8.3 Rigid Bodies and Euler’s Dynamical Equations
Moments and products of inertia, the inertia tensor and its principal axes, motion about a fixed axis, rolling and the instantaneous centre, and Euler's dynamical equations — with the torque-free symmetric top and the instability of rotation about the intermediate axis.
A particle needs three numbers to say where it is; a rigid body needs six — three for the position of its centre of mass and three for its orientation. The translational half of the motion is easy: the centre of mass moves like a particle carrying the total mass and acted on by the total external force. The rotational half is where all the interesting behaviour lives. A spinning book flips over when you spin it about its middle axis but not about the other two; a spinning plate tossed into the air wobbles twice as fast as it spins; a rolling ball reaches the bottom of a slope before a rolling hoop of the same mass and radius; a wheel mounted slightly off its axis of symmetry shakes its bearings to pieces. All of these are explained by one object — the inertia tensor — and one set of equations — Euler's dynamical equations.
This section builds both from scratch. We define moments and products of inertia, assemble them into a symmetric matrix, and use the spectral theorem for real symmetric matrices (Unit 1, linear algebra) to find the principal axes. We then compute the angular momentum and kinetic energy of a rigid body, treat motion about a fixed axis (the compound pendulum) and plane motion (rolling, the instantaneous centre), and finally derive Euler's equations and use them for the torque-free symmetric top and the stability of rotation about principal axes. The Lagrangian machinery of §8.1 and §8.2 is used freely: once $T$ is known, a rigid body is just a system with a few generalised coordinates. CSIR NET asks about this material in a very recognisable way — a moment of inertia to compute quickly (cube about a diagonal, disc about a tangent, rolling acceleration), a statement about principal axes to accept or reject, or a property of Euler's equations; the counterexamples box below collects the traps.
Moments and Products of Inertia
Throughout, a rigid body is either a finite system of particles $m_i$ at positions $\mathbf r_i$, or a continuum with density $\rho$; sums $\sum_i m_i(\cdots)$ become integrals $\int\rho(\cdots)\,dV$ and we write both as $\sum m(\cdots)$. $M$ is the total mass and $G$ the centre of mass, $\mathbf r_G=\frac1M\sum m\,\mathbf r$. Vectors are bold ($\mathbf r$, $\mathbf v$, $\mathbf L$, $\mathbf N$ for torque), except the angular velocity, written $\vec\omega$ with components $\omega_1,\omega_2,\omega_3$.
A rigid body is a system of particles whose mutual distances do not change. A free rigid body has $6$ degrees of freedom; with one point fixed it has $3$; rotating about a fixed axis, $1$; in plane motion, $3$.
- The moment of inertia about a line $\ell$ is $I_\ell=\sum m\,d^2$, where $d$ is the distance of the mass element from $\ell$. The radius of gyration $k$ about $\ell$ is defined by $I_\ell=Mk^2$.
- With respect to axes $Oxyz$, the moments about the coordinate axes are $$I_{xx}=\sum m(y^2+z^2),\quad I_{yy}=\sum m(z^2+x^2),$$ $$I_{zz}=\sum m(x^2+y^2),$$ and the products of inertia are $I_{xy}=\sum m\,xy$, $\ I_{yz}=\sum m\,yz$, $\ I_{zx}=\sum m\,zx$.
Moments are always $\ge0$; products can have either sign, and depend on the origin and on the directions of the axes.
Proof. The distance $d$ of the point $\mathbf r$ from the line $\{t\mathbf n\}$ is the length of the component of $\mathbf r$ perpendicular to $\mathbf n$, so $d^2=|\mathbf r|^2-(\mathbf n\cdot\mathbf r)^2$. Since $|\mathbf n|=1$, $|\mathbf r|^2=\mathbf n^T(|\mathbf r|^2E)\mathbf n$ and $(\mathbf n\cdot\mathbf r)^2=\mathbf n^T(\mathbf r\mathbf r^T)\mathbf n$. Hence $d^2=\mathbf n^T(|\mathbf r|^2E-\mathbf r\mathbf r^T)\mathbf n$, and summing over the body gives $I_{\mathbf n}=\mathbf n^T\mathcal I_O\mathbf n$. Expanding the quadratic form with the entries of Definition 8.15 gives the second formula. ∎
So the nine numbers in $\mathcal I_O$ (six of them independent) determine the moment about every line through $O$. Two classical theorems move us between points and between dimensions.
Proof. Write the position of a mass element relative to $O$ as $\mathbf r=\mathbf a+\mathbf s$, where $\mathbf s$ is its position relative to $G$; by definition of $G$, $\sum m\,\mathbf s=\mathbf 0$. Then $|\mathbf r|^2E-\mathbf r\mathbf r^T$ equals $$\big(|\mathbf a|^2E-\mathbf a\mathbf a^T\big)+\big(|\mathbf s|^2E-\mathbf s\mathbf s^T\big)$$ $$+\big(2(\mathbf a\cdot\mathbf s)E-\mathbf a\mathbf s^T-\mathbf s\mathbf a^T\big).$$ The last bracket is linear in $\mathbf s$, so its sum over the body vanishes: every term contains the factor $\sum m\,\mathbf s=\mathbf 0$. Summing the other two brackets gives $M(|\mathbf a|^2E-\mathbf a\mathbf a^T)+\mathcal I_G$. For a line $\ell$ in direction $\mathbf n$ through $O$, Theorem 8.25 gives $I_\ell=I_{\ell_G}+M\big(|\mathbf a|^2-(\mathbf n\cdot\mathbf a)^2\big)$, and $|\mathbf a|^2-(\mathbf n\cdot\mathbf a)^2=h^2$ is the squared distance between $\ell$ and $\ell_G$. ∎
Proof. For any body, $I_{xx}+I_{yy}=\sum m(y^2+2z^2+x^2)$, which is $I_{zz}+2\sum mz^2$. For a lamina in $z=0$ every element has $z=0$, so the extra term vanishes, and so do $I_{zx}=\sum mzx$ and $I_{yz}=\sum myz$. ∎
Combining the three theorems with a few integrals gives every standard moment. The ones below are worth knowing by heart ($M$ = mass; axes through the centre unless stated).
The Inertia Tensor and Principal Axes
Because $\mathcal I_O$ is real and symmetric, we can rotate the axes so that it becomes diagonal. In those axes every product of inertia vanishes, and the formula of Theorem 8.25 collapses to a sum of three squares. These are the axes in which the dynamics is simplest.
