Rotational Kinetic Energy

On 31 January 1958, Explorer 1 rode into orbit as America's first satellite — a slim, pencil-shaped cylinder spinning proudly about its long axis at 750 rpm. Within hours, ground stations noticed something alarming in the radio signal: the spin was decaying, and the satellite was settling into a lazy end-over-end tumble. Nothing had touched it. No torque acted on it. Its angular momentum was exactly conserved — and yet its rotational kinetic energy had collapsed to less than one percent of the launch value, bled away by four flexing whip antennas. The satellite had discovered, before the theorists in the program did, that a spinning body's energy is not one number but a quadratic form: the same angular momentum can be carried expensively about one axis and cheaply about another, and dissipation always finds the cheap way. That quadratic form is this article's subject. For a rigid body, the kinetic energy splits cleanly into a translational piece $\tfrac12 MV^2$ and a rotational piece $T_{\mathrm{rot}} = \tfrac12\,\boldsymbol\omega^{\mathsf T}\mathbf I\,\boldsymbol\omega$ , where the inertia tensor $\mathbf I$ acts as a price list: it tells you, direction by direction, what a unit of spin costs. Diagonalize it — work in principal axes — and the price list becomes three numbers $I_1, I_2, I_3$ . Mastering this one expression buys you the energetics of everything that rolls, spins, or tumbles: why a sphere outraces a hoop downhill, how a flywheel plant stores megawatt-hours, why every modern spinning spacecraft is designed as a "major-axis spinner" — and the geometric heart of free rotation, where surfaces of constant energy and constant angular momentum intersect to steer $\boldsymbol\omega$ itself.

Key takeaways

The kinetic energy of a rigid body splits into a translational part $\tfrac12 MV^2$ for the center of mass and a rotational part that depends only on the angular velocity and how the mass is distributed — but the split is exact only about the center of mass or a fixed point.
Rotational kinetic energy is the quadratic form $T_{\mathrm{rot}}=\tfrac12\,\boldsymbol\omega^{\mathsf T}\mathbf{I}\,\boldsymbol\omega$ , with $\mathbf{I}$ the inertia tensor about the chosen point and $\boldsymbol\omega$ the angular velocity.
In principal axes the cross terms vanish and $T_{\mathrm{rot}}=\tfrac12(I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2)$ — a clean weighted sum over the three principal moments.
Because $\mathbf L=\mathbf I\boldsymbol\omega$ , one also has $T_{\mathrm{rot}}=\tfrac12\,\boldsymbol\omega\cdot\mathbf L$ and $T_{\mathrm{rot}}=\sum_i L_i^2/2I_i$ — at fixed angular momentum, spin about the largest moment is the cheapest state, which is where dissipating bodies end up.
Surfaces of constant $T_{\mathrm{rot}}$ in $\boldsymbol\omega$ -space form an ellipsoid; for torque-free motion both $T_{\mathrm{rot}}$ and $\mathbf L$ are conserved, and their two ellipsoids intersect in the curve (the polhode) that the tip of $\boldsymbol\omega$ must trace — Poinsot's construction.
Key idea: Rotational kinetic energy is a quadratic form $T_{\mathrm{rot}}=\tfrac12\,\boldsymbol\omega^{\mathsf T}\mathbf I\,\boldsymbol\omega$ in the angular velocity, with the inertia tensor as the metric; in principal axes it becomes $\tfrac12\sum_i I_i\omega_i^2$ .

Where this lives today

The quadratic form is quietly load-bearing across modern engineering. Grid-scale flywheel plants — hundreds of carbon-fiber rotors in vacuum pits, of the kind that buffer frequency on the New York and Beijing grids — are businesses built on $T=\tfrac12 I\omega^2$ plus one materials fact: the stored energy per unit mass of a rim scales as $\sigma/\rho$ , the specific strength of the rim material, because hoop stress grows as $\rho\,\omega^2R^2$ exactly as fast as the energy density. That single scaling is why flywheel development is materials science, not motor engineering, and why composite rims displaced steel decades ago. In spaceflight, the dissipation theorem is written into design rules. Every purely spin-stabilized spacecraft since the Explorer 1 anomaly is either a major-axis spinner (spun about $I_{\max}$ , the energy floor, passively safe) or carries active nutation damping and control authority to fight the energy slide. The same argument, run in reverse on natural bodies, tells astronomers that a tumbling asteroid must have been disturbed recently: internal friction relaxes rotation toward the principal-axis energy floor on a computable timescale, so an object still tumbling — like 4179 Toutatis, or interstellar visitor 1I/'Oumuamua, whose excited spin state is still argued over in the 2020s — is wearing a clock. And attitude engineers reading out reaction-wheel budgets on observatories like JWST are doing nothing more exotic than moving terms between $\tfrac12MV^2$ , $\tfrac12\boldsymbol\omega^{\mathsf T}\mathbf I\boldsymbol\omega$ , and the wheels' own spin energies — König's split, kept honest to the milliradian.

