed Spaces

Metric axioms§ 7.1 →
$d(x,y)\ge0$, $d(x,y)=0\iff x=y$; $\ d(x,y)=d(y,x)$; $\ d(x,z)\le d(x,y)+d(y,z)$; hence $|d(x,z)-d(y,z)|\le d(x,y)$.
Hölder and Minkowski§ 7.1 →
$\tfrac1p+\tfrac1q=1$, $1\lt p\lt\infty$: $\ \sum|x_iy_i|\le\|x\|_p\|y\|_q$; $\ \|x+y\|_p\le\|x\|_p+\|y\|_p$ for $1\le p\le\infty$.
Comparing $\ell^p$ sizes§ 7.1 →
$\mathbb{R}^n$: $\|x\|_\infty\le\|x\|_p\le n^{1/p}\|x\|_\infty$. Sequences: $p\lt q\Rightarrow\ell^p\subsetneq\ell^q$, $\|x\|_q\le\|x\|_p$.
Banach contraction principle§ 7.1 →
$X$ complete, $d(Tx,Ty)\le k\,d(x,y)$, $0\le k\lt1$: unique fixed point $x^*$, and $d(x_n,x^*)\le\dfrac{k^n}{1-k}\,d(x_0,x_1)$.
Baire category theorem§ 7.1 →
$X$ complete, $U_n$ open dense $\Rightarrow\bigcap_nU_n$ dense. So $X\ne\bigcup F_n$ with all $F_n$ closed and $F_n^\circ=\varnothing$; $\mathbb{Q}$ is not a $G_\delta$ in $\mathbb{R}$.
Compactness in metric spaces§ 7.2 →
compact $\iff$ sequentially compact $\iff$ every infinite subset has a limit point $\iff$ complete and totally bounded.
Heine–Borel in $\mathbb{R}^n$§ 7.2 →
$K\subseteq\mathbb{R}^n$ compact $\iff$ closed and bounded. False in $\ell^2$, $C[0,1]$, infinite discrete spaces.
Compact sets in $\ell^p$§ 7.2 →
Closed $S\subseteq\ell^p$ is compact $\iff$ bounded and $\sup_{x\in S}\sum_{n\gt N}|x_n|^p\to0$. Hilbert cube $\{|x_n|\le\tfrac1n\}$: compact; unit ball: not ($\|e_n-e_m\|_p=2^{1/p}$).
Continuous maps on compact spaces§ 7.2 →
$f(K)$ compact; $f$ real attains max/min; $f$ uniformly continuous; a continuous bijection from compact $X$ is a homeomorphism; $K$ compact, $F$ closed, disjoint $\Rightarrow d(K,F)\gt0$.
Lebesgue number; Edelstein§ 7.2 →
Every open cover of a compact metric space has $\lambda\gt0$ with each $B(x,\lambda)$ in one member. Compact $X$, $d(Tx,Ty)\lt d(x,y)$ ($x\ne y$) $\Rightarrow$ unique fixed point.
Connected sets in ℝ and continuous images§ 7.3 →
$A\subseteq\mathbb{R}$ connected $\iff$ $A$ is an interval; $\ X$ connected, $f$ continuous $\Rightarrow f(X)$ connected; $\ f:X\to\mathbb{R}$ $\Rightarrow f(X)$ is an interval.
Building connected sets§ 7.3 →
$\bigcap A_i\ne\varnothing\Rightarrow\bigcup A_i$ connected; $\ A\subseteq B\subseteq\overline A$, $A$ connected $\Rightarrow B$ connected; $\ X\times Y$ connected $\iff X,Y$ connected.
Path-connected versus connected§ 7.3 →
path-connected $\Rightarrow$ connected; $\ U\subseteq\mathbb{R}^n$ open: connected $\iff$ path-connected $\iff$ polygonally connected; $\ \{(x,\sin\frac1x)\}\cup(\{0\}\times[-1,1])$ connected, not path-connected.
Components of matrix groups§ 7.3 →
$GL_n(\mathbb{R})$: $2$ ($\det\gt0$, $\det\lt0$); $\ O(n)$: $2$ ($\det=\pm1$); $\ GL_n(\mathbb{C})$, $SL_n(\mathbb{R})$, $SO(n)$, $U(n)$, nilpotents, singular matrices: connected.