A line through $O$ with unit direction $\mathbf n$ is a principal axis of inertia at $O$ if $\mathcal I_O\mathbf n=I\mathbf n$ for some scalar $I$ (an eigenvector of $\mathcal I_O$); $I=I_{\mathbf n}$ is the corresponding principal moment of inertia. Equivalently, the products of inertia $I_{xz}$, $I_{yz}$ vanish when the $z$-axis is taken along $\mathbf n$. The principal moments at $O$ are usually called $A,B,C$; a body with $A=B$ is a symmetric top about the third axis, and one with $A=B=C$ a spherical top.
When $\mathcal I_O$ is positive definite, the momental ellipsoid at $O$ is the surface $\mathbf x^T\mathcal I_O\mathbf x=1$. Along the direction $\mathbf n$ it lies at distance $1/\sqrt{I_{\mathbf n}}$ from $O$; its axes are the principal axes, with semi-axes $1/\sqrt A$, $1/\sqrt B$, $1/\sqrt C$.
- $\mathcal I_O$ is symmetric and positive semidefinite: $I_{\mathbf n}=\sum m\,|\mathbf n\times\mathbf r|^2\ge0$. It is positive definite unless all the mass lies on one line through $O$.
- At every point $O$ there are three mutually perpendicular principal axes; the principal moments $A,B,C$ are real and $\ge0$. If $A,B,C$ are distinct, the principal axes at $O$ are unique.
- In principal axes $I_{\mathbf n}=A\lambda^2+B\mu^2+C\nu^2$; the largest and smallest moments about lines through $O$ are the largest and smallest principal moments.
- (Triangle inequality) $A+B\ge C$, $B+C\ge A$, $C+A\ge B$, with $A+B=C$ if and only if the body is a lamina in the plane of the first two principal axes.
Proof. (1) $|\mathbf n\times\mathbf r|^2=|\mathbf r|^2-(\mathbf n\cdot\mathbf r)^2=d^2$, so $\mathbf n^T\mathcal I_O\mathbf n=\sum m\,|\mathbf n\times\mathbf r|^2\ge0$ for unit $\mathbf n$, hence for all $\mathbf n$ by scaling. If it vanishes for some $\mathbf n\ne\mathbf 0$, every mass element has $\mathbf n\times\mathbf r=\mathbf 0$, i.e. lies on the line $\mathbb R\mathbf n$.
(2) By the spectral theorem a real symmetric matrix has an orthonormal basis of eigenvectors with real eigenvalues; the eigenvalues are values $\mathbf n^T\mathcal I_O\mathbf n$ at unit eigenvectors, hence $\ge0$ by (1). If the eigenvalues are distinct, each eigenspace is a line, so the axes are unique.
(3) In an orthonormal eigenbasis $\mathcal I_O=\operatorname{diag}(A,B,C)$, so $\mathbf n^T\mathcal I_O\mathbf n=A\lambda^2+B\mu^2+C\nu^2$ with $\lambda^2+\mu^2+\nu^2=1$: a weighted average of $A,B,C$, which lies between the smallest and the largest.
(4) In principal axes, Theorem 8.27's identity reads $A+B=C+2\sum m\,z^2\ge C$, with equality iff every element has $z=0$. The other two inequalities are the same argument with the axes renamed. ∎
Symmetry usually tells us the principal axes before we compute anything.
- If the plane $z=0$ through $O$ is a plane of mirror symmetry of the mass distribution ($\rho(x,y,-z)=\rho(x,y,z)$), then $I_{zx}=I_{yz}=0$: the normal $Oz$ is a principal axis at $O$. In particular, for a lamina in a plane, the normal to the plane is principal at every point of it, and a line of symmetry of a lamina is a principal axis at each of its points.
- If $Oz$ is an axis of rotational symmetry of order $n\ge2$ (the body coincides with itself after a rotation by $2\pi/n$ about $Oz$), then $Oz$ is a principal axis at every point $O$ of it.
- If the order is $n\ge3$, moreover, the two transverse principal moments at $O$ are equal and every line through $O$ perpendicular to $Oz$ is a principal axis.
Proof. (1) The mass elements at $(x,y,z)$ and $(x,y,-z)$ pair off, and their contributions to $\sum m\,zx$ and $\sum m\,yz$ cancel. Then the third column of $\mathcal I_O$ is $(0,0,I_{zz})^T$, so $\mathbf e_z$ is an eigenvector. For a line of symmetry $\ell$ of a lamina, apply (1) to the plane through $\ell$ perpendicular to the lamina: its normal, which lies in the lamina perpendicular to $\ell$, is principal; $\ell$ itself is then principal because it is perpendicular to two principal axes (the normal to the lamina and the one just found), and the eigenvectors of a symmetric matrix can be taken orthogonal.
(2)–(3) Let $R$ be the rotation by $\alpha=2\pi/n$ about $Oz$. Moving each mass element $\mathbf r\mapsto R\mathbf r$ does not change the body, so $\mathcal I_O=\sum m(|R\mathbf r|^2E-R\mathbf r\mathbf r^TR^T)$, and this equals $R\,\mathcal I_OR^T$. Write $\mathcal I_O=\begin{pmatrix}P&\mathbf u\\\mathbf u^T&I_{zz}\end{pmatrix}$ with $P$ a symmetric $2\times2$ block and $\mathbf u=-(I_{zx},I_{yz})^T$, and let $R_2$ be the $2\times2$ rotation by $\alpha$. The relation $R\mathcal I_OR^T=\mathcal I_O$ says $R_2PR_2^T=P$ and $R_2\mathbf u=\mathbf u$. Since $0\lt\alpha\lt2\pi$, $R_2$ fixes no nonzero vector, so $\mathbf u=\mathbf 0$ and $Oz$ is principal. If $n\ge3$, then also $\alpha\ne\pi$, so $R_2$ maps no line through the origin to itself. But $R_2$ maps each eigenspace of $P$ onto an eigenspace of $P$ with the same eigenvalue (as $PR_2=R_2P$); if $P$ had two distinct eigenvalues, its two eigenlines would be fixed by $R_2$. Hence $P=\beta E_2$ for a scalar $\beta$: both transverse moments equal $\beta$, $I_{xy}=0$ for every pair of perpendicular transverse axes, and every transverse line is principal. ∎
So a square or an equilateral-triangle lamina has the same moment about every line in its plane through its centre, and a cube (three axes of order $4$ through $G$) has $\mathcal I_G$ a multiple of $E$, as Example 2 found by computation. A rectangle (order $2$) does not: its two in-plane moments $\tfrac1{12}Mb^2\ne\tfrac1{12}Ma^2$.