The physical picture

Think of the inertia tensor as a machine that, when fed a spin direction, quotes you an energy price. Spin a broomstick about its long axis — the mass hugs the axis, the moment of inertia is tiny, and a brisk rotation costs almost nothing. Spin the same broomstick end-over-end at the same angular speed and every gram of it must be hauled around a wide circle: the price is orders of magnitude higher. Same body, same $|\boldsymbol\omega|$ , wildly different energy. The full directional version of that price list is $T_{\mathrm{rot}} = \tfrac12\,\boldsymbol\omega^{\mathsf T}\mathbf I\,\boldsymbol\omega$ : a quadratic form whose "metric" is the inertia tensor built in the sibling article on the moment of inertia tensor.

Figure

One body, two prices — the same angular speed costs far more end-over-end than about the long axis

Geometrically, fix the energy and ask which angular velocities you can afford: the answer is an ellipsoid in $\boldsymbol\omega$ -space with semi-axes $\sqrt{2T/I_i}$ . A long thin body has a long ellipsoid axis along its symmetry direction — lots of $\omega$ is permitted there, because that rotation is cheap. This surface is the star of Poinsot's picture of free rotation, where it rolls without slipping on a fixed plane. The relation $T_{\mathrm{rot}}=\tfrac12\,\boldsymbol\omega\cdot\mathbf L$ says the energy is half the overlap between spin and angular momentum. When $\boldsymbol\omega$ and $\mathbf L$ are parallel — rotation about a principal axis — the overlap is maximal for given magnitudes and the motion is steady. When they are misaligned, $\boldsymbol\omega$ wanders around $\mathbf L$ even though $\mathbf L$ itself is nailed to the inertial sky. And because energy is a scalar built from a tensor and a vector, it is invariant under any rotation of the coordinate axes: compute it in whatever frame is convenient — almost always principal axes — and the number is the same in every frame. One more picture, the one Explorer 1 taught the hard way. Freeze the angular momentum (no external torque can change it) and write the energy as $T=\sum_i L_i^2/2I_i$ . The same $|\mathbf L|$ costs the most energy when carried by the smallest moment and the least when carried by the largest. Give a real, slightly flexible body any way to dissipate — sloshing fuel, flexing antennas — and it will slide down that energy ladder at fixed $\mathbf L$ , ending up spinning about its axis of maximum moment of inertia. Pencil-shaped spinners are doomed; frisbee-shaped spinners are forever.

Figure

Torque-free rotation lives on the intersection of two ellipsoids — constant energy and constant angular momentum

From the particle sum to the quadratic form

Everything starts from the definition that kinetic energy is additive over the body's constituents. For any collection of masses,

T=\frac12\sum_i m_i\,\mathbf v_i\cdot\mathbf v_i \qquad (1)

where: $m_i$ — mass of the $i$ -th constituent; $\mathbf v_i$ — its velocity in the inertial frame. Nothing rigid yet; this holds for a gas, a galaxy, or a gearbox. Rigidity enters through the kinematic statement that every point of the body shares one translational velocity and one angular velocity. If $\mathbf V$ is the velocity of a chosen reference point and $\mathbf r_i'$ the position of particle $i$ relative to that point,

\mathbf v_i=\mathbf V+\boldsymbol\omega\times\mathbf r_i' \qquad (2)

where: $\boldsymbol\omega$ — the body's angular velocity vector, the same for all particles (a body-wide property, as established for rigid bodies); $\boldsymbol\omega\times\mathbf r_i'$ — the sweep velocity of particle $i$ around the reference point. This one constraint is what collapses $3N$ velocity degrees of freedom to six. Substituting (2) into (1) and choosing the reference point wisely produces the master decomposition of rigid-body energetics — the König split:

T=\tfrac12 M V^2+T_{\mathrm{rot}},\qquad T_{\mathrm{rot}}=\tfrac12\sum_i m_i\,(\boldsymbol\omega\times\mathbf r_i')^2 \qquad (3)

where: $M$ — total mass; $V$ — speed of the center of mass. The translational term is a point particle's energy; the rotational term knows nothing about where the body is going, only how it is turning. The derivation below shows why the cross term dies — and why it dies only for the center of mass or a fixed point.