Open subsets of ℝ§ 7.3 →
$G\subseteq\mathbb{R}$ open $\Rightarrow G=\bigsqcup_{j}I_j$, countably many disjoint open intervals (its components).
Norms on ℝⁿ§ 7.4 →
$\|x\|_\infty\le\|x\|_2\le\|x\|_1\le\sqrt n\,\|x\|_2\le n\|x\|_\infty$; $\ $ all norms on a finite-dimensional space are equivalent.
Complete or not§ 7.4 →
Banach: $\mathbb{F}^n$, $\ell^p$ $(1\le p\le\infty)$, $c_0$, $(C(K),\|\cdot\|_\infty)$, $(C^1,\|f\|_\infty+\|f'\|_\infty)$. $\ $ Not: $(C[0,1],\|\cdot\|_1)$, $(C[0,1],\|\cdot\|_2)$, $c_{00}$, polynomials.
Operator norms§ 7.4 →
$\|T\|=\sup_{\|x\|\le1}\|Tx\|$; $\ \|A\|_{1}=\max_j\sum_i|a_{ij}|$, $\ \|A\|_{\infty}=\max_i\sum_j|a_{ij}|$, $\ \|A\|_2=\sqrt{\lambda_{\max}(A^*A)}$; $\ \|\mathcal{K}\|=\max_x\int_0^1|K(x,t)|\,dt$ on $C[0,1]$.
Bounded operators and Neumann series§ 7.4 →
linear $T$: continuous $\iff$ bounded; $\ \|ST\|\le\|S\|\|T\|$; $\ X$ Banach, $\|T\|\lt1\Rightarrow(I-T)^{-1}=\sum_{n\ge0}T^n$, $\ \|(I-T)^{-1}\|\le\frac1{1-\|T\|}$.
Compactness in normed spaces and in C(K)§ 7.4 →
closed unit ball compact $\iff\dim X\lt\infty$; $\ \mathcal{F}\subseteq C(K)$ relatively compact $\iff$ pointwise bounded and equicontinuous ($K$ compact).
Stone–Weierstrass§ 7.4 →
$\mathcal{A}\subseteq C(K,\mathbb{R})$ algebra, $1\in\mathcal{A}$, separates points $\Rightarrow\overline{\mathcal{A}}=C(K)$ (complex: also $\bar f\in\mathcal{A}$). $\ C[a,b]$, $\ell^p$ ($p\lt\infty$) separable; $\ell^\infty$ not.

Chapter 8 · Vector Spaces and Linear Transformations

Dimension formula§ 8.1 →
$U,W$ finite-dimensional: $\ \dim(U+W)=\dim U+\dim W-\dim(U\cap W)$; in $\dim V=n$: $\ \dim(U\cap W)\ge\dim U+\dim W-n$.
Direct sums§ 8.1 →
$U+W$ direct $\iff U\cap W=\{0\}\iff\dim(U+W)=\dim U+\dim W$. For $k\ge3$ summands: $W_j\cap\sum_{i\ne j}W_i=\{0\}$ for every $j$ (pairwise is not enough).
Quotient space§ 8.1 →
$u+W=v+W\iff u-v\in W$; $\ \dim V/W=\dim V-\dim W$ ($\dim V\lt\infty$).
Standard dimensions§ 8.1 →
$\dim M_{m\times n}=mn$, $\dim P_n=n+1$, $\dim\operatorname{Sym}_n=\tfrac{n(n+1)}2$, $\dim\operatorname{Skew}_n=\tfrac{n(n-1)}2$, trace $0$: $n^2-1$, zero row and column sums: $(n-1)^2$.
Counting over $\mathbb{F}_q$§ 8.1 →
$\dim V=n$ over $\mathbb{F}_q$: $|V|=q^n$; ordered bases $\prod_{i=0}^{n-1}(q^n-q^i)$; $k$-dim subspaces $\binom nk_q=\prod_{i=0}^{k-1}\dfrac{q^n-q^i}{q^k-q^i}=\binom n{n-k}_q$.
