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 · First-Order Ordinary Differential Equations
Picard–Lindelöf theorem§ 1.1 →
$f$ continuous on $R=\{|x-x_0|\le a,|y-y_0|\le b\}$, $|f|\le M$, Lipschitz in $y$ with constant $L$ $\Rightarrow$ unique solution on $|x-x_0|\le h=\min(a,b/M)$; $\ |\varphi-\varphi_n|\le\tfrac ML\tfrac{(Lh)^{n+1}}{(n+1)!}e^{Lh}$
Picard iterates§ 1.1 →
$\varphi_0=y_0$, $\ \varphi_{n+1}(x)=y_0+\int_{x_0}^xf(t,\varphi_n(t))\,dt$
Lipschitz test§ 1.1 →
$D$ convex in $y$, $|\partial f/\partial y|\le L$ on $D$ $\Rightarrow$ $|f(x,y_1)-f(x,y_2)|\le L|y_1-y_2|$
Gronwall and continuous dependence§ 1.1 →
$u\le C+\int_{x_0}^xku$, $k\ge0$ $\Rightarrow u\le Ce^{\int_{x_0}^xk}$. $\ |f-g|\le\varepsilon$, $f$ $L$-Lipschitz: $|y-z|\le|y_0-z_0|e^{L|x-x_0|}+\tfrac\varepsilon L\left(e^{L|x-x_0|}-1\right)$
Standard examples§ 1.1 →
$y'=y^2$, $y(0)=y_0\gt0$: $y=\tfrac{y_0}{1-y_0x}$ on $(-\infty,\tfrac1{y_0})$. $\ y'=|y|^\alpha$, $y(0)=0$: unique iff $\alpha\ge1$; for $0\lt\alpha\lt1$ also $y=((1-\alpha)(x-c))^{1/(1-\alpha)}$, $x\ge c$.
Solution tools§ 1.1 →
$y'+Py=Q$: $y=e^{-\int P}\left(\int Qe^{\int P}+C\right)$. Bernoulli $y'+Py=Qy^n$: $v=y^{1-n}$. Exact iff $M_y=N_x$; $\mu(x)=e^{\int\frac{M_y-N_x}{N}dx}$.
Clairaut's equation§ 1.2 →
$y=px+f(p)$: general $y=cx+f(c)$; singular ($f''\ne0$): $x=-f'(p)$, $y=f(p)-pf'(p)$
Lagrange's equation§ 1.2 →
$y=xg(p)+f(p)$: $\ (p-g(p))\tfrac{dx}{dp}-g'(p)x=f'(p)$; plus the lines $y=p_0x+f(p_0)$ with $g(p_0)=p_0$
Discriminant loci§ 1.2 →
$c$-disc.: eliminate $c$ from $\varphi=0,\ \varphi_c=0$ — typically $EN^2C^3$. $\ p$-disc.: eliminate $p$ from $F=0,\ F_p=0$ — typically $ET^2C$. Singular solutions lie on the $p$-discriminant; test each factor in the equation.
Orthogonal trajectories§ 1.2 →
Cartesian: $p\mapsto-\tfrac1p$. Polar: $\tfrac{dr}{d\theta}\mapsto-r^2\tfrac{d\theta}{dr}$ (eliminate $c$ first)
Liouville's formula§ 1.3 →
$\Phi'=A(t)\Phi$, $A$ continuous: $\det\Phi(t)=\det\Phi(t_0)\exp\int_{t_0}^t\operatorname{tr}A(s)\,ds$; Abel: $W=W_0e^{-\int p}$ for $y''+py'+qy=0$
$e^{tA}$ for $2\times2$ matrices§ 1.3 →
$\lambda$ double: $e^{\lambda t}(I+t(A-\lambda I))$; $\ \lambda_1\ne\lambda_2$: $\frac{e^{\lambda_1t}(A-\lambda_2I)-e^{\lambda_2t}(A-\lambda_1I)}{\lambda_1-\lambda_2}$; $\ \alpha\pm i\beta$: $e^{\alpha t}\left(\cos\beta t\,I+\frac{\sin\beta t}{\beta}(A-\alpha I)\right)$
Generalised eigenvectors§ 1.3 →
$v\in\ker(A-\lambda I)^m$: $\ e^{tA}v=e^{\lambda t}\sum_{k=0}^{m-1}\frac{t^k}{k!}(A-\lambda I)^kv$
Variation of parameters for systems§ 1.3 →
$x'=A(t)x+b$: $x=\Phi(t)\Phi(t_0)^{-1}x_0+\Phi(t)\int_{t_0}^t\Phi(s)^{-1}b(s)\,ds$; constant $A$: $x=e^{(t-t_0)A}x_0+\int_{t_0}^te^{(t-s)A}b(s)\,ds$
Planar classification ($\tau=\operatorname{tr}A$, $\Delta=\det A$)§ 1.3 →
$\Delta\lt0$ saddle; $\Delta\gt0$: $\tau^2\ge4\Delta$ node, $\tau^2\lt4\Delta$ spiral ($\tau\ne0$), $\tau=0$ centre; asymptotically stable iff $\tau\lt0$, $\Delta\gt0$
Stability of $x'=Ax$§ 1.3 →
Asymptotically stable iff all $\operatorname{Re}\lambda\lt0$; stable iff all $\operatorname{Re}\lambda\le0$ with $\operatorname{Re}\lambda=0$ semisimple. Linearisation: $\operatorname{Re}\lambda(Df(x^*))\lt0$ ⇒ $x^*$ asymptotically stable.