Angular Momentum, Kinetic Energy and Motion about a Fixed Axis
To do dynamics we need the velocity field of a moving rigid body. The key fact is that, at each instant, it is a translation plus a rotation about an axis — and the rotation does not depend on which point we use as reference.
Proof. Fix an orthonormal frame $\mathbf e_1(t),\mathbf e_2(t),\mathbf e_3(t)$ attached to the body at $Q$. A body point has $\mathbf r_P=\mathbf r_Q+\sum_iX_i\mathbf e_i$ with constant $X_i$ (rigidity). Differentiating $\mathbf e_i\cdot\mathbf e_j=\delta_{ij}$ gives $\dot{\mathbf e}_i\cdot\mathbf e_j=-\mathbf e_i\cdot\dot{\mathbf e}_j$, so the matrix $W_{ij}=\dot{\mathbf e}_i\cdot\mathbf e_j$ is antisymmetric. Put $$\vec\omega=W_{23}\mathbf e_1+W_{31}\mathbf e_2+W_{12}\mathbf e_3 ;$$ a direct check on each basis vector gives $\dot{\mathbf e}_i=\vec\omega\times\mathbf e_i$. For instance $\dot{\mathbf e}_1=W_{12}\mathbf e_2+W_{13}\mathbf e_3$, while $$\vec\omega\times\mathbf e_1=W_{31}\,\mathbf e_2\times\mathbf e_1+W_{12}\,\mathbf e_3\times\mathbf e_1,$$ which is $-W_{31}\mathbf e_3+W_{12}\mathbf e_2$, the same since $W_{13}=-W_{31}$. Hence $\mathbf v_P=\mathbf v_Q+\sum X_i\,\vec\omega\times\mathbf e_i$, that is, $\mathbf v_P=\mathbf v_Q+\vec\omega\times(\mathbf r_P-\mathbf r_Q)$. If $\vec\omega'$ also works then $(\vec\omega-\vec\omega')\times(\mathbf r_P-\mathbf r_Q)=\mathbf 0$ for all $P$, which forces $\vec\omega=\vec\omega'$ unless all the vectors $\mathbf r_P-\mathbf r_Q$ are parallel. Finally $\vec\omega$ is the same for every reference point: subtracting the formulas for references $Q$ and $Q'$ shows that $\vec\omega$ found from $Q$ also works from $Q'$. ∎
- If the point $O$ of the body is fixed, then $$\mathbf L_O=\mathcal I_O\,\vec\omega,$$ $$T=\tfrac12\,\vec\omega^T\mathcal I_O\,\vec\omega=\tfrac12\,\vec\omega\cdot\mathbf L_O=\tfrac12I_{\hat\omega}|\vec\omega|^2,$$ where $I_{\hat\omega}$ is the moment about the instantaneous axis. In principal axes, $\mathbf L_O=(A\omega_1,B\omega_2,C\omega_3)$ and $T=\tfrac12(A\omega_1^2+B\omega_2^2+C\omega_3^2)$.
- (König) For any motion, $$T=\tfrac12Mv_G^2+\tfrac12\,\vec\omega^T\mathcal I_G\,\vec\omega,$$ $$\mathbf L_O=M\,\mathbf r_G\times\mathbf v_G+\mathcal I_G\,\vec\omega$$ ($O$ a fixed origin): translation of $G$ plus rotation about $G$.
- $\mathbf L_O$ is parallel to $\vec\omega$ if and only if $\vec\omega$ is along a principal axis at $O$.
Proof. (1) With $O$ fixed as origin, $\mathbf v=\vec\omega\times\mathbf r$. By the vector triple product $$\mathbf r\times(\vec\omega\times\mathbf r)=|\mathbf r|^2\vec\omega-(\mathbf r\cdot\vec\omega)\mathbf r=(|\mathbf r|^2E-\mathbf r\mathbf r^T)\vec\omega ;$$ summing, $\mathbf L_O=\mathcal I_O\vec\omega$. Next $$|\vec\omega\times\mathbf r|^2=|\vec\omega|^2|\mathbf r|^2-(\vec\omega\cdot\mathbf r)^2=\vec\omega^T(|\mathbf r|^2E-\mathbf r\mathbf r^T)\vec\omega,$$ so $T=\tfrac12\sum m|\mathbf v|^2=\tfrac12\vec\omega^T\mathcal I_O\vec\omega=\tfrac12\vec\omega\cdot\mathbf L_O$, and with $\vec\omega=|\vec\omega|\hat\omega$ Theorem 8.25 gives $\tfrac12I_{\hat\omega}|\vec\omega|^2$.
(2) Write $\mathbf r=\mathbf r_G+\mathbf s$ with $\sum m\mathbf s=\mathbf 0$; by Theorem 8.30, $\mathbf v=\mathbf v_G+\vec\omega\times\mathbf s$. Then $$\begin{aligned}\sum m|\mathbf v|^2&=Mv_G^2+2\mathbf v_G\cdot\Big(\vec\omega\times\sum m\mathbf s\Big)\\&\quad+\sum m|\vec\omega\times\mathbf s|^2 ;\end{aligned}$$ the middle term is $0$ and the last is $\vec\omega^T\mathcal I_G\vec\omega$ as in (1). Similarly, expanding $\sum m(\mathbf r_G+\mathbf s)\times(\mathbf v_G+\vec\omega\times\mathbf s)$, the cross terms vanish for the same reason and what is left is $M\mathbf r_G\times\mathbf v_G+\sum m\,\mathbf s\times(\vec\omega\times\mathbf s)$, which is $M\mathbf r_G\times\mathbf v_G+\mathcal I_G\vec\omega$.
(3) $\mathcal I_O\vec\omega=c\,\vec\omega$ is exactly the eigenvector condition. ∎
Part (3) is behind every unbalanced wheel: if a body is made to turn about a fixed axis that is not principal, $\mathbf L$ is not along the axis, it is carried round with the body, and $\dot{\mathbf L}\ne\mathbf 0$ even at constant spin — the bearings must supply that torque (Example 6).
- $I\ddot\theta+Mgh\sin\theta=0$, with $I=M(k^2+h^2)$; small oscillations have period $\tau=2\pi\sqrt{\dfrac{k^2+h^2}{gh}}$, the period of a simple pendulum of length $\ell=h+\dfrac{k^2}{h}$ (the equivalent simple pendulum);
- the point on the line $OG$ at distance $\ell$ from $O$ (the centre of oscillation) has the property that suspension about a parallel axis through it gives the same period;
- the period is least, $2\pi\sqrt{2k/g}$, when $h=k$.