Rank–nullity§ 8.2 →
$T:V\to W$, $\dim V\lt\infty$: $\ \dim V=\operatorname{rank}T+\operatorname{nullity}T$. If $\dim V=\dim W\lt\infty$: injective $\iff$ surjective $\iff$ bijective.
Matrix of a linear map§ 8.2 →
Column $j$ of $[T]_{\mathcal{B}}^{\mathcal{C}}$ is $[Tb_j]_{\mathcal{C}}$; $\ [Tv]_{\mathcal{C}}=[T]_{\mathcal{B}}^{\mathcal{C}}[v]_{\mathcal{B}}$; $\ [ST]_{\mathcal{B}}^{\mathcal{D}}=[S]_{\mathcal{C}}^{\mathcal{D}}[T]_{\mathcal{B}}^{\mathcal{C}}$; $\ \dim\mathcal{L}(V,W)=mn$.
Change of basis§ 8.2 →
$P$ = new basis in old coordinates: $[v]_{\text{old}}=P[v]_{\text{new}}$, $\ [T]_{\text{new}}=P^{-1}[T]_{\text{old}}P$. Similar matrices share rank, trace, det, characteristic polynomial.
First isomorphism theorem§ 8.2 →
$V/\ker T\cong\operatorname{im}T$ via $v+\ker T\mapsto Tv$.
Duality§ 8.2 →
$f_i(b_j)=\delta_{ij}$; $\ f=\sum f(b_i)f_i$; $\ \dim V^*=\dim V$; $\ \dim W^0=\dim V-\dim W$; $\ [T^{t}]=[T]^{T}$ in dual bases.
Projections and nilpotents§ 8.2 →
$E^2=E\Rightarrow V=\operatorname{im}E\oplus\ker E$, $\operatorname{rank}E=\operatorname{tr}E$ (over $\mathbb{R},\mathbb{C}$). $N$ nilpotent on $\dim n$: $N^n=0$, $(I-N)^{-1}=\sum_{k\ge0}N^k$.
Rank rules§ 8.3 →
$\operatorname{rank}A+\operatorname{rank}B-n\le\operatorname{rank}(AB)\le\min(\operatorname{rank}A,\operatorname{rank}B)$ ($A$: $m\times n$, $B$: $n\times p$); $\ \operatorname{rank}(AB)+\operatorname{rank}(BC)\le\operatorname{rank}B+\operatorname{rank}(ABC)$; $\ \operatorname{rank}(A+B)\le\operatorname{rank}A+\operatorname{rank}B$; $\ \operatorname{rank}A^T=\operatorname{rank}A$; real $A$: $\operatorname{rank}(A^TA)=\operatorname{rank}A$.
Determinant rules§ 8.3 →
$\det(AB)=\det A\det B$, $\ \det A^T=\det A$, $\ \det(cA)=c^n\det A$; swap: $\times(-1)$, scale a row by $c$: $\times c$, $R_i+cR_j$: unchanged; $\ \det\begin{pmatrix}A&B\\0&D\end{pmatrix}=\det A\det D$; $A$ invertible: $\det\begin{pmatrix}A&B\\C&D\end{pmatrix}=\det A\det(D-CA^{-1}B)$.
Adjugate ($n\ge2$)§ 8.3 →
$A\operatorname{adj}A=(\det A)I$; $\ \det\operatorname{adj}A=(\det A)^{n-1}$; $\ \operatorname{adj}(cA)=c^{n-1}\operatorname{adj}A$; $\ \operatorname{adj}\operatorname{adj}A=(\det A)^{n-2}A$; $\ \operatorname{rank}\operatorname{adj}A=n,\,1,\,0$ for $\operatorname{rank}A=n,\ n-1,\ \le n-2$.
Linear systems§ 8.3 →
$Ax=b$ ($A$: $m\times n$, rank $r$) is consistent $\iff\operatorname{rank}A=\operatorname{rank}[A\mid b]$; solutions $x_0+N(A)$, dimension $n-r$ ($q^{n-r}$ of them over $\mathbb{F}_q$). Cramer ($\det A\ne0$): $x_i=\det A_i/\det A$.