Chapter 2 · Linear Ordinary Differential Equations
Abel's formula§ 2.1 →
Solutions of $y^{(n)}+p_{n-1}y^{(n-1)}+\dots+p_0y=0$ ($p_k$ continuous, normal form): $W(x)=W(x_0)\exp\left(-\int_{x_0}^xp_{n-1}\right)$; $W\equiv0$ or $W$ never $0$
Reduction of order§ 2.1 →
$y_1\ne0$ solves $y''+py'+qy=0$: $\ y_2=y_1\displaystyle\int\frac{e^{-\int p\,dx}}{y_1^2}\,dx$, $\ W(y_1,y_2)=e^{-\int p}$
Constant coefficients§ 2.1 →
Root $\lambda$ of multiplicity $m$: $x^ke^{\lambda x}$, $k\lt m$; real equation, roots $\alpha\pm i\beta$: $x^ke^{\alpha x}\cos\beta x$, $x^ke^{\alpha x}\sin\beta x$
Euler–Cauchy equations§ 2.1 →
$x^2y''+b_1xy'+b_0y=0$: $r^2+(b_1-1)r+b_0=0$; $\ x^{r_1},x^{r_2}$; double $r$: $x^r,\ x^r\log x$; $\ \alpha\pm i\beta$: $x^\alpha\cos(\beta\log x)$, $x^\alpha\sin(\beta\log x)$
Normal form§ 2.1 →
$y=u\,e^{-\frac12\int p}$ turns $y''+py'+qy=0$ into $u''+Qu=0$, $\ Q=q-\tfrac14p^2-\tfrac12p'$; $\ Q\le0$ ⇒ at most one zero
Equation with given solutions $y_1,y_2$§ 2.1 →
$W=W(y_1,y_2)\ne0$ on $J$: $\ y''-\dfrac{W'}{W}y'+\dfrac{y_1'y_2''-y_1''y_2'}{W}y=0$
Variation of parameters§ 2.2 →
$y''+py'+qy=r$ (normal form): $\ y_p=-y_1\displaystyle\int\frac{y_2r}{W}\,dx+y_2\int\frac{y_1r}{W}\,dx$; order $n$: $y_p=\sum_ky_k\int\frac{W_k}{W}r$
Kernel (Duhamel)§ 2.2 →
$y_p=\displaystyle\int_{x_0}^xK(x,t)r(t)\,dt$, $\ K=\frac{y_1(t)y_2(x)-y_1(x)y_2(t)}{W(t)}$; $\ y''+\omega^2y=r$: $K=\frac{\sin\omega(x-t)}{\omega}$
Undetermined coefficients§ 2.2 →
$r=e^{\alpha x}(R_1\cos\beta x+R_2\sin\beta x)$, $\deg R_i\le k$, $\alpha+i\beta$ a root of multiplicity $s$: $\ y_p=x^se^{\alpha x}(A_1\cos\beta x+A_2\sin\beta x)$, $\deg A_i\le k$
Operator rules $1/f(D)$§ 2.2 →
$\frac{1}{f(D)}e^{ax}=\frac{e^{ax}}{f(a)}$ or $\frac{x^se^{ax}}{f^{(s)}(a)}$; $\ \frac{1}{g(D^2)}\sin ax=\frac{\sin ax}{g(-a^2)}$; $\ \frac{1}{D^2+a^2}\sin ax=-\frac{x}{2a}\cos ax$; $\ \frac1{f(D)}e^{ax}V=e^{ax}\frac{1}{f(D+a)}V$
Frobenius at a regular singular point§ 2.2 →
$r(r-1)+p_0r+q_0=0$; $\ r_1-r_2\notin\mathbb{Z}$: two series; $r_1=r_2$: $\log$ always; $r_1-r_2\in\mathbb{N}$: $\log$ possible
Legendre and Bessel§ 2.2 →
$P_n=\frac{1}{2^nn!}\frac{d^n}{dx^n}(x^2-1)^n$, $\int_{-1}^1P_mP_n=\frac{2\delta_{mn}}{2n+1}$; $\ J_\nu=\sum\frac{(-1)^k(x/2)^{2k+\nu}}{k!\,\Gamma(k+\nu+1)}$, $(x^\nu J_\nu)'=x^\nu J_{\nu-1}$, $J_0'=-J_1$
Self-adjoint form§ 2.3 →
$a_2y''+a_1y'+a_0y+\lambda by=0$ ($a_2\gt0$) $\iff(py')'+qy+\lambda ry=0$, $\ p=e^{\int a_1/a_2}$, $q=\frac{a_0p}{a_2}$, $r=\frac{bp}{a_2}$
Green's formula and orthogonality§ 2.3 →
$\int_a^b(uLv-vLu)=\big[p(uv'-u'v)\big]_a^b$; regular SL: $\lambda$ real, simple, $\lambda_1\lt\lambda_2\lt\cdots\to\infty$, $\int_a^br\varphi_m\varphi_n=0$ ($m\ne n$), $\varphi_n$ has $n-1$ zeros in $(a,b)$
Eigenvalues of $y''+\lambda y=0$ on $[0,L]$§ 2.3 →
D–D: $\left(\frac{n\pi}{L}\right)^2$, $n\ge1$; N–N: $\left(\frac{n\pi}{L}\right)^2$, $n\ge0$; D–N or N–D: $\left(\frac{(2n-1)\pi}{2L}\right)^2$; periodic on $[-L,L]$: $\left(\frac{n\pi}{L}\right)^2$, double for $n\ge1$
Rayleigh quotient§ 2.3 →
$\lambda_1=\min\dfrac{-[puu']_a^b+\int_a^b(pu'^2-qu^2)}{\int_a^bru^2}$; $\ \lambda\ge0$ if $q\le0$, $\alpha_1\alpha_2\le0$, $\beta_1\beta_2\ge0$
Sturm comparison: zero spacing§ 2.3 →
$u''+Q(x)u=0$, $0\lt m^2\le Q\le M^2$: consecutive zeros satisfy $\frac{\pi}{M}\le x_2-x_1\le\frac{\pi}{m}$
Singular SL families§ 2.3 →
$\int_{-1}^1P_mP_n=\frac{2\delta_{mn}}{2n+1}$; $\int_0^1xJ_\nu(j_kx)J_\nu(j_mx)=\frac{\delta_{km}}2J_{\nu+1}(j_k)^2$; weights: Chebyshev $(1-x^2)^{-1/2}$, Hermite $e^{-x^2}$, Laguerre $e^{-x}$
Green's function of $(pu')'+qu=f$§ 2.4 →
$B_a[u_1]=0$, $B_b[u_2]=0$, $C=p\,(u_1u_2'-u_1'u_2)$ (constant $\ne0$): $\ G(x,t)=\dfrac{u_1(\min(x,t))\,u_2(\max(x,t))}{C}$, $\ [\partial_xG]_{x=t}=\dfrac1{p(t)}$, $\ u=\int_a^bG(x,t)f(t)\,dt$