Proof. Every point moves on a circle about the axis with $\mathbf v=\dot\theta\,\mathbf e_z\times\mathbf r$, so $L_z=\sum m(xv_y-yv_x)$, which is $\dot\theta\sum m(x^2+y^2)=I\dot\theta$, and $T=\tfrac12\dot\theta^2\sum m(x^2+y^2)$. The $z$-component of $\dot{\mathbf L}_O=\mathbf N_O$ gives $I\ddot\theta=N_z$: the reaction of a smooth axis acts at points of the axis and has no moment about it. (Equivalently, $\theta$ is a single generalised coordinate and Lagrange's equations of §8.1 with $L=\tfrac12I\dot\theta^2-V(\theta)$ give $I\ddot\theta=-V'(\theta)$.)
(1) Gravity acts at $G$; with $\theta$ the angle between $OG$ and the downward vertical, $N_z=-Mgh\sin\theta$, and $I=M(k^2+h^2)$ by Steiner. For small $\theta$, $\ddot\theta+\tfrac{gh}{k^2+h^2}\theta=0$, so $\tau=2\pi\sqrt{(k^2+h^2)/(gh)}=2\pi\sqrt{\ell/g}$ with $\ell=h+k^2/h$. (2) The centre of oscillation is at distance $h'=\ell-h=k^2/h$ from $G$ on the other side, and $h'+k^2/h'=k^2/h+h=\ell$: the same equivalent length. (3) $h+k^2/h\ge2k$ by the AM–GM inequality, with equality iff $h=k$. ∎
Part (2) is the principle of Kater's reversible pendulum, which measured $g$ to about one part in $10^5$ for a century: one finds two knife-edges on opposite sides of $G$ giving equal periods, and then $\ell$ is just the distance between them (Exercise III.3).
Two-Dimensional Motion: Rolling and the Instantaneous Centre
A wheel rolling along a road, a ladder sliding down a wall, a cylinder rolling down a slope: in all of these every point moves parallel to one fixed plane. Then $\vec\omega$ is perpendicular to that plane, $\vec\omega=\dot\theta\,\mathbf e_z$, and the configuration is described by three numbers $(x_G,y_G,\theta)$.
- A rigid body is in plane motion if the velocities of all its points are parallel to a fixed plane (the $xy$-plane); then $\vec\omega=\dot\theta\,\mathbf e_z$, where $\theta$ is the angle turned by a line fixed in the body.
- An instantaneous centre (instantaneous centre of rotation) is a point $I$ of the plane, regarded as rigidly attached to the body, whose velocity is zero at the instant considered.
- A body rolls without slipping on a fixed surface if the material point of the body in contact with the surface has zero velocity. For a wheel of radius $a$ with centre $G$ on a straight line this is the constraint $\dot x_G=a\dot\phi$ ($\phi$ = angle turned, measured clockwise when the wheel moves in the $+x$ direction), which integrates to the holonomic constraint $x_G=a\phi+\text{const}$. (A ball rolling on a plane in three dimensions is the standard non-holonomic example of §8.1.)
Proof. By Theorem 8.30, $\mathbf v_P=\mathbf v_Q+\omega\mathbf e_z\times\mathbf d$ with $\mathbf d=\mathbf r_P-\mathbf r_Q$ in the plane. Setting $\mathbf v_P=\mathbf 0$ and taking $\mathbf e_z\times$ of both sides, using $\mathbf e_z\times(\mathbf e_z\times\mathbf d)=-\mathbf d$ for $\mathbf d\perp\mathbf e_z$, gives $\mathbf e_z\times\mathbf v_Q-\omega\mathbf d=\mathbf 0$, i.e. $\mathbf d=\mathbf e_z\times\mathbf v_Q/\omega$; conversely this $\mathbf d$ makes $\mathbf v_P=\mathbf 0$ (it is the unique solution of a nonsingular linear equation). Using $I$ as the reference point in Theorem 8.30 gives $\mathbf v_P=\omega\mathbf e_z\times(\mathbf r_P-\mathbf r_I)$, which is perpendicular to $\mathbf r_P-\mathbf r_I$ with length $|\omega|\,IP$. Then $T=\tfrac12\sum m\,\omega^2\,IP^2=\tfrac12I_I\omega^2$. If $\omega=0$, Theorem 8.30 gives $\mathbf v_P=\mathbf v_Q$ for all $P$. ∎
As the motion proceeds, $I$ traces a curve fixed in space (the space centrode) and a curve fixed in the body (the body centrode); the motion is exactly the rolling of the body centrode on the space centrode. For a wheel rolling on a road, the space centrode is the road and the body centrode is the rim.
Proof. The first equation is the law of motion of the centre of mass. For the second, let $\mathbf s=\mathbf r-\mathbf r_A$ for body points and $\mathbf H_A=\sum m\,\mathbf s\times\dot{\mathbf s}$. Since $A$ is a body point, $\dot{\mathbf s}=\omega\mathbf e_z\times\mathbf s$ and, as in Theorem 8.31, $\mathbf H_A=I_A\omega\,\mathbf e_z$ with $I_A$ constant. Differentiating, $\dot{\mathbf H}_A=\sum m\,\mathbf s\times\ddot{\mathbf s}=\sum m\,\mathbf s\times(\ddot{\mathbf r}-\mathbf a_A)$. Now $\sum m\,\mathbf s\times\ddot{\mathbf r}=\mathbf N_A$ (internal forces cancel), and $\sum m\,\mathbf s\times\mathbf a_A=M(\mathbf r_G-\mathbf r_A)\times\mathbf a_A$. Taking $z$-components gives the formula; the correction term vanishes in the three listed cases. For a round wheel of radius $a$ with $G$ at its centre, rolling on a line, the contact point $C$ has $$\mathbf a_C=\mathbf a_G+\ddot\theta\,\mathbf e_z\times(\mathbf r_C-\mathbf r_G)-\omega^2(\mathbf r_C-\mathbf r_G);$$ the rolling condition makes the first two terms cancel (both are horizontal, of size $a|\ddot\theta|$ and opposite), leaving $\mathbf a_C=\omega^2(\mathbf r_G-\mathbf r_C)$, directed along $CG$. ∎
Notice what the theorem does not say: the instantaneous centre has zero velocity but in general nonzero acceleration (for the rolling wheel, $a\omega^2$ towards the centre), and taking moments about it is wrong for, say, a wheel whose centre of mass is off-centre.