Derivation AThe Koenig split and the inertia tensor as the energy kernel

We are deriving the two central results of this article in one sweep: the clean decomposition (3) of a rigid body's kinetic energy into translation plus rotation, and the tensor form (4) of the rotational piece. The strategy is direct: insert the rigid velocity field (2) into the particle sum (1), expand the square, and watch which terms survive; then reduce the surviving rotational sum with Lagrange's identity for the squared cross product. The step to watch is the middle (cross) term of the expansion — it vanishes only because the first mass moment about the center of mass is zero, which is precisely what "center of mass" means. Nothing here assumes principal axes, small rotations, or any symmetry: the results are exact for any rigid body.

Step 1. Insert the velocity field into the kinetic energy and expand the square:

T=\tfrac12\sum_i m_i\big(\mathbf V+\boldsymbol\omega\times\mathbf r_i'\big)^2 = \tfrac12 M V^2+\mathbf V\cdot\Big(\boldsymbol\omega\times\sum_i m_i\mathbf r_i'\Big)+\tfrac12\sum_i m_i(\boldsymbol\omega\times\mathbf r_i')^2

Why the middle term could be pulled out of the sum: $\mathbf V$ and $\boldsymbol\omega$ are body-wide constants, so only $\sum_i m_i \mathbf r_i'$ — the first moment of the mass distribution — survives inside the bracket.

Step 2. Choose the reference point to be the center of mass. Then by definition $\sum_i m_i\mathbf r_i'=0$ , the cross term dies, and the split (3) is exact. Why the fixed-point case also works: if instead the reference point is fixed in space, $\mathbf V=0$ kills the cross term (and the translational term) regardless of where the mass sits — which is why a pendulum pivot is as good a reference as the center of mass.

Step 3. Reduce the rotational sum with Lagrange's identity $(\mathbf a\times\mathbf b)^2=a^2b^2-(\mathbf a\cdot\mathbf b)^2$ :

(\boldsymbol\omega\times\mathbf r_i')^2=\omega^2 r_i'^2-(\boldsymbol\omega\cdot\mathbf r_i')^2 = \omega_a\omega_b\big(\delta_{ab}\,r_i'^2-x_{ia}'x_{ib}'\big)

Why the index form: both terms are quadratic in $\boldsymbol\omega$ , and writing $\omega^2 = \omega_a\omega_b\delta_{ab}$ and $(\boldsymbol\omega\cdot\mathbf r')^2 = \omega_a\omega_b\,x_a'x_b'$ factors the $\boldsymbol\omega$ -dependence completely out of the geometry.

Step 4. Sum over the constituents. The mass-weighted bracket is exactly the inertia tensor about the center of mass:

T_{\mathrm{rot}}=\tfrac12\,\omega_a\,\omega_b\sum_i m_i\big(\delta_{ab}\,r_i'^2-x_{ia}'x_{ib}'\big)=\tfrac12\,\omega_a I_{ab}\,\omega_b, \qquad I_{ab}=\sum_i m_i\big(\delta_{ab}r_i'^2-x_{ia}'x_{ib}'\big)

Why this is satisfying: the same tensor that the angular-momentum route defines through $\mathbf L = \mathbf I\boldsymbol\omega$ appears here as the kernel of the energy quadratic form — one object, two jobs. For a continuous body, replace the sum by $I_{ab}=\int\rho(\mathbf r)(\delta_{ab}r^2-x_ax_b)\,\mathrm d^3r$ . *Note: contracting $\tfrac12\boldsymbol\omega\cdot(\mathbf I\boldsymbol\omega)$ reproduces the same expression, confirming $T_{\mathrm{rot}}=\tfrac12\boldsymbol\omega\cdot\mathbf L$ .*