Standard Green's functions of $-u''=f$ on $[0,1]$§ 2.4 →
$u(0)=u(1)=0$: $\min(x,t)\big(1-\max(x,t)\big)$; $\ u(0)=u'(1)=0$: $\min(x,t)$; $\ u'(0)=u(1)=0$: $1-\max(x,t)$; for $u''=f$ change the sign
Fredholm alternative§ 2.4 →
If $L\varphi=0$, $B_a\varphi=B_b\varphi=0$, $\varphi\not\equiv0$: $\ L[u]=f$ solvable $\iff\int_a^bf\varphi\,dx=0$; then $u=u^*+c\varphi$. Neumann $-u''=f$ on $[0,1]$: $\int_0^1f=0$, $G_M=\tfrac{x^2+t^2}2-\max(x,t)+\tfrac13$
Eigenfunction expansion of $G$§ 2.4 →
$L[u]+\mu ru=f$: $\ G(x,t;\mu)=\displaystyle\sum_n\frac{\varphi_n(x)\varphi_n(t)}{\mu-\lambda_n}$ ($\int r\varphi_n^2=1$); $\ \int G(x,x;\mu)\,r=\sum\frac1{\mu-\lambda_n}$
Causal Green's function (IVP)§ 2.4 →
$u(a)=\dots=u^{(n-1)}(a)=0$: $\ G(x,t)=K(x,t)$ for $t\le x$, $0$ for $t\gt x$; e.g. $u''+\omega^2u=f$: $\ G=\frac{\sin\omega(x-t)}{\omega}H(x-t)$
Chapter 3 · First-Order Partial Differential Equations
Eliminating an arbitrary function§ 3.1 →
$\varphi(u,v)=0\ \Rightarrow\ P'p+Q'q=R'$, $\ (P',Q',R')=\nabla u\times\nabla v=\left(\frac{\partial(u,v)}{\partial(y,z)},\frac{\partial(u,v)}{\partial(z,x)},\frac{\partial(u,v)}{\partial(x,y)}\right)$
Lagrange's method for $Pp+Qq=R$§ 3.1 →
$\frac{dx}{P}=\frac{dy}{Q}=\frac{dz}{R}$; two independent first integrals $u=c_1$, $v=c_2$ ($Pu_x+Qu_y+Ru_z=0$, $\nabla u\times\nabla v\ne0$) $\Rightarrow$ general integral $\varphi(u,v)=0$
Multipliers§ 3.1 →
$\ell P+mQ+nR=0$ and $\ell\,dx+m\,dy+n\,dz=dw$ $\Rightarrow$ $w=c$; also $\frac{dx}{P}=\frac{\ell dx+m dy+n dz}{\ell P+mQ+nR}$
$a\,p+b\,q=0$§ 3.1 →
$z$ is constant on the curves $\frac{dx}{a}=\frac{dy}{b}$: $z=f(u(x,y))$; e.g. $yp-xq=0\Rightarrow z=f(x^2+y^2)$, $xp+yq=nz\Rightarrow z=x^nf(y/x)$
Surfaces orthogonal to $f(x,y,z)=c$§ 3.1 →
$f_x\,p+f_y\,q=f_z$ $\ $ (normals $(p,q,-1)\perp\nabla f$)
Charpit's equations for $F(x,y,z,p,q)=0$§ 3.2 →
$\dfrac{dx}{F_p}=\dfrac{dy}{F_q}=\dfrac{dz}{pF_p+qF_q}=\dfrac{dp}{-(F_x+pF_z)}=\dfrac{dq}{-(F_y+qF_z)}$; a first integral $G=a$ with $\partial(F,G)/\partial(p,q)\ne0$, then $dz=p\,dx+q\,dy$
Compatibility of $F=0$, $G=0$ ($J=\partial(F,G)/\partial(p,q)\ne0$)§ 3.2 →
$\frac{\partial(F,G)}{\partial(x,p)}+p\frac{\partial(F,G)}{\partial(z,p)}+\frac{\partial(F,G)}{\partial(y,q)}+q\frac{\partial(F,G)}{\partial(z,q)}=0$ on $F=G=0$
Standard forms§ 3.2 →
$f(p,q)=0$: $z=ax+h(a)y+c$; $\ z=px+qy+f(p,q)$: $z=ax+by+f(a,b)$; $\ f(z,p,q)=0$: $q=ap$, $\int\frac{dz}{p(z,a)}=x+ay+b$; $\ f_1(x,p)=f_2(y,q)=a$: $z=\int p\,dx+\int q\,dy+b$
Singular integral§ 3.2 →
Eliminate $a,b$ from $z=\phi$, $\phi_a=0$, $\phi_b=0$; it lies in $F=F_p=F_q=0$ (check it solves!)
Cauchy problem for $Pp+Qq=R$§ 3.2 →
Unique local solution if $J=x_0'Q-y_0'P\ne0$; if $J\equiv0$: none, or infinitely many when $\Gamma$ is characteristic. Burgers $z_y+zz_x=0$, $z(x,0)=g$: $z=g(x-zy)$, breaks at $y^*=-1/\min g'$
Chapter 4 · Second-Order Partial Differential Equations
Type of $Au_{xx}+Bu_{xy}+Cu_{yy}+\dots=0$§ 4.1 →
$B^2-4AC\gt0$ hyperbolic, $=0$ parabolic, $\lt0$ elliptic ($A,B,C$ not all $0$); under $(\xi,\eta)$ with $J\ne0$: $\ \bar B^2-4\bar A\bar C=J^2(B^2-4AC)$
Characteristics and canonical forms§ 4.1 →
$A\left(\frac{dy}{dx}\right)^2-B\frac{dy}{dx}+C=0$; $\ \xi,\eta=$ first integrals: $u_{\xi\eta}=\Phi$ (hyp.), $u_{\eta\eta}=\Phi$ (par.), $u_{\alpha\alpha}+u_{\beta\beta}=\Phi$ (ell., $\alpha+i\beta$ complex first integral)
CF of $F(D,D')z=0$§ 4.1 →
$(D-mD'-c)^r\ \to\ e^{cx}\sum_{j=0}^{r-1}x^j\varphi_j(y+mx)$; homogeneous: roots of $F(m,1)=0$; irreducible: $\sum Ae^{ax+by}$ with $F(a,b)=0$
Particular integrals§ 4.1 →