Euler's Dynamical Equations
For a body turning about a fixed point $O$ (or about $G$), the law $\dot{\mathbf L}=\mathbf N$ is simple in space but $\mathbf L=\mathcal I\vec\omega$ is simple in the body, where $\mathcal I$ is constant. Euler's idea was to write $\dot{\mathbf L}$ in components along principal axes moving with the body. The only tool needed is the rule for differentiating a vector given by its body components: if $\mathbf Q=\sum Q_i\mathbf e_i$ with $\mathbf e_i$ fixed in the body, then by Theorem 8.30 $\dot{\mathbf e}_i=\vec\omega\times\mathbf e_i$ and $$\frac{d\mathbf Q}{dt}=\sum_i\dot Q_i\,\mathbf e_i+\vec\omega\times\mathbf Q .$$ (For $\mathbf Q=\vec\omega$ the extra term is $\vec\omega\times\vec\omega=\mathbf 0$: the angular acceleration has the same components in both frames.)
Proof. In principal axes $\mathbf L=A\omega_1\mathbf e_1+B\omega_2\mathbf e_2+C\omega_3\mathbf e_3$ (Theorem 8.31), with $A,B,C$ constant because the axes move with the body. By the differentiation rule, $\dot{\mathbf L}=(A\dot\omega_1,B\dot\omega_2,C\dot\omega_3)+\vec\omega\times\mathbf L$ in body components, and $$\vec\omega\times\mathbf L=\begin{pmatrix}(C-B)\,\omega_2\omega_3\\(A-C)\,\omega_3\omega_1\\(B-A)\,\omega_1\omega_2\end{pmatrix}.$$ Equating $\dot{\mathbf L}$ with $\mathbf N$ component by component gives the three equations. The law $\dot{\mathbf L}=\mathbf N$ is valid about a fixed point and also about $G$ (even when $G$ accelerates), by Theorem 8.31(2). ∎
The equations are cyclic in $(1,2,3)\to(2,3,1)\to(3,1,2)$ together with $(A,B,C)$ — remember one and rotate the indices. They are nonlinear, which is why a free rigid body can do surprising things.
Proof. $\dot{\mathbf L}=\mathbf N=\mathbf 0$, so $\mathbf L$ is a constant vector and $L^2$ is constant. Multiply Euler's equations (with $N_i=0$) by $\omega_1,\omega_2,\omega_3$ and add: the left side becomes $\tfrac{d}{dt}T$ and the right side is $\omega_1\omega_2\omega_3$ times $(B-C)+(C-A)+(A-B)=0$, so $T$ is constant. (Multiplying instead by $A\omega_1,B\omega_2,C\omega_3$ gives $\tfrac{d}{dt}\tfrac{L^2}2=\omega_1\omega_2\omega_3\,S$ with $S=A(B-C)+B(C-A)+C(A-B)$, which is $0$ again.) Finally, dividing each equation by its moment and adding, $$\tfrac{d}{dt}\tfrac{|\vec\omega|^2}2=\omega_1\omega_2\omega_3\Big[\tfrac{B-C}A+\tfrac{C-A}B+\tfrac{A-B}C\Big],$$ and the bracket equals $\dfrac{(A-B)(A-C)(B-C)}{ABC}$, which is the stated expression; it is nonzero when $A,B,C$ are distinct. ∎
Proof. The third Euler equation reads $C\dot\omega_3=(A-B)\omega_1\omega_2=0$, so $\omega_3=n$. The first two become $A\dot\omega_1=(A-C)n\,\omega_2$ and $A\dot\omega_2=(C-A)n\,\omega_1$, i.e. $\dot\omega_1=-\Omega\omega_2$, $\dot\omega_2=\Omega\omega_1$. For $w=\omega_1+i\omega_2$ this is $\dot w=i\Omega w$, so $w=w(0)e^{i\Omega t}$. In space, $$\begin{aligned}\mathbf L&=A(\omega_1\mathbf e_1+\omega_2\mathbf e_2)+Cn\,\mathbf e_3\\&=A\vec\omega+(C-A)n\,\mathbf e_3,\end{aligned}$$ which shows that $\mathbf L,\vec\omega,\mathbf e_3$ are coplanar. Solving, $\vec\omega=\big(\mathbf L-(C-A)n\,\mathbf e_3\big)/A$, and since $\mathbf e_3$ is fixed in the body, $\dot{\mathbf e}_3=\vec\omega\times\mathbf e_3=\tfrac1A\mathbf L\times\mathbf e_3$: $\mathbf e_3$ rotates about the constant vector $\mathbf L$ with angular velocity $\mathbf L/A$. ∎
The Earth is very nearly an oblate symmetric top with $(C-A)/A\approx1/305$. If its rotation axis is slightly off the figure axis, Theorem 8.37 predicts that the pole wanders around the figure axis with period $2\pi/\Omega\approx305$ days (the Euler period). The observed wobble — the Chandler wobble, discovered in 1891 — has a period of about $433$ days; the difference is due to the Earth not being rigid.
Proof. With $\omega_2=\omega_3=0$ all products in Euler's equations vanish, so $\vec\omega=\Omega\mathbf e_1$ is a solution. Linearisation: put $\omega_1=\Omega+\xi_1$ and keep first-order terms in $\xi_1,\omega_2,\omega_3$: $A\dot\xi_1=0$, $B\dot\omega_2=(C-A)\Omega\,\omega_3$, $C\dot\omega_3=(A-B)\Omega\,\omega_2$. Differentiating the second and substituting the third gives the displayed equation; the coefficient is negative iff $(C-A)(A-B)\lt0$, i.e. iff $A$ is not between $B$ and $C$.
The linear analysis decides instability (a positive coefficient gives an exponentially growing solution, and the full nonlinear system follows it while the perturbation is small). For nonlinear stability in the extreme cases use the integrals of Theorem 8.36: the combination $$L^2-2AT=B(B-A)\,\omega_2^2+C(C-A)\,\omega_3^2$$ is conserved. If $A$ is the smallest moment, both coefficients are positive, so $\omega_2^2$ and $\omega_3^2$ stay bounded by a constant multiple of their (small) initial values for all time; if $A$ is the largest, both are negative and the same holds. Conservation of $T$ then keeps $\omega_1$ close to $\Omega$. If $A$ is intermediate the coefficients have opposite signs and the conserved quantity confines nothing. ∎
Spin a book (or a phone, or a tennis racket) about each of its three symmetry axes in turn: about the longest and the shortest it spins cleanly; about the middle axis it flips over every turn or so. Interactive 2 integrates Euler's equations so you can watch this — and see the two integrals stay constant while $\vec\omega$ swings wildly.