The result of that derivation deserves its own display, because it is the working formula of the whole subject:

T_{\mathrm{rot}}=\frac12\,\omega_a\,I_{ab}\,\omega_b=\frac12\,\boldsymbol\omega^{\mathsf T}\mathbf I\,\boldsymbol\omega \qquad (4)

where: $I_{ab}$ — components of the inertia tensor about the reference point; $\omega_a$ — Cartesian components of $\boldsymbol\omega$ , with summation over repeated indices. Because $\mathbf I$ is symmetric and positive-(semi)definite, $T_{\mathrm{rot}}\ge 0$ always, vanishing only when $\boldsymbol\omega=0$ (or points along an axis of zero moment, as for an idealized line of masses). Diagonalizing $\mathbf I$ — the subject of the sibling article on principal axes and principal moments — kills the off-diagonal (product-of-inertia) terms and produces the form everyone memorizes:

T_{\mathrm{rot}}=\tfrac12\left(I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2\right) \qquad (5)

where: $I_1,I_2,I_3$ — the principal moments (eigenvalues of $\mathbf I$ ); $\omega_1,\omega_2,\omega_3$ — components of $\boldsymbol\omega$ along the principal axes. Every term is separately non-negative: there are no energy "discounts" from cross terms once you are in the right frame. For a sphere all three moments are equal and (5) collapses to $\tfrac12 I\omega^2$ for any axis. Contracting (4) with $\mathbf L=\mathbf I\boldsymbol\omega$ instead gives two equivalent expressions, each the natural tool for a different question:

T_{\mathrm{rot}}=\tfrac12\,\boldsymbol\omega\cdot\mathbf L \qquad (6)

where: $\mathbf L$ — angular momentum about the same reference point. Energy is half the projection of $\mathbf L$ on the spin. When bearings force rotation about a non-principal axis, $\mathbf L$ tilts away from $\boldsymbol\omega$ and only its component along $\boldsymbol\omega$ carries energy — the transverse part is what makes the bearings shake.

T_{\mathrm{rot}}=\frac{1}{2}\left(\frac{L_1^2}{I_1}+\frac{L_2^2}{I_2}+\frac{L_3^2}{I_3}\right) \qquad (7)

where: $L_i=I_i\omega_i$ — angular momentum components along principal axes. This is the form to reach for when $\mathbf L$ is the conserved quantity: at fixed $|\mathbf L|$ , the energy is smallest when all the angular momentum is carried by the largest moment ( $T=L^2/2I_{\max}$ ) and largest on the smallest moment ( $T=L^2/2I_{\min}$ ). That inequality is the entire physics of the Explorer 1 flip, derived properly below. Finally, two expressions that connect this article outward. Fixing $T_{\mathrm{rot}}$ in (5) defines the ellipsoid of affordable angular velocities,

\frac{\omega_1^2}{2T/I_1}+\frac{\omega_2^2}{2T/I_2}+\frac{\omega_3^2}{2T/I_3}=1 \qquad (8)

where: $T$ — the fixed value of $T_{\mathrm{rot}}$ ; the semi-axes are $\sqrt{2T/I_i}$ , longest along the smallest moment. Up to normalization this is the inertia ellipsoid studied in its own sibling article. And substituting the Euler-angle resolution of $\boldsymbol\omega$ (from the sibling article on angular velocity in terms of Euler angles) turns (5) into

T_{\mathrm{rot}}=\tfrac12 I_1\big(\dot\phi\sin\theta\sin\psi+\dot\theta\cos\psi\big)^2+\tfrac12 I_2\big(\dot\phi\sin\theta\cos\psi-\dot\theta\sin\psi\big)^2+\tfrac12 I_3\big(\dot\phi\cos\theta+\dot\psi\big)^2 \qquad (9)

where: $\phi,\theta,\psi$ — the Euler angles (precession, tilt, spin) and $\dot\phi,\dot\theta,\dot\psi$ their rates. This is the kinetic part of every top Lagrangian: for the symmetric case $I_1=I_2$ it simplifies to $\tfrac12 I_1(\dot\theta^2+\dot\phi^2\sin^2\theta)+\tfrac12 I_3(\dot\phi\cos\theta+\dot\psi)^2$ , the launching point for the heavy symmetric top, its precession, and its nutation.