$\frac{e^{ax+by}}{F(a,b)}$ if $F(a,b)\ne0$; $\ \frac{x\,e^{ax+by}}{F_X(a,b)}$ if $F(a,b)=0\ne F_X(a,b)$; $\ \frac{1}{(D-mD')^r}\varphi(y+mx)=\frac{x^r}{r!}\varphi(y+mx)$; $\sin(ax+by)$: $D^2\to-a^2$, $DD'\to-ab$, $D'^2\to-b^2$
One factor at a time§ 4.1 →
$\frac{1}{D-mD'}h=\int h(x,\,w-mx)\,dx\,\Big|_{w=y+mx}$
Harmonic functions§ 4.2 →
$\Delta u=0$: $\ u(\mathbf x_0)=\frac1{2\pi}\int_{-\pi}^{\pi}u(\mathbf x_0+\rho\mathbf e_\theta)\,d\theta$; $\ \Omega$ bounded: $\min_{\partial\Omega}u\le u\le\max_{\partial\Omega}u$; polar $\Delta u=u_{rr}+\frac1ru_r+\frac1{r^2}u_{\theta\theta}$
Fourier series on $(-L,L)$§ 4.2 →
$a_n=\frac1L\int_{-L}^{L}f\cos\frac{n\pi x}L\,dx$, $\ b_n=\frac1L\int_{-L}^{L}f\sin\frac{n\pi x}L\,dx$; piecewise $C^1$: sum $=\frac12(f(x^+)+f(x^-))$; $\ \frac1L\int f^2=\frac{a_0^2}2+\sum(a_n^2+b_n^2)$
Rectangle $(0,a)\times(0,b)$, data $u(x,0)=f$§ 4.2 →
$u=\sum b_n\frac{\sinh(n\pi(b-y)/a)}{\sinh(n\pi b/a)}\sin\frac{n\pi x}a$, $\ b_n=\frac2a\int_0^af\sin\frac{n\pi x}a\,dx$ ($f$ continuous, piecewise $C^1$, $f(0)=f(a)=0$)
Disc $r\lt R$: Poisson's formula§ 4.2 →
$u=\frac{a_0}2+\sum\big(\frac rR\big)^n(a_n\cos n\theta+b_n\sin n\theta)=\frac1{2\pi}\int_{-\pi}^{\pi}\frac{R^2-r^2}{R^2-2Rr\cos(\theta-\varphi)+r^2}f(\varphi)\,d\varphi$
Neumann compatibility§ 4.2 →
$\Delta u=F$, $\partial_nu=g$: $\ \iint_\Omega F\,dA=\oint_{\partial\Omega}g\,ds$; $u$ unique up to a constant
Heat equation on $[0,L]$§ 4.3 →
$u=0$ at ends: $u=\sum b_ne^{-kn^2\pi^2t/L^2}\sin\frac{n\pi x}L$; insulated ends: $u=\frac{a_0}2+\sum a_ne^{-kn^2\pi^2t/L^2}\cos\frac{n\pi x}L\to\frac1L\int_0^Lf$
Principles for the heat equation§ 4.3 →
$\max_{\overline{Q_T}}u=\max_{\Gamma_T}u$ (bottom and sides); $\ \frac{d}{dt}\int_0^Lu^2=-2k\int_0^Lu_x^2\le0$ for $u=0$ or $u_x=0$ at the ends
d'Alembert's formula§ 4.3 →
$u=\frac12\big[f(x+ct)+f(x-ct)\big]+\frac1{2c}\int_{x-ct}^{x+ct}g(s)\,ds$; $u(x_0,t_0)$ depends on data in $[x_0-ct_0,x_0+ct_0]$
String with fixed ends§ 4.3 →
$u=\sum\big(a_n\cos\frac{n\pi ct}L+b_n\sin\frac{n\pi ct}L\big)\sin\frac{n\pi x}L$, $\ b_n=\frac{2}{n\pi c}\int_0^Lg\sin\frac{n\pi x}L$; $\ E=\frac12\int(u_t^2+c^2u_x^2)=\frac L4\sum\omega_n^2(a_n^2+b_n^2)$
Heat kernel§ 4.3 →
$\Phi=\frac{1}{\sqrt{4\pi kt}}e^{-x^2/(4kt)}$, $\ u=\int_{-\infty}^\infty\Phi(x-y,t)f(y)\,dy$
Chapter 5 · Numerical Analysis
Order of convergence§ 5.1 →
$\lim\dfrac{|e_{n+1}|}{|e_n|^p}=C\in(0,\infty)$ ($C\lt1$ if $p=1$); correct digits: $d_{n+1}\approx p\,d_n-\log_{10}C$
Bisection ($f$ continuous, $f(a)f(b)\lt0$)§ 5.1 →
$|c_n-\alpha|\le\dfrac{b-a}{2^n}$; $\ n\ge\log_2\dfrac{b-a}{\varepsilon}$ steps
Fixed-point iteration $x_{n+1}=g(x_n)$§ 5.1 →
$|g'(\alpha)|\lt1$: converges, $\dfrac{e_{n+1}}{e_n}\to g'(\alpha)$; $\ |g'|\le L\lt1$ on $[a,b]$, $g([a,b])\subseteq[a,b]$: $|x_n-\alpha|\le\dfrac{L^n}{1-L}|x_1-x_0|$; $\ g'(\alpha)=\dots=g^{(p-1)}(\alpha)=0\ne g^{(p)}(\alpha)$: order $p$
Newton–Raphson§ 5.1 →
$x_{n+1}=x_n-\dfrac{f(x_n)}{f'(x_n)}$; simple root: $\dfrac{e_{n+1}}{e_n^2}\to\dfrac{f''(\alpha)}{2f'(\alpha)}$; $m$-fold root: $\dfrac{e_{n+1}}{e_n}\to1-\dfrac1m$, restored by $x_n-m\dfrac{f(x_n)}{f'(x_n)}$
Secant method and regula falsi§ 5.1 →
$x_{n+1}=x_n-f(x_n)\dfrac{x_n-x_{n-1}}{f(x_n)-f(x_{n-1})}$; secant: $e_{n+1}\approx\dfrac{f''}{2f'}e_ne_{n-1}$, order $\tfrac{1+\sqrt5}2$; regula falsi (fixed end $\beta$): linear, $C=1-\dfrac{f'(\alpha)(\beta-\alpha)}{f(\beta)}$
Newton for a system§ 5.1 →
$J(\mathbf{x}^{(n)})\,\mathbf{d}=-F(\mathbf{x}^{(n)})$, $\ \mathbf{x}^{(n+1)}=\mathbf{x}^{(n)}+\mathbf{d}$; quadratic if $J$ is nonsingular at the root
Gauss elimination§ 5.2 →
$m_{ik}=\dfrac{a^{(k)}_{ik}}{a^{(k)}_{kk}}$; pivots $u_{kk}=\dfrac{\Delta_k}{\Delta_{k-1}}$; cost $\dfrac{n^3}3+n^2-\dfrac n3$ mult./div. and $\dfrac{n^3}3+\dfrac{n^2}2-\dfrac{5n}6$ add. ($n=3$: $17$, $11$)
LU and Cholesky§ 5.2 →