Proof (sketch). The total rotation is the sum of three rotations, so $\vec\omega=\dot\phi\,\mathbf z+\dot\theta\,\mathbf n+\dot\psi\,\mathbf e_3$. In body components, $\mathbf e_3=(0,0,1)$; the line of nodes, at angle $\psi$ behind $\mathbf e_1$ in the $\mathbf e_1\mathbf e_2$-plane, is $\mathbf n=(\cos\psi,-\sin\psi,0)$; and $\mathbf z$, at angle $\theta$ from $\mathbf e_3$ in the plane perpendicular to $\mathbf n$, is $(\sin\theta\sin\psi,\sin\theta\cos\psi,\cos\theta)$. Adding gives the three formulas. Then $\omega_1^2+\omega_2^2=\dot\theta^2+\dot\phi^2\sin^2\theta$, so $T=\tfrac A2(\omega_1^2+\omega_2^2)+\tfrac C2\omega_3^2$ takes the stated form; $V=Mgl\cos\theta$ is the height of $G$ times $Mg$. Neither $\phi$ nor $\psi$ appears in $L$, so their momenta are conserved (§8.1, cyclic coordinates): $p_\psi=\partial L/\partial\dot\psi=C\omega_3$, $p_\phi=\partial L/\partial\dot\phi=A\dot\phi\sin^2\theta+C\omega_3\cos\theta$. Lagrange's equation for $\theta$ is $$A\ddot\theta=A\dot\phi^2\sin\theta\cos\theta-C\omega_3\dot\phi\sin\theta+Mgl\sin\theta ;$$ with $\ddot\theta=0$ and $\sin\theta\ne0$ this is the quadratic for $\dot\phi$, whose smaller root is $\dfrac{C\omega_3-\sqrt{C^2\omega_3^2-4AMgl\cos\theta}}{2A\cos\theta}\approx\dfrac{Mgl}{C\omega_3}$ by expanding the square root. ∎
- Two sign conventions. Here $I_{xy}=\sum m\,xy$ and the tensor entry is $-I_{xy}$; many books call the entry itself the product of inertia. Formulas such as $I_{\mathbf n}=\dots-2I_{xy}\lambda\mu$ change sign accordingly.
- Perpendicular axes: laminae only. For a cube, $I_{xx}+I_{yy}=\tfrac13Ma^2\ne\tfrac16Ma^2=I_{zz}$ at $G$. In general $I_{xx}+I_{yy}\ge I_{zz}$.
- Steiner only through $G$. $I_\ell=I_{\ell_G}+Mh^2$ needs $\ell_G$ through the centre of mass. Between two other parallel axes, $I_{\ell_1}-I_{\ell_2}=M(h_1^2-h_2^2)$, not $Md_{12}^2$: for a rod, the moment about an end ($\tfrac13M\ell^2$) is not the moment about the other end plus $M\ell^2$.
- Products of inertia also shift. $I_{xy}^O=I_{xy}^G+Ma_xa_y$: the sides of a rectangle are principal at $G$ but not at a corner.
- Symmetry of order 2 is not enough for equal transverse moments (rectangle); order $\ge3$ is (square, equilateral triangle, regular hexagon).
- $\mathcal I_G=\beta E$ does not make $\mathcal I_O$ isotropic for $O\ne G$ — and, conversely, $\mathcal I_O$ can be isotropic at a point other than $G$: for a uniform disc, at the point of its axis at distance $\tfrac a2$ from the centre every moment equals $\tfrac12Ma^2$.
- $\mathbf L$ is not parallel to $\vec\omega$ unless $\vec\omega$ is principal (the plate spinning about a diagonal, Example 6). And $T=\tfrac12I\omega^2$ needs $I$ about the instantaneous axis.
- The instantaneous centre is not at rest in the sense of acceleration. The contact point of a wheel rolling steadily has acceleration $a\omega^2$ towards the centre. Taking moments about the instantaneous centre is legitimate only when its acceleration points towards $G$ (Theorem 8.34).
- Rolling friction is static friction. It is not $\mu R$ in general (it is whatever rolling requires), and it does no work; a sliding-and-spinning ball loses energy until it rolls (Exercise III.7).
- Euler's equations need principal axes fixed in the body. In non-principal axes, or with components along space axes, extra terms appear. Note $\dot{\vec\omega}$ is the same in both frames, but $\dot{\mathbf L}$ is not.
- Torque-free does not mean constant $\vec\omega$. $\mathbf L$ is fixed; $\vec\omega$ moves in the body and in space, and even $|\vec\omega|$ varies when $A,B,C$ are distinct (Theorem 8.36).
- "Stable" about the largest and smallest axes is a statement about a rigid body. With internal dissipation (flexible parts, fuel sloshing), energy decreases at fixed $L$ and only rotation about the axis of largest moment survives: the first American satellite, Explorer 1 (1958), designed to spin about its long axis of least moment, went into a tumble soon after launch.
Exercises
- 1 For a uniform rod of length $\ell$, find by integration the moments about the perpendicular through its centre and through an end, and the moment about a line through its centre making angle $\alpha$ with the rod.
- 2 For a uniform disc of radius $a$, find the moment about a diameter, about a tangent in its plane, and about the line through a point of the rim perpendicular to the disc.
- 3 Find the moment of inertia of a uniform solid sphere about a diameter by slicing it into discs.
- 4 A uniform solid right circular cone has height $h$ and base radius $c$. Find its moment about its axis, about a line through the apex perpendicular to the axis, and about the parallel line through $G$ (which is at distance $\tfrac34h$ from the apex).
- 5 A uniform right-angled triangular lamina has legs $a$ (along $Ox$) and $b$ (along $Oy$) meeting at $O$. Find $I_{xx},I_{yy},I_{xy}$ at $O$, and then at the centroid (axes parallel to the legs).
- 6 For a uniform cube of edge $a$, find the moment about an edge and about a line joining the midpoints of two opposite edges of a face (through the centre of that face, parallel to the other edges).
- 7 Use Theorem 8.29 to show that the moment of inertia of a uniform cube about any line through its centre is $\tfrac16Ma^2$, without computing the tensor.
- 8 Prove that $I_{xx}+I_{yy}\ge I_{zz}$ for any body and any axes (not only principal ones), with equality iff the body lies in the plane $z=0$.
- 9 For a lamina in the plane $z=0$, prove that $I_{xy}^2\le\sum m\,x^2\cdot\sum m\,y^2=I_{yy}I_{xx}$. When does equality hold?
- 1 For a uniform rectangular plate $a\times b$ with axes along the sides at a corner $O$, find the in-plane principal directions and principal moments at $O$. Evaluate for $a=2$, $b=1$.
- 2 For the isosceles right triangle ($a=b$ in I.5), find the principal axes and moments at the centroid in the plane of the lamina.
- 3 Show that a uniform equilateral-triangle lamina of side $s$ has moment $\tfrac1{24}Ms^2$ about every line through its centroid in its plane.
- 4 The L-shaped lamina of Interactive 1 consists of the unit squares filling $[0,3]\times[0,1]$ and $[0,1]\times[1,3]$, mass $M$ in total. Find $I_{xx},I_{yy},I_{xy}$ at the outer corner $O=(0,0)$, the principal axes and moments in the plane at $O$, and explain why the line $y=x$ is principal both at $O$ and at the centroid.
- 5 A uniform cube turns about a fixed edge with angular velocity $\omega$. Find $\mathbf L$ about a corner on that edge, the angle between $\mathbf L$ and $\vec\omega$, and $T$.
- 6 Use König's theorem to find the kinetic energy of a disc and of a ring, each of mass $M$, rolling without slipping with centre speed $v$.
- 7 Prove that if two principal moments at $O$ are equal, $A=B$, then every line through $O$ in the plane of the corresponding principal axes is principal.
- 8 Prove that if a body has two non-parallel axes of rotational symmetry of order $\ge3$ through $G$, then $\mathcal I_G=\beta E$. Deduce that a uniform regular tetrahedron is a spherical top at its centre.
- 1 A uniform rod of length $\ell$ swings about a horizontal axis through a point at distance $h$ from its centre. Find the period of small oscillations as a function of $h$, and its minimum value.
- 2 A uniform disc of radius $a$ swings in its own vertical plane about a horizontal axis through a point of its rim. Find the length of the equivalent simple pendulum.
- 3 (Kater.) A compound pendulum has the same period about two parallel axes on opposite sides of $G$, at distances $h_1\ne h_2$ from $G$. Prove that $h_1h_2=k^2$ and that the equivalent length is $h_1+h_2$.
- 4 A light string wound round a uniform pulley (a disc of mass $4\ \text{kg}$, radius $0.1\ \text{m}$, free to turn about its fixed horizontal axis) carries a $2\ \text{kg}$ mass hanging freely. Find the acceleration of the mass and the tension.
- 5 A thin cylindrical shell and a solid sphere roll from rest down a $30^\circ$ incline, $2\ \text{m}$ along the slope. Find the speed of each at the bottom.
- 6 Find the least coefficient of friction that lets a uniform solid cylinder roll without slipping down an incline of angle $\alpha$. What happens if $\mu$ is smaller?
- 7 A billiard ball (uniform sphere, radius $a$) is set sliding with speed $v_0$ and no spin on a table with coefficient of friction $\mu$. Find when it starts to roll, its speed then, the distance travelled while slipping, and the fraction of kinetic energy lost.
- 8 A wheel of radius $a$ rolls along a line with centre speed $v$. Find the velocity of the topmost point and of the point level with the centre at the front, and show that every point of the rim moves at right angles to the line joining it to the contact point.
- 9 A uniform ladder of length $2a$ stands on a smooth floor against a smooth vertical wall, making angle $\theta_0$ with the wall, and is released from rest. (a) Locate the instantaneous centre and show that its space and body centrodes are circles. (b) Prove that the ladder leaves the wall when its top has fallen to two-thirds of its initial height.
- 1 Derive the third Euler equation directly from $\dot{\mathbf L}=\mathbf N$, writing $\dot{\mathbf L}$ in body components.
- 2 A uniform cube turns with constant angular velocity $\omega$ about a fixed edge (no gravity). Find the torque about a corner on that edge that the bearings must provide.
- 3 A torque-free symmetric body has $A=B=2\ \text{kg m}^2$, $C=3\ \text{kg m}^2$ and $\omega_3=4\ \text{rad/s}$. Find the rate and period at which $\vec\omega$ circles the symmetry axis in the body frame.
- 4 Take $(C-A)/A=1/305$ for the Earth and a spin of one turn per day. Predict the period of the free wobble of the rotation axis about the figure axis.
- 5 Using only Euler's equations, prove that for $\mathbf N=\mathbf 0$ the functions $L^2-2AT$ and $L^2-2CT$ are constant, and express them in terms of $\omega_2,\omega_3$ and $\omega_1,\omega_2$ respectively.
- 6 A box with $A:B:C=1:2:3$ spins freely at $\Omega$ about each principal axis in turn. For each axis find the linearised wobble frequency or growth rate.
- 7 From the Eulerian-angle formulas, verify $\omega_1^2+\omega_2^2=\dot\theta^2+\dot\phi^2\sin^2\theta$, and show that a heavy symmetric top precessing steadily at a fixed tilt $\theta\in(0,\tfrac\pi2)$ has a precession rate $\dot\phi$ satisfying $A\dot\phi^2\cos\theta-C\omega_3\dot\phi+Mgl=0$; deduce that steady precession is possible only if $C^2\omega_3^2\ge4AMgl\cos\theta$.
E1. CSIR Part B GATE MA A uniform solid cube has mass $M$ and edge $a$. Its moment of inertia about a diagonal of one of its faces is (A) $\tfrac16Ma^2$; (B) $\tfrac5{12}Ma^2$; (C) $\tfrac23Ma^2$; (D) $\tfrac7{12}Ma^2$.
E2. CSIR Part B A thin uniform spherical shell rolls without slipping down a rough plane inclined at $30^\circ$ to the horizontal. The acceleration of its centre is (A) $\tfrac g2$; (B) $\tfrac{3g}{10}$; (C) $\tfrac{5g}{14}$; (D) $\tfrac g3$.
E3. CSIR Part B A rigid body moves with no external torque. Its principal moments at $G$ are $A=B=4$ and $C=6$ (in $\text{kg m}^2$), and $\omega_3=10\ \text{rad/s}$. In the body frame the angular velocity vector revolves around the symmetry axis with period (A) $\tfrac{2\pi}5\ \text{s}$; (B) $\tfrac{2\pi}{15}\ \text{s}$; (C) $\tfrac{4\pi}5\ \text{s}$; (D) $\tfrac\pi5\ \text{s}$.
E4. CSIR Part C GATE MA Which of the following statements about inertia tensors are true? (One or more options may be correct.) (A) The principal moments $A,B,C$ at any point satisfy $A+B\ge C$. (B) If a lamina lies in the $xy$-plane and $I_{xy}=0$ at its centroid, then the $x$-, $y$- and $z$-axes through the centroid are principal axes. (C) If the three principal moments at $G$ are equal, the body has the same moment of inertia about every line through every point. (D) If a body has the same moment of inertia about every line through a point $O$, then $O=G$.
E5. CSIR Part C Consider the torque-free motion of a rigid body about its centre of mass, with principal moments $A,B,C$. Which of the following are true? (A) If $A\lt B\lt C$, steady rotation about the principal axis with moment $B$ is unstable. (B) For every body, $A\omega_1^2+B\omega_2^2+C\omega_3^2$ is constant. (C) For every body, $\omega_1^2+\omega_2^2+\omega_3^2$ is constant. (D) If $A=B$, then both $\omega_3$ and $\omega_1^2+\omega_2^2$ are constant.
E6. CSIR Part C A uniform disc of mass $M$ and radius $a$ rolls without slipping along a straight line with constant centre speed $v$. Which of the following are true? (A) The material point of the disc in contact with the line has zero velocity and zero acceleration. (B) The topmost point of the disc has speed $2v$. (C) The kinetic energy is $\tfrac34Mv^2$. (D) The material point in contact has acceleration of magnitude $v^2/a$ directed towards the centre.
Suggested plan — 3 lectures × 60 min.
- Inertia. Definitions 8.14–8.15 and Theorem 8.25 (the one-line proof via $d^2=|\mathbf r|^2-(\mathbf n\cdot\mathbf r)^2$); Steiner and the perpendicular axis theorem; Examples 1–2 on the board (the cube's tensor is a beautiful use of $\tfrac{11}{12}E-\tfrac14J$). Principal axes, Theorems 8.28–8.29, Example 3. Project Interactive 1: start with the rectangle at its centre (product identically zero along the sides), then move $O$ to the corner and ask the class where the principal axis will be before pressing Snap.
- Kinematics and plane motion. Theorem 8.30 (spend time on why $\vec\omega$ does not depend on the reference point), Theorem 8.31 and König; fixed axis and the compound pendulum with Example 4 (bring a metre rule and time it). Plane motion, instantaneous centre, the warning in Theorem 8.34, Example 5 and the ladder (Exercise III.9) — the "two-thirds" answer always surprises.
- Euler. Derive Euler's equations from the rotating-frame rule; Example 6 (dynamic balancing) and Example 7 (the wobbling plate — toss a paper plate). Theorems 8.36–8.38 with Interactive 2 and a live demonstration with a book held shut by a rubber band. Ten minutes on Eulerian angles and gyroscopic precession (Theorem 8.39). Finish with E4 and E5.
Discussion questions. (1) A wheel is statically balanced (its centre of mass is on the axle) but still vibrates at speed. What is wrong, and how do the clip-on weights fix it? (2) Can a body be a spherical top about a point other than its centre of mass? (E4(D) says yes — for which shapes?) (3) Why does a satellite with flexible antennas end up spinning about its axis of largest moment, although Theorem 8.38 says the smallest is also stable?
Misconceptions to watch for. Using the perpendicular axis theorem for solids; applying Steiner between two axes neither of which passes through $G$; assuming $\mathbf L\parallel\vec\omega$; taking the friction on a rolling body to be $\mu R$; believing the instantaneous centre has zero acceleration and taking moments about it carelessly; writing Euler's equations in non-principal or space-fixed axes; thinking that torque-free motion means constant $\vec\omega$.
Suggested homework. I(2, 4, 5, 7), II(1, 4, 5, 8), III(1, 3, 5, 7, 9), IV(2, 3, 6); Exam corner E2, E4, E6.
Research question. How long does a phone spun about its middle axis take to flip over — and how does that time depend on how carefully it is spun?
Why it matters. In 1985 the cosmonaut Vladimir Dzhanibekov, on the Salyut 7 space station, saw a wing nut spinning off its bolt flip over again and again at regular intervals. The same instability governs the attitude of satellites and spacecraft (which is why their designers care about which axis has the largest moment), twisting somersaults in diving and gymnastics, and the tumbling of asteroids. It is a clean example of a homoclinic orbit: the motion lingers near an unstable equilibrium and then makes a quick excursion.
What to do.
- Measure a phone (or a book) and model it as a uniform cuboid; compute $A\lt B\lt C$ about its centre.
- Integrate Euler's equations (RK4 with a small step, or
scipy.integrate.solve_ivp) from $\vec\omega(0)=\Omega\mathbf e_2+\varepsilon(\mathbf e_1+\mathbf e_3)$ for $\varepsilon=10^{-1},10^{-2},10^{-3},10^{-4}$, and record the time between successive sign changes of $\omega_2$ (successive flips). - Plot the flip time against $\log(1/\varepsilon)$. Predict the slope from the growth rate $s$ of Theorem 8.38 (think about how long a trajectory needs to leave, and later to approach, a saddle point at distance $\varepsilon$) and compare.
- Check that $2T$ and $L^2$ are conserved to many digits; then deliberately make the step too large and watch them drift.
- Optional: film a real toss with a slow-motion camera and compare the observed number of turns per flip with your simulation.
Tools: Python with NumPy/SciPy (or Interactive 2 for a first look); a ruler; optionally a phone camera.
Hand in: A one-page note with your $A,B,C$, the table of flip times, the fitted slope against the predicted one, and a sentence on why a real phone does not flip exactly periodically.
Going further. The torque-free motion is solvable exactly in Jacobi elliptic functions; derive the exact flip period as a complete elliptic integral and recover the logarithmic law as the modulus tends to $1$. Read about the "tennis racket theorem" and the geometric phase it produces.
🧪 Self-check quiz
Q1. The relation $I_{zz}=I_{xx}+I_{yy}$ holds
Q2. For a body turning about a fixed point $O$, the angular momentum $\mathbf L_O$ is parallel to $\vec\omega$
Q3. Which is the first of Euler's dynamical equations (principal axes, moments $A,B,C$)?
Q4. A free rigid body with distinct principal moments $A\lt B\lt C$ is set spinning about a principal axis. Steady rotation is stable about
Q5. A uniform solid cylinder rolls without slipping down a plane inclined at $\alpha$. The acceleration of its axis is