Special determinants§ 8.3 →
$\det(x_i^{\,j-1})=\prod_{i\lt j}(x_j-x_i)$; $\ \det(aI_n+bJ_n)=a^{n-1}(a+nb)$; $\ \det(I_m+AB)=\det(I_n+BA)$, so $\det(I+uv^T)=1+v^Tu$.
Special matrices§ 8.3 →
Nilpotent $N\in M_n$: $N^n=0$, $\operatorname{tr}N=\det N=0$, $\det(I+N)=1$, $(I-N)^{-1}=\sum_{k\ge0}N^k$. Idempotent $P$: $\mathbb{F}^n=C(P)\oplus N(P)$, $\operatorname{rank}P+\operatorname{rank}(I-P)=n$, $\operatorname{tr}P=\operatorname{rank}P$ (char $0$). Orthogonal $Q$: $Q^{-1}=Q^T$, $\det Q=\pm1$.

Chapter 9 · Eigenvalues and Canonical Forms

Characteristic polynomial§ 9.1 →
$p_A(x)=\det(xI-A)=x^n-E_1x^{n-1}+E_2x^{n-2}-\cdots+(-1)^nE_n$, $E_k=$ sum of principal $k\times k$ minors; $E_1=\operatorname{tr}A=\sum\lambda_i$, $E_n=\det A=\prod\lambda_i$. For $n=2$: $x^2-(\operatorname{tr}A)x+\det A$.
Multiplicities§ 9.1 →
$g(\lambda)=n-\operatorname{rank}(A-\lambda I)$; $\ 1\le g(\lambda)\le a(\lambda)$; $\ \sum a(\lambda)=n$ iff $p_A$ splits.
Cayley–Hamilton and $m_A$§ 9.1 →
$p_A(A)=0$; $\ f(A)=0\iff m_A\mid f$; if $p_A=\prod(x-\lambda_i)^{a_i}$ then $m_A=\prod(x-\lambda_i)^{e_i}$, $1\le e_i\le a_i$; block diagonal: $m=\operatorname{lcm}$ of the blocks.
Eigenvalues of $f(A)$§ 9.1 →
Over $\mathbb{C}$: $p_{f(A)}(x)=\prod\big(x-f(\lambda_i)\big)$; $\ \operatorname{tr}A^k=\sum\lambda_i^k$; $\ \sigma(A^{-1})=\{\lambda^{-1}\}$.
$AB$ versus $BA$§ 9.1 →
$A$: $m\times n$, $B$: $n\times m$: $\ x^np_{AB}(x)=x^mp_{BA}(x)$; $\ \det(I_m+AB)=\det(I_n+BA)$; $\ p_{uv^T}(x)=x^{n-1}(x-v^Tu)$.
Companion matrix§ 9.1 →
$p_{C(f)}=m_{C(f)}=f$; $\ m_A=p_A\iff A\sim C(p_A)\iff A$ has a cyclic vector.
Diagonalisability§ 9.2 →
$A$ diagonalisable over $\mathbb{F}$ $\iff$ basis of eigenvectors $\iff\sum g(\lambda)=n\iff p_A$ splits and $g(\lambda)=a(\lambda)$ for all $\lambda$ $\iff m_A=\prod(x-\lambda_i)$ with distinct $\lambda_i\in\mathbb{F}$.
Using $A=PDP^{-1}$§ 9.2 →
$A^k=PD^kP^{-1}$, $\ f(A)=Pf(D)P^{-1}$, $\ e^{tA}=Pe^{tD}P^{-1}$; columns of $P$ are eigenvectors.
Primary decomposition§ 9.2 →
$m_T=\prod q_i^{e_i}$: $\ V=\bigoplus_i\ker q_i(T)^{e_i}$; split case: $\dim\ker(T-\lambda_iI)^{e_i}=a(\lambda_i)$; projections $E_i=g_i(T)f_i(T)$ from $\sum g_if_i=1$, $f_i=m_T/q_i^{e_i}$.
Triangular forms and Schur§ 9.2 →
Triangularisable $\iff p_A$ splits; $A\in M_n(\mathbb{C})$: $A=UTU^*$, $U$ unitary; $\ \sum|\lambda_i|^2\le\sum|a_{ij}|^2$, equality iff $A$ is unitarily diagonalisable.