Doolittle: $u_{kj}=a_{kj}-\sum_{s\lt k}l_{ks}u_{sj}$, $l_{ik}=\dfrac{a_{ik}-\sum_{s\lt k}l_{is}u_{sk}}{u_{kk}}$; $\det A=\prod u_{kk}$; Cholesky (SPD): $l_{kk}=\sqrt{a_{kk}-\sum_{s\lt k}l_{ks}^2}$
Jacobi and Gauss–Seidel ($A=L+D+U$)§ 5.2 →
$T_J=-D^{-1}(L+U)$, $\ T_{GS}=-(D+L)^{-1}U$; converge for all $x^{(0)}$ iff $\rho(T)\lt1$; strictly diagonally dominant $\Rightarrow$ both converge; SPD $\Rightarrow$ Gauss–Seidel converges; tridiagonal: $\rho(T_{GS})=\rho(T_J)^2$
$2\times2$ case $\begin{pmatrix}a&b\\c&d\end{pmatrix}$§ 5.2 →
$\rho(T_J)=\sqrt{\left|\tfrac{bc}{ad}\right|}$, $\ \rho(T_{GS})=\left|\tfrac{bc}{ad}\right|$
SOR§ 5.2 →
$\rho(T_\omega)\ge|\omega-1|$; $\ \omega_{\rm opt}=\dfrac{2}{1+\sqrt{1-\rho_J^2}}$, $\ \rho(T_{\omega_{\rm opt}})=\omega_{\rm opt}-1$
Condition number§ 5.2 →
$\kappa(A)=\|A\|\,\|A^{-1}\|\ge1$; $\ \dfrac{\|\delta x\|}{\|x\|}\le\kappa(A)\dfrac{\|\delta b\|}{\|b\|}$
Difference operators (step $h$)§ 5.3 →
$\Delta=E-1$, $\ \nabla=1-E^{-1}$, $\ \delta=E^{1/2}-E^{-1/2}$, $\ \Delta\nabla=\Delta-\nabla=\delta^2$; $\ \Delta^kf_i=\sum_j(-1)^{k-j}\binom kjf_{i+j}$; $\ \deg p=n$, leading coeff. $a$: $\Delta^np=n!\,a\,h^n$, $\Delta^{n+1}p=0$
Lagrange form and error ($n+1$ distinct nodes, $f\in C^{n+1}$)§ 5.3 →
$p_n=\sum_iy_iL_i$, $\ L_i(x)=\prod_{j\ne i}\dfrac{x-x_j}{x_i-x_j}$; $\ f(x)-p_n(x)=\dfrac{f^{(n+1)}(\xi)}{(n+1)!}\prod_{i=0}^n(x-x_i)$; linear: $\le\tfrac18M_2h^2$; equal spacing: $\le\dfrac{M_{n+1}h^{n+1}}{4(n+1)}$
Divided differences and Newton's form§ 5.3 →
$p_n(x)=\sum_{k=0}^nf[x_0,\dots,x_k]\prod_{j\lt k}(x-x_j)$; $\ f[x_0,\dots,x_n]=\sum_i\dfrac{f(x_i)}{\prod_{j\ne i}(x_i-x_j)}=\dfrac{f^{(n)}(\xi)}{n!}$; $\ f[x,x]=f'(x)$
Newton's forward and backward formulas§ 5.3 →
$x=x_0+sh$: $p=\sum_k\binom sk\Delta^kf_0$; $\ x=x_n+sh$: $p=f_n+s\nabla f_n+\dfrac{s(s+1)}{2!}\nabla^2f_n+\cdots$; $\ f[x_0,\dots,x_k]=\dfrac{\Delta^kf_0}{k!\,h^k}$
Hermite and cubic spline interpolation§ 5.3 →
Hermite ($f,f'$ at $n+1$ nodes): degree $\le2n+1$, error $\dfrac{f^{(2n+2)}(\xi)}{(2n+2)!}\omega_{n+1}(x)^2$; spline moments $M_i=s''(x_i)$: $h_iM_{i-1}+2(h_i+h_{i+1})M_i+h_{i+1}M_{i+1}=6(d_{i+1}-d_i)$, natural: $M_0=M_n=0$
Chebyshev nodes on $[-1,1]$§ 5.3 →
$x_i=\cos\dfrac{(2i+1)\pi}{2n+2}$; $\ \max\big|\prod(x-x_i)\big|=2^{-n}$ (minimal); $\ \|f-p_n\|\le\dfrac{M_{n+1}}{2^n(n+1)!}$
Difference formulas and their errors§ 5.4 →
$f'(x_0)=\dfrac{f_1-f_0}{h}-\dfrac h2f''(\xi)$; $\ f'(x_0)=\dfrac{f_1-f_{-1}}{2h}-\dfrac{h^2}6f'''(\xi)$; $\ f''(x_0)=\dfrac{f_1-2f_0+f_{-1}}{h^2}-\dfrac{h^2}{12}f^{(4)}(\xi)$; optimal central step with data error $\varepsilon$: $h^*=(3\varepsilon/M_3)^{1/3}$
Basic rules (one panel)§ 5.4 →
Trapezoid $\tfrac h2(f_0+f_1)$, error $-\tfrac{h^3}{12}f''$; midpoint $hf(m)$, error $+\tfrac{h^3}{24}f''$; Simpson $\tfrac h3(f_0+4f_1+f_2)$, error $-\tfrac{h^5}{90}f^{(4)}$; 3/8 rule $\tfrac{3h}8(f_0+3f_1+3f_2+f_3)$, error $-\tfrac{3h^5}{80}f^{(4)}$ (error $=\int f-Q$)
Composite rules, $h=(b-a)/n$§ 5.4 →
$I-T_n=-\tfrac{(b-a)h^2}{12}f''(\xi)$; $\ I-M_n=\tfrac{(b-a)h^2}{24}f''(\xi)$; $\ I-S_n=-\tfrac{(b-a)h^4}{180}f^{(4)}(\xi)$ ($n$ even); $\ T_{2n}=\tfrac12(T_n+M_n)$, $S_{2n}=\tfrac13(T_n+2M_n)$
Gauss–Legendre on $[-1,1]$§ 5.4 →
$n=2$: $f(-\tfrac1{\sqrt3})+f(\tfrac1{\sqrt3})$, error $\tfrac{1}{135}f^{(4)}(\xi)$; $\ n=3$: $\tfrac59f(-\sqrt{\tfrac35})+\tfrac89f(0)+\tfrac59f(\sqrt{\tfrac35})$; exact for degree $2n-1$; on $[a,b]$: $x=\tfrac{a+b}2+\tfrac{b-a}2t$, factor $\tfrac{b-a}2$
Richardson and Romberg§ 5.4 →
$A(h)=A+ch^p+\cdots\Rightarrow\dfrac{2^pA(h/2)-A(h)}{2^p-1}$; Romberg $R(k,j)=R(k,j-1)+\dfrac{R(k,j-1)-R(k-1,j-1)}{4^j-1}$; $T(h)-I=\tfrac{h^2}{12}[f'(b)-f'(a)]-\tfrac{h^4}{720}[f'''(b)-f'''(a)]+\cdots$
Euler and modified Euler (Heun)§ 5.5 →