Which form answers which question

you want	use	notes
energy of a body rolling or flying	König split, eq. (3)	valid about CM; add $\tfrac12MV^2$
energy from $\boldsymbol\omega$ in any axes	quadratic form, eq. (4)	needs the full tensor with products of inertia
energy from $\boldsymbol\omega$ in principal axes	eq. (5)	the default for tops and tumbling
energy when $\mathbf L$ is what you know	eq. (7)	conserved- $\mathbf L$ problems, dissipation arguments
quick scalar check, fixed axis	$T=\tfrac12 I\omega^2$	only for a fixed or principal axis
building a Lagrangian for a top	Euler-angle form, eq. (9)	cyclic $\phi,\psi$ give conserved momenta

Free rotation: what conservation of T and L allows

For a body floating free of torques — a tossed book, a tumbling asteroid, a satellite — the two quantities this article and its angular-momentum sibling have built are both constants of the motion. Their conservation is not an extra assumption; it follows from the torque-free Euler equations themselves, and jointly they imprison the angular velocity on a one-dimensional curve.

Go deeper · Derivation B · 2 minConservation of rotational energy and the Poinsot constraint

Derivation BConservation of rotational energy and the Poinsot constraint

The goal is to show that torque-free rotation conserves both $T_{\mathrm{rot}}$ and $L^2$ when each is written in body-frame components, and to extract the geometric consequence: the tip of $\boldsymbol\omega$ is confined to the intersection of two fixed ellipsoids — the polhode. The strategy is to differentiate each candidate invariant along the flow of the torque-free Euler equations $I_i\dot\omega_i=(I_j-I_k)\omega_j\omega_k$ (indices cyclic) and watch the cyclic structure annihilate the sum. The subtle point to appreciate: $\mathbf L$ is constant in the inertial frame, but its body-frame components $L_i$ rotate — only the combination $L^2 = \sum_i L_i^2$ survives as a body-frame constant, because rotations preserve lengths.

Step 1. Differentiate the energy $T=\tfrac12\sum_i I_i\omega_i^2$ along the motion:

\frac{\mathrm dT}{\mathrm dt}=\sum_i I_i\,\omega_i\,\dot\omega_i = \omega_1\omega_2\omega_3\big[(I_2-I_3)+(I_3-I_1)+(I_1-I_2)\big]=0

Why the bracket telescopes to zero: substituting each torque-free Euler equation $I_i\dot\omega_i=(I_j-I_k)\omega_j\omega_k$ makes every term proportional to the same product $\omega_1\omega_2\omega_3$ , and the moment differences sum cyclically to zero — energy conservation is hard-wired into the antisymmetric structure of the equations.

Step 2. Differentiate $L^2=\sum_i I_i^2\omega_i^2$ the same way:

\frac{\mathrm d(L^2)}{\mathrm dt}=2\sum_i I_i^2\,\omega_i\,\dot\omega_i = 2\,\omega_1\omega_2\omega_3\big[I_1(I_2-I_3)+I_2(I_3-I_1)+I_3(I_1-I_2)\big]=0

Why this bracket also vanishes: expand it — $I_1I_2-I_1I_3+I_2I_3-I_1I_2+I_1I_3-I_2I_3=0$ identically. The body-frame components of $\mathbf L$ move, but their squared sum cannot.

Step 3. Read the two invariants as surfaces in $\boldsymbol\omega$ -space:

\sum_i I_i\,\omega_i^2=2T \quad\text{(energy ellipsoid)},\qquad \sum_i I_i^2\,\omega_i^2=L^2 \quad\text{(momentum ellipsoid)}

where: both are ellipsoids fixed in the body frame, with semi-axes $\sqrt{2T/I_i}$ and $L/I_i$ respectively — the momentum ellipsoid is the more eccentric of the two, since its axes involve $I_i$ squared.