Commuting matrices§ 9.2 →
$A,B$ diagonalisable: common eigenbasis $\iff AB=BA$. Commuting complex $A,B$: simultaneously (unitarily) triangularisable.
Jordan blocks from ranks§ 9.3 →
$r_k=\operatorname{rank}(A-\lambda I)^k$, $r_0=n$: number of blocks $J_j(\lambda)$ with $j\ge k$ is $r_{k-1}-r_k$; with $j=k$ exactly it is $r_{k-1}-2r_k+r_{k+1}$. The ranks stabilise at $n-a_\lambda$.
What the Jordan form records§ 9.3 →
For an eigenvalue $\lambda$: number of blocks $=g_\lambda=n-\operatorname{rank}(A-\lambda I)$; total size $=a_\lambda$; largest block $=$ exponent of $(x-\lambda)$ in $m_A$. Diagonalisable $\iff$ all blocks $1\times1$.
Powers of a Jordan block§ 9.3 →
$J_k(\lambda)^m=\sum_{j=0}^{k-1}\binom{m}{j}\lambda^{m-j}N^j$, $N=J_k(0)$; $\ \operatorname{rank}N^j=\max(k-j,0)$.
Counting similarity classes§ 9.3 →
With $p_A=\prod(x-\lambda_i)^{a_i}$, $m_A=\prod(x-\lambda_i)^{m_i}$: $\prod_iP(a_i,m_i)$ classes ($P(a,m)$ = partitions of $a$ with largest part $m$). Nilpotent $n\times n$: $p(n)$ classes; $p(1),\dots,p(8)=1,2,3,5,7,11,15,22$.
Rational canonical form§ 9.3 →
$A\sim\operatorname{diag}(C(f_1),\dots,C(f_r))$, $f_1\mid\cdots\mid f_r$, $f_r=m_A$, $\prod f_i=p_A$; $p_{C(f)}=m_{C(f)}=f$. From Jordan data: $f_r=\prod_\lambda(x-\lambda)^{k_1(\lambda)}$, $f_{r-1}=\prod_\lambda(x-\lambda)^{k_2(\lambda)}$, …

Chapter 10 · Inner Product Spaces and Quadratic Forms

Cauchy–Schwarz and the parallelogram law§ 10.1 →
$|\langle x,y\rangle|\le\|x\|\,\|y\|$ (equality iff $x,y$ dependent); $\|x+y\|^2+\|x-y\|^2=2\|x\|^2+2\|y\|^2$; real polarisation $\langle x,y\rangle=\tfrac14\left(\|x+y\|^2-\|x-y\|^2\right)$
Gram–Schmidt (independent $v_1,v_2,\dots$)§ 10.1 →
$u_k=v_k-\sum_{j\lt k}\langle v_k,e_j\rangle e_j,\quad e_k=u_k/\|u_k\|$; $A=QR$ with $r_{jk}=\langle a_k,q_j\rangle$, $r_{kk}=\|u_k\|\gt0$
Orthogonal projection onto a finite-dimensional $W$§ 10.1 →
$P_Wx=\sum_j\langle x,e_j\rangle e_j$ ($e_j$ orthonormal basis of $W$); $\operatorname{dist}(x,W)^2=\|x\|^2-\sum_j|\langle x,e_j\rangle|^2$; for $W=C(A)$ with independent columns $P=A(A^*A)^{-1}A^*$; onto a line $P=\dfrac{vv^*}{v^*v}$
Bessel and Parseval§ 10.1 →
$\sum_k|\langle x,e_k\rangle|^2\le\|x\|^2$ for any orthonormal $(e_k)$; equality for all $x$ iff $(e_k)$ is complete
The adjoint§ 10.1 →
$\langle Tx,y\rangle=\langle x,T^*y\rangle$; $[T^*]=[T]^*$ in orthonormal bases; $\ker T^*=(\operatorname{im}T)^\perp$, $\operatorname{im}T^*=(\ker T)^\perp$
Least squares: normal equations§ 10.1 →
$\min\|Ax-b\|\iff A^*A\hat x=A^*b$; unique iff $\operatorname{rank}A=n$, then $\hat x=(A^*A)^{-1}A^*b$
Special classes and their spectra§ 10.2 →
Hermitian $A^*=A$: $\lambda\in\mathbb{R}$; skew-Hermitian $A^*=-A$: $\lambda\in i\mathbb{R}$; unitary $A^*A=I$: $|\lambda|=1$; normal $A^*A=AA^*$: $Ax=\lambda x\iff A^*x=\bar\lambda x$. For normal $A$ these implications reverse.
Spectral theorem§ 10.2 →
$A\in M_n(\mathbb{C})$: $A=UDU^*$ ($U$ unitary) $\iff A$ normal. $A\in M_n(\mathbb{R})$: $A=QDQ^T$ ($Q$ orthogonal, $D$ real) $\iff A^T=A$. $\ A=\sum\lambda_jP_j$, $f(A)=\sum f(\lambda_j)P_j$.
Positive (semi)definite§ 10.2 →
$A=A^*\succeq0\iff$ all $\lambda\ge0\iff A=B^*B$; unique $\sqrt A=\sum\sqrt{\lambda_j}P_j\succeq0$. $A\succ0\Rightarrow a_{ii}\gt0$, $|a_{ij}|^2\lt a_{ii}a_{jj}$.
SVD and polar decomposition§ 10.2 →
$A=U\Sigma V^*$, $\sigma_i=\sqrt{\lambda_i(A^*A)}$; $\ \|A\|=\sigma_1$, $\|A\|_F^2=\sum\sigma_i^2$, $|\det A|=\prod\sigma_i$; normal $A$: $\sigma_i=|\lambda_i|$, $\|A\|=\rho(A)$. Polar: $A=WP$, $P=\sqrt{A^*A}$.
Locating eigenvalues§ 10.2 →
Gershgorin: $\sigma(A)\subseteq\bigcup_i\{|z-a_{ii}|\le\sum_{j\ne i}|a_{ij}|\}$. Hermitian: $\lambda_{\min}\le\dfrac{x^*Ax}{x^*x}\le\lambda_{\max}$. Strictly diagonally dominant $\Rightarrow$ invertible.
Matrix of a quadratic form§ 10.3 →
$q=\sum a_{ii}x_i^2+2\sum_{i\lt j}a_{ij}x_ix_j=x^TAx$; $\ B(x,y)=\tfrac12\big(q(x+y)-q(x)-q(y)\big)$; $\ x=Py\Rightarrow A\mapsto P^TAP$ (congruence).
Inertia, rank, signature§ 10.3 →
$A\cong\operatorname{diag}(I_{n_+},-I_{n_-},0_{n_0})$, $(n_+,n_-,n_0)$ = sign counts of eigenvalues (invariant). Rank $n_++n_-$, signature $n_+-n_-$. Real classes: $\tfrac{(n+1)(n+2)}2$; complex symmetric: $n+1$.
Sylvester's criterion and Jacobi§ 10.3 →
$A\succ0\iff D_1,\dots,D_n\gt0$; $\ A\prec0\iff(-1)^kD_k\gt0$; $\ A\succeq0\iff$ all principal minors $\ge0$. All $D_k\ne0$: $q\cong\sum\frac{D_k}{D_{k-1}}y_k^2$.
Conics $ax^2+2hxy+by^2+2gx+2fy+c=0$§ 10.3 →
$\delta=ab-h^2$, $\Delta=\det\begin{pmatrix}a&h&g\\h&b&f\\g&f&c\end{pmatrix}$. $\Delta\ne0$: $\delta\gt0$ ellipse (real iff $(a+b)\Delta\lt0$), $\delta=0$ parabola, $\delta\lt0$ hyperbola. $\Delta=0$: degenerate. Centre form $\lambda_1u^2+\lambda_2v^2+\Delta/\delta=0$.
Central quadrics $x^TAx=1$§ 10.3 →
Inertia $(3,0,0)$ ellipsoid; $(2,1,0)$ hyperboloid of one sheet; $(1,2,0)$ of two sheets; $(2,0,1)$ elliptic cylinder; $(1,1,1)$ hyperbolic cylinder.