Euler: $y_{n+1}=y_n+hf(x_n,y_n)$; Heun: $y^*=y_n+hf_n$, $y_{n+1}=y_n+\tfrac h2\big[f_n+f(x_{n+1},y^*)\big]$; midpoint: $y_{n+1}=y_n+hf\big(x_n+\tfrac h2,y_n+\tfrac h2f_n\big)$
Classical Runge–Kutta (RK4)§ 5.5 →
$k_1=f(x_n,y_n)$, $k_2=f(x_n+\tfrac h2,y_n+\tfrac h2k_1)$, $k_3=f(x_n+\tfrac h2,y_n+\tfrac h2k_2)$, $k_4=f(x_n+h,y_n+hk_3)$; $\ y_{n+1}=y_n+\tfrac h6(k_1+2k_2+2k_3+k_4)$
Order and global error§ 5.5 →
$T_{n+1}=O(h^{p+1})$ (local) $\Rightarrow|e_n|\le\tfrac{Ch^p}{\Lambda}\big(e^{\Lambda(x_n-x_0)}-1\big)$ (global $O(h^p)$); Euler $p=1$, $T=\tfrac{h^2}2y''(\xi)$; Heun/midpoint $p=2$; RK4 $p=4$
Absolute stability on $y'=\lambda y$, $z=h\lambda$§ 5.5 →
Euler $R=1+z$, interval $(-2,0)$; RK2 $R=1+z+\tfrac{z^2}2$, $(-2,0)$; RK4 $R=\sum_{k\le4}\tfrac{z^k}{k!}$, $(-2.785,0)$; backward Euler $R=\tfrac1{1-z}$ (A-stable); trapezoidal $\tfrac{1+z/2}{1-z/2}$ (A-stable)
Picard and Taylor methods§ 5.5 →
$\varphi_{n+1}=y_0+\int_{x_0}^xf(t,\varphi_n)\,dt$, $|y-\varphi_n|\le\tfrac{ML^n|x-x_0|^{n+1}}{(n+1)!}$; Taylor: $y_{n+1}=\sum_{k\le p}\tfrac{h^k}{k!}y_n^{(k)}$, $y''=f_x+ff_y$
Chapter 6 · Calculus of Variations
Euler–Lagrange equation§ 6.1 →
Weak extremum of $\int_a^bF(x,y,y')\,dx$, fixed ends $\Rightarrow$ $F_y-\dfrac{d}{dx}F_{y'}=0$; expanded: $F_{y'y'}y''+F_{yy'}y'+F_{xy'}-F_y=0$
First integrals§ 6.1 →
No $y$: $F_{y'}=c$. No $x$ (Beltrami): $F-y'F_{y'}=c$. $F=F(y')$: lines. $F=M+Ny'$: $M_y=N_x$ (no ODE)
Classical extremals§ 6.1 →
$\sqrt{1+y'^2}$: lines; $\frac{\sqrt{1+y'^2}}{\sqrt y}$: cycloids $x=a(\theta-\sin\theta)$, $y=a(1-\cos\theta)$; $y\sqrt{1+y'^2}$: catenaries $y=c\cosh\frac{x-d}{c}$; $\frac{\sqrt{1+y'^2}}{y}$: circles centred on the $x$-axis
Euler–Poisson and Ostrogradsky§ 6.1 →
$F(x,y,y',y'')$: $F_y-(F_{y'})'+(F_{y''})''=0$ ($y,y'$ fixed at both ends). $\iint F(x,y,u,u_x,u_y)$: $F_u-\partial_xF_{u_x}-\partial_yF_{u_y}=0$
Free and variable end points§ 6.1 →
$y(b)$ free: $F_{y'}=0$ at $b$ (with $+\Gamma'(y(b))$ for a term $\Gamma(y(b))$). End on $y=\varphi(x)$: $F+(\varphi'-y')F_{y'}=0$; for $f\sqrt{1+y'^2}$: orthogonality $y'\varphi'=-1$
Isoperimetric problems§ 6.1 →
Extremise $\int F$ with $\int G=\ell$, $y$ not an extremal of $\int G$ $\Rightarrow$ Euler–Lagrange for $F+\lambda G$, $\lambda$ constant
Second variation, Legendre and Jacobi§ 6.2 →
$\delta^2J=\int(P\eta'^2+Q\eta^2)$, $P=F_{y'y'}$, $Q=F_{yy}-(F_{yy'})'$. Weak min $\Rightarrow P\ge0$. $P\gt0$ + no zero of $-(Pu')'+Qu=0$, $u(a)=0$, $u'(a)=1$ in $(a,b]$ $\Rightarrow$ strict weak min
Weierstrass excess function§ 6.2 →
$E=F(x,y,q)-F(x,y,p)-(q-p)F_{y'}(x,y,p)$; strong min $\Rightarrow E(x,y,y',q)\ge0$ for all $q$. $F$ jointly convex in $(y,y')$ $\Rightarrow$ extremals are global minima
Energy principle§ 6.2 →
$-(pu')'+qu=f$, $u(a)=u(b)=0$ ($p\gt0$, $q\ge0$) $\iff$ $u$ minimises $\int_a^b(pv'^2+qv^2-2fv)$; $\ I[u+\eta]=I[u]+\int(p\eta'^2+q\eta^2)$
Rayleigh–Ritz§ 6.2 →
$u_n=\sum c_j\varphi_j$, $Kc=F$, $K_{ij}=\int(p\varphi_i'\varphi_j'+q\varphi_i\varphi_j)$, $F_i=\int f\varphi_i$; $I[u_n]\ge I[u]$; $\det(K-\Lambda M)=0$ gives $\Lambda_k\ge\lambda_k$
Dirichlet principle§ 6.2 →
$-\Delta u=f$, $u=g$ on $\partial\Omega$ $\iff$ $u$ minimises $\iint(\tfrac12|\nabla v|^2-fv)$ with $v=g$ on $\partial\Omega$; Neumann $\partial_nu=h$ is natural, needs $\iint f+\oint h=0$
Chapter 7 · Linear Integral Equations
IVP → Volterra equation§ 7.1 →
$y''+q(x)y=g$, $y(a)=\alpha$, $y'(a)=\beta$ $\iff$ $\displaystyle y(x)=\alpha+\beta(x-a)+\int_a^x(x-t)\big[g(t)-q(t)y(t)\big]dt$; with a term $py'$: kernel $(x-t)[p'(t)-q(t)]-p(t)$, free term gains $p(a)\alpha(x-a)$
Cauchy's repeated-integration formula§ 7.1 →
$\displaystyle\int_a^x\!\!\cdots\!\int_a^{x_1}g=\frac{1}{(n-1)!}\int_a^x(x-t)^{n-1}g(t)\,dt$