Step 4. Intersect them. A point $\boldsymbol\omega(t)$ obeying both constraints for all time must move along the intersection curve of the two ellipsoids — the polhode. For the intersection to be nonempty at all, the constants must satisfy $2TI_{\min}\le L^2\le 2TI_{\max}$ , with equality pinning $\boldsymbol\omega$ to a principal axis. Why this already contains the famous stability result: near the largest and smallest principal axes the polhodes are small closed loops (stable wobble), while through the intermediate axis they degenerate into a separatrix — the geometry behind the tennis-racket tumble, developed fully in the article on Euler's equations of motion. *Note: setting $I_1=I_2=I_3$ makes the two ellipsoids concentric spheres, which are compatible only if $L^2 = 2TI$ — and indeed a spherical top has $\boldsymbol\omega$ fixed, tracing a degenerate point polhode.*

The payoff of carrying both invariants is Poinsot's rolling picture: the energy ellipsoid, glued to the body, rolls without slipping on a plane fixed in space (the invariable plane, perpendicular to $\mathbf L$ ), the contact point tracing the polhode on the ellipsoid and the herpolhode on the plane. No integral needs to be evaluated — the geometry is the solution. The quantitative body-frame solution, with its body cone and space cone, is the business of the free-symmetric-top article.

Dissipation at fixed L: why spinning bodies flatten their spin

Real bodies are never perfectly rigid. Fuel sloshes, antennas flex, rock creeps. Internal friction converts mechanical energy to heat — but it exerts no external torque, so $\mathbf L$ stays put while $T$ leaks away. Equation (7) then becomes a ratchet with an inevitable end state.

Go deeper · Derivation C · 2 minEnergy dissipation at fixed angular momentum selects the major axis

Derivation CEnergy dissipation at fixed angular momentum selects the major axis

We want the end state of a slightly non-rigid, freely spinning body: the configuration that minimizes rotational kinetic energy at fixed angular momentum magnitude $L$ . The strategy is elementary but decisive — write the energy in the $\mathbf L$ -based form (7), treat the $L_i$ as the free variables subject to $\sum_i L_i^2 = L^2$ , and compare the energy of angular momentum "parked" on each principal axis. This three-line argument, not any detailed friction model, is what doomed Explorer 1 and now governs every spin-stabilized spacecraft design; watch how it never needs to know how the energy is dissipated, only that it monotonically decreases.

Step 1. Write the energy with $\mathbf L$ as the variable, ordering the moments $I_1\le I_2\le I_3$ :

T=\frac{L_1^2}{2I_1}+\frac{L_2^2}{2I_2}+\frac{L_3^2}{2I_3},\qquad L_1^2+L_2^2+L_3^2=L^2

Step 2. Bound the energy between the extreme axes. Since $1/2I_3\le 1/2I_i\le 1/2I_1$ for every $i$ , multiplying the constraint through gives

\frac{L^2}{2I_3}\;\le\; T\;\le\;\frac{L^2}{2I_1}

Why the bounds are attained only on principal axes: the lower bound requires every unit of $L^2$ to sit in the $1/2I_3$ slot, i.e. $\mathbf L$ entirely along the largest-moment axis; the upper bound likewise pins $\mathbf L$ to the smallest-moment axis.

Step 3. Let dissipation act. Internal friction reduces $T$ monotonically but cannot touch $L$ ; the state slides down the inequality chain until it hits the floor $T_{\min}=L^2/2I_3$ , where nothing more can be extracted. The body ends up spinning about its axis of maximum moment of inertia — a flat spin. Why minor-axis spin is doomed even though it is a valid steady rotation of a perfectly rigid body: for a rigid body the spin about $I_1$ sits at an energy maximum at fixed $L$ ; the tiniest flexibility opens a downhill path, so the configuration is only as stable as the body is rigid.

Step 4. Quantify the drop for a pencil-shaped body ( $I_1\ll I_3$ ): flipping from minor-axis spin to the flat spin releases the fraction

\frac{T_{\mathrm{initial}}-T_{\mathrm{final}}}{T_{\mathrm{initial}}}=1-\frac{I_1}{I_3}

which for Explorer 1 ( $I_1/I_3\approx 0.009$ ) means over 99% of the launch spin energy went into heating four whip antennas within hours. The end state matches the observed behavior of Explorer 1 (1958) and the "major-axis rule" of spacecraft dynamics; the bound structure agrees with the polhode geometry of the previous derivation — flat spin is the polhode of least energy at fixed L.

Limiting cases

Case	Result	Meaning
fixed axis	$T_{\mathrm{rot}}=\tfrac12 I\omega^2$	scalar moment $I=\int r_\perp^2\,\mathrm dm$ about that axis; the elementary formula, exact whenever the axis direction is enforced
principal axes	$T_{\mathrm{rot}}=\tfrac12(I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2)$	cross terms vanish; the cleanest general form
spherical top	$T_{\mathrm{rot}}=\tfrac12 I\,\vert\boldsymbol\omega\vert^2$	$I_1=I_2=I_3$ ; energy is isotropic in spin direction
symmetric top	$T_{\mathrm{rot}}=\tfrac12 I_1(\omega_1^2+\omega_2^2)+\tfrac12 I_3\omega_3^2$	two equal moments; the workhorse of precession theory
pure translation	$\boldsymbol\omega=0\Rightarrow T=\tfrac12MV^2$	the rotational term switches off entirely
relativistic check	$T_{\mathrm{rot}}\ll Mc^2$ required	Newtonian form is the leading term when peripheral speeds obey $\omega R\ll c$

Common trapThe formula T equals one half I omega squared works about any point

It does not — and this is the most common energy error in rigid-body problems. The clean split $T=\tfrac12MV^2+T_{\mathrm{rot}}$ holds only when the reference point is the center of mass (first mass moment vanishes) or a fixed point ( $\mathbf V=0$ ). About any other point a cross term $M\,\mathbf V\cdot(\boldsymbol\omega\times\mathbf r_{\mathrm{cm}})$ survives and silently corrupts the count. A second trap hides in the scalar shortcut: $T_{\mathrm{rot}}=\tfrac12 I\omega^2$ with one number $I$ is legitimate only for rotation about a fixed axis or a principal axis. For a skewed spin the products of inertia contribute, $\mathbf L$ is not parallel to $\boldsymbol\omega$ , and only the full quadratic form (4) — or equivalently $\tfrac12\boldsymbol\omega\cdot\mathbf L$ — gives the right energy.

Worked examples

1 · The downhill race (confidence builder)

A solid sphere, a solid cylinder, and a hoop, each of mass $M$ and radius $R$ , roll without slipping from rest down a ramp of vertical drop $h=1.2\ \mathrm m$ . Find each finishing speed and the fraction of energy locked in rotation.

Step 1. Rolling without slipping ties spin to travel: $\omega=v/R$ . With the moment about the symmetry axis written $I=\beta MR^2$ , the König split (3) gives total energy $T=\tfrac12Mv^2+\tfrac12(\beta MR^2)(v/R)^2=\tfrac12Mv^2(1+\beta)$ .

Step 2. Energy conservation down the drop: $Mgh=\tfrac12Mv^2(1+\beta)$ , so $v=\sqrt{2gh/(1+\beta)}$ — mass and radius both cancel; only the shape factor $\beta$ matters.

Step 3. Insert $\beta$ : sphere $\tfrac25$ , cylinder $\tfrac12$ , hoop $1$ . With $2gh=2\times9.81\times1.2=23.54\ \mathrm{m^2/s^2}$ : sphere $v=\sqrt{23.54/1.4}=\boxed{4.10\ \mathrm{m/s}}$ , cylinder $v=\sqrt{23.54/1.5}=\boxed{3.96\ \mathrm{m/s}}$ , hoop $v=\sqrt{23.54/2}=\boxed{3.43\ \mathrm{m/s}}$ .

Step 4. The rotational share of the energy is $\beta/(1+\beta)$ : 28.6% for the sphere, 33.3% for the cylinder, 50% for the hoop. The hoop finishes last not because it is heavier or slower to start but because half of every joule it gains must be spent on spinning — the price list of eq. (5) applied at the checkout.

2 · Explorer 1's expensive lesson (exam level)

Model Explorer 1 as a uniform cylinder of mass $m=14\ \mathrm{kg}$ , length $\ell=2.05\ \mathrm m$ , radius $R=0.08\ \mathrm m$ , spun about its long axis at $750\ \mathrm{rpm}$ . Flexible antennas dissipate energy at fixed angular momentum. Find the initial energy, the final state, the final tumble rate, and the energy dissipated.

Step 1. Principal moments of a uniform cylinder: about the long axis $I_3=\tfrac12 mR^2=\tfrac12(14)(0.08)^2=0.0448\ \mathrm{kg\,m^2}$ ; about a transverse axis through the CM $I_\perp=\tfrac{1}{12}m(3R^2+\ell^2)=\tfrac{14}{12}(3\times0.0064+4.2025)=4.93\ \mathrm{kg\,m^2}$ . The ratio $I_3/I_\perp=0.0091$ — a pencil.

Step 2. Initial spin: $\omega_3=750\times2\pi/60=78.5\ \mathrm{rad/s}$ . Angular momentum $L=I_3\omega_3=0.0448\times78.5=3.52\ \mathrm{kg\,m^2/s}$ ; initial energy $T_0=\tfrac12 I_3\omega_3^2=\tfrac12(0.0448)(78.5)^2=\boxed{138\ \mathrm J}$ .

Step 3. Dissipation at fixed $L$ drives the body to minimum energy, $T=L^2/2I_{\max}$ — a flat, end-over-end spin about the transverse axis. Final energy: $T_f=L^2/2I_\perp=(3.52)^2/(2\times4.93)=\boxed{1.26\ \mathrm J}$ .

Step 4. Final tumble rate: $\omega_\perp=L/I_\perp=3.52/4.93=0.714\ \mathrm{rad/s}$ , i.e. 6.8 rpm — one stately somersault every 8.8 seconds, the slow end-over-end tumble the ground stations observed.

Step 5. Energy dissipated: $\Delta T=T_0-T_f=138-1.26\approx\boxed{137\ \mathrm J}$ , a fraction $1-I_3/I_\perp=99.1\%$ of the launch value, delivered as heat to the antennas. The satellite kept every last unit of its angular momentum and lost essentially all of its rotational energy — the sharpest possible demonstration that $T$ and $\mathbf L$ are independent bookkeeping columns.

Origins · 1 minLeonhard Euler and how this was discovered

Leonhard Euler

1707–1783 · St. Petersburg & Berlin academies

Euler owns rigid-body rotation the way Newton owns gravity. In his 1765 treatise Theoria motus corporum solidorum seu rigidorum he introduced the moments of inertia, proved that every body possesses three mutually perpendicular principal axes, and wrote the rotational energy as the sum $\tfrac12\sum I_i\omega_i^2$ that still opens every course on the subject. The equations of motion bearing his name — and the angles, and the theorems — came from a man who produced roughly half his life's output after going almost completely blind in 1766, dictating calculations to assistants in St. Petersburg. When he showed that torque-free rotation conserves both the energy quadratic and the momentum quadratic, he handed the 19th century the two ellipsoids from which Poinsot would later build the entire geometry of free rotation without solving a single differential equation.

History

1686

— Leibniz champions "vis viva"

mv^2

as the true measure of motion, seeding the concept that becomes kinetic energy.

1751

— Samuel König states the decomposition of a system's kinetic energy into center-of-mass motion plus motion about the center of mass.

1765

— Euler's Theoria motus corporum solidorum founds rigid-body dynamics: principal axes, moments of inertia, and the rotational energy quadratic.

1834

— Poinsot publishes his rolling-ellipsoid construction, turning free rotation into pure geometry of the energy and momentum surfaces.

1958

— Explorer 1 flips from minor-axis spin to flat tumble within hours, teaching spacecraft engineers the energy-sink rule at fixed angular momentum.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 9 (Angular Momentum, Part II) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 5 · Landau & Lifshitz, Mechanics, Vol. 1, §32 · Taylor, Classical Mechanics, Ch. 10 Key results used throughout this article: the König decomposition and vanishing cross term; the Lagrange-identity expansion giving $I_{ab}=\sum m_i(\delta_{ab}r^2-x_ax_b)$ ; the rolling-energy factors $7/10$ (solid sphere) and $1$ (hoop) from standard moments of inertia; Earth's $T_{\mathrm{rot}}\approx 2.1\times10^{29}\,\mathrm J$ from $I\approx8.0\times10^{37}\,\mathrm{kg\,m^2}$ and $\omega\approx7.29\times10^{-5}\,\mathrm{s^{-1}}$ . Worked-example values (downhill speeds $4.10/3.96/3.43\ \mathrm{m/s}$ for $h=1.2$ m; Explorer 1 figures $I_3=0.0448$ , $I_\perp=4.93\ \mathrm{kg\,m^2}$ , $T_0=138$ J, $T_f=1.26$ J, final tumble $6.8$ rpm) are illustrative arithmetic computed from the boxed formulas with the stated model dimensions.

Builds on: Moment of Inertia Tensor · Products of Inertia and the Inertia Matrix · Angular