BVP → Fredholm equation§ 7.1 →
$L[y]+\lambda ry=h$, $B_a[y]=B_b[y]=0$, $G$ = Green's function of $L$ $\iff$ $y=\int_a^bGh\,dt-\lambda\int_a^bG\,r\,y\,dt$; e.g. $y''+\lambda y=0$, $y(0)=y(1)=0$ $\iff$ $y=\lambda\int_0^1\min(x,t)(1-\max(x,t))\,y\,dt$
Separable kernels and the Fredholm alternative§ 7.1 →
$K=\sum a_i(x)b_i(t)$: $y=f+\lambda\sum c_ia_i$, $(I-\lambda A)c=F$, $A_{ij}=\int b_ia_j$, $F_i=\int b_if$. Characteristic numbers: $\det(I-\lambda A)=0$; there, solvable $\iff\int fz=0$ for all $z=\lambda\int K(t,x)z(t)\,dt$
Volterra equations: successive approximations§ 7.1 →
$y=\sum\lambda^n\varphi_n$, $\varphi_n=\int_a^xK\varphi_{n-1}$, $|\varphi_n|\le\|f\|\frac{M^n(x-a)^n}{n!}$: unique solution for every $\lambda$; no characteristic numbers. Convolution: $Y=\dfrac{F}{1-\lambda\hat k}$
Abel's equation§ 7.1 →
$\displaystyle\int_0^x\frac{y(t)\,dt}{\sqrt{x-t}}=f(x)\Rightarrow y=\frac1\pi\frac{d}{dx}\int_0^x\frac{f(t)\,dt}{\sqrt{x-t}}$
Iterated kernels and the resolvent§ 7.2 →
$K_{n+1}(x,t)=\int_a^bK(x,s)K_n(s,t)\,ds$, $\ R(x,t;\lambda)=\sum_{n\ge1}\lambda^{n-1}K_n$, $\ y=f+\lambda\int_a^bR(x,t;\lambda)f(t)\,dt$; $\ R=K+\lambda\int KR=K+\lambda\int RK$
Convergence and bounds§ 7.2 →
Neumann series converges if $|\lambda|M(b-a)\lt1$ or $|\lambda|B\lt1$, $B^2=\iint|K|^2$; true radius $=|\lambda_1|$. Every characteristic number: $|\lambda|\ge1/B$; multiplicity $\le|\lambda|^2B^2$
Resolvent of a rank-one kernel§ 7.2 →
$K=a(x)b(t)$, $\alpha=\int_a^bab$: $\ K_n=\alpha^{n-1}K$, $\ R=\dfrac{K}{1-\lambda\alpha}$; e.g. $xt$ on $[0,1]$: $\dfrac{3xt}{3-\lambda}$
Volterra resolvents (entire in $\lambda$)§ 7.2 →
$|K_n|\le\frac{M^n(x-t)^{n-1}}{(n-1)!}$. $K\equiv1$: $R=e^{\lambda(x-t)}$; $\ e^{x-t}$: $R=e^{(1+\lambda)(x-t)}$; $\ x-t$: $R=\frac{\sinh\sqrt\lambda(x-t)}{\sqrt\lambda}$; convolution: $\hat R=\frac{\hat k}{1-\lambda\hat k}$
Symmetric kernels§ 7.2 →
$\lambda_n$ real, $\varphi_n$ orthonormal; $\ \sum\frac1{\lambda_n^2}=\iint K^2$; Mercer ($K$ positive): $K=\sum\frac{\varphi_n(x)\varphi_n(t)}{\lambda_n}$, $\sum\frac1{\lambda_n}=\int K(x,x)\,dx$; $\ y=f+\lambda\sum\frac{(f,\varphi_n)}{\lambda_n-\lambda}\varphi_n$
Chapter 8 · Classical Mechanics
Lagrange's equations (holonomic, ideal constraints)§ 8.1 →
$\dfrac{d}{dt}\dfrac{\partial T}{\partial\dot q_j}-\dfrac{\partial T}{\partial q_j}=Q_j$; for $Q_j=-\partial V/\partial q_j$: $\dfrac{d}{dt}\dfrac{\partial L}{\partial\dot q_j}=\dfrac{\partial L}{\partial q_j}$, $L=T-V$
Degrees of freedom and generalised force§ 8.1 →
$n=3N-k$ ($k$ independent holonomic constraints); $\ Q_j=\sum_i\mathbf F_i\cdot\dfrac{\partial\mathbf r_i}{\partial q_j}$; equilibrium $\iff Q_j=0$
Kinetic energy in generalised coordinates§ 8.1 →
$T=\tfrac12\sum a_{jk}\dot q_j\dot q_k+\sum b_j\dot q_j+T_0$, $\ a_{jk}=\sum_i m_i\dfrac{\partial\mathbf r_i}{\partial q_j}\cdot\dfrac{\partial\mathbf r_i}{\partial q_k}$ positive definite; scleronomic $\Rightarrow T=T_2$
Cyclic coordinates and momenta§ 8.1 →
$p_j=\dfrac{\partial L}{\partial\dot q_j}$; $\ \dfrac{\partial L}{\partial q_j}=0\Rightarrow p_j=\text{const}$
Energy integral§ 8.1 →
$h=\sum\dot q_j\dfrac{\partial L}{\partial\dot q_j}-L=T_2-T_0+V$, $\ \dfrac{dh}{dt}=-\dfrac{\partial L}{\partial t}$; scleronomic, $V(q)$: $h=T+V$
Central force: Binet and effective potential§ 8.1 →
$\dfrac{d^2u}{d\theta^2}+u=-\dfrac{m}{\ell^2u^2}F\!\left(\tfrac1u\right)$, $u=\tfrac1r$; $\ V_{\text{eff}}=V+\dfrac{\ell^2}{2mr^2}$
Hamiltonian and canonical equations§ 8.2 →
$p_j=\dfrac{\partial L}{\partial\dot q_j}$, $H=\sum p_j\dot q_j-L$; $\ \dot q_j=\dfrac{\partial H}{\partial p_j}$, $\dot p_j=-\dfrac{\partial H}{\partial q_j}$, $\dfrac{dH}{dt}=\dfrac{\partial H}{\partial t}=-\dfrac{\partial L}{\partial t}$
Hamiltonian of a quadratic Lagrangian§ 8.2 →
$L=\tfrac12\dot q^{\mathsf T}A\dot q+b^{\mathsf T}\dot q+T_0-V\Rightarrow H=\tfrac12(p-b)^{\mathsf T}A^{-1}(p-b)-T_0+V$; scleronomic, $V(q)$: $H=T+V$
Hamilton's principle§ 8.2 →
$\delta\displaystyle\int_{t_1}^{t_2}L(q,\dot q,t)\,dt=0$ (fixed $q(t_1),q(t_2)$) $\iff$ Lagrange's equations
Poisson brackets§ 8.2 →
$[f,g]=\sum\big(f_{q_j}g_{p_j}-f_{p_j}g_{q_j}\big)$, $[q_j,p_k]=\delta_{jk}$; $\ \dot f=[f,H]+\partial_tf$
Canonical transformation test ($n=1$)§ 8.2 →
$(q,p)\mapsto(Q,P)$ canonical $\iff[Q,P]_{q,p}=Q_qP_p-Q_pP_q=1$
Least action and Routhian§ 8.2 →
$\delta\displaystyle\int\sqrt{2(E-V)}\,ds=0$ at fixed $E$; $\ R=L-\sum_c\beta_c\dot q_c$
Inertia tensor and moment about a line§ 8.3 →
$\mathcal I_O=\sum m(|\mathbf r|^2E-\mathbf r\mathbf r^T)$, entries $I_{xx}=\sum m(y^2+z^2)$, $-I_{xy}=-\sum mxy$; $\ I_{\mathbf n}=\mathbf n^T\mathcal I_O\mathbf n=I_{xx}\lambda^2+I_{yy}\mu^2+I_{zz}\nu^2-2I_{yz}\mu\nu-2I_{zx}\nu\lambda-2I_{xy}\lambda\mu$
Parallel and perpendicular axes§ 8.3 →
$\mathcal I_O=\mathcal I_G+M(|\mathbf a|^2E-\mathbf a\mathbf a^T)$, $\mathbf a=\overrightarrow{OG}$: $I_\ell=I_{\ell_G}+Mh^2$, $I^O_{xy}=I^G_{xy}+Ma_xa_y$; lamina in $z=0$: $I_{zz}=I_{xx}+I_{yy}$; always $A+B\ge C$
Standard moments (through $G$)§ 8.3 →
rod $\tfrac1{12}M\ell^2$; ring $Ma^2$ (axis); disc $\tfrac12Ma^2$ (axis), $\tfrac14Ma^2$ (diameter); sphere $\tfrac25Ma^2$; shell $\tfrac23Ma^2$; cuboid $\tfrac1{12}M(b^2+c^2)$; cube: $\tfrac16Ma^2$ about every line through $G$
$\mathbf L$, $T$, fixed axis§ 8.3 →
$\mathbf L_O=\mathcal I_O\vec\omega$, $T=\tfrac12\vec\omega^T\mathcal I_O\vec\omega$ ($O$ fixed); $T=\tfrac12Mv_G^2+\tfrac12\vec\omega^T\mathcal I_G\vec\omega$; compound pendulum $\tau=2\pi\sqrt{(k^2+h^2)/(gh)}$, $\ell_{\rm eq}=h+k^2/h$, least at $h=k$
Plane motion and rolling§ 8.3 →
$M\ddot{\mathbf r}_G=\mathbf F$, $I_G\ddot\theta=N_G$; $\mathbf v_P=\omega\mathbf e_z\times(\mathbf r_P-\mathbf r_I)$; rolling down slope $\alpha$: $\ddot x=\dfrac{g\sin\alpha}{1+k^2/a^2}$, needs $\mu\ge\dfrac{k^2\tan\alpha}{k^2+a^2}$
Euler's equations and free tops§ 8.3 →
$A\dot\omega_1-(B-C)\omega_2\omega_3=N_1$ (cyclic); $\mathbf N=\mathbf 0$: $2T$, $L^2$ constant; $A=B$: $\omega_3=n$, $\Omega=(C-A)n/A$ in the body, axis precesses at $L/A$ in space; rotation about the intermediate axis is unstable
Equilibrium and stability§ 8.4 →
Equilibrium $\iff\nabla V(q^0)=0$; strict minimum of $V$ $\Rightarrow$ stable (Lagrange–Dirichlet); Hessian with a negative eigenvalue $\Rightarrow$ unstable; rotating systems: use $U=V-T_0$
Small oscillations§ 8.4 →
$L_2=\tfrac12\dot\eta^T\mathsf T\dot\eta-\tfrac12\eta^T\mathsf V\eta$, $\ \mathsf T=a(q^0)$, $\mathsf V=$ Hessian of $V$; $\ \mathsf T\ddot\eta+\mathsf V\eta=0$; $\ \det(\mathsf V-\omega^2\mathsf T)=0$; $\ \sum\omega_k^2=\operatorname{tr}(\mathsf T^{-1}\mathsf V)$, $\prod\omega_k^2=\det\mathsf V/\det\mathsf T$
Normal modes and coordinates§ 8.4 →
$\mathsf V\mathbf a_k=\omega_k^2\mathsf T\mathbf a_k$, $\ \mathbf a_j^T\mathsf T\mathbf a_k=\delta_{jk}$; $\ \xi=P^T\mathsf T\eta$, $L_2=\tfrac12\sum(\dot\xi_k^2-\omega_k^2\xi_k^2)$; Rayleigh: $\omega_1^2=\min\dfrac{\eta^T\mathsf V\eta}{\eta^T\mathsf T\eta}$
Coupled and double pendulums§ 8.4 →
Spring-coupled pair: $\omega_1^2=\tfrac g\ell$ $(1,1)$, $\omega_2^2=\tfrac g\ell+\tfrac{2k}m$ $(1,-1)$, transfer time $\pi/(\omega_2-\omega_1)$; double pendulum (equal $m,\ell$): $\omega^2=(2\mp\sqrt2)\tfrac g\ell$, modes $(1,\pm\sqrt2)$
Molecule and chain§ 8.4 →
$m$–$M$–$m$: $\omega^2=0,\ \tfrac km,\ \tfrac km\big(1+\tfrac{2m}M\big)$; chain of $N$ between walls: $\omega_r=2\sqrt{k/m}\,\sin\dfrac{r\pi}{2(N+1)}$, modes $\sin\dfrac{jr\pi}{N+1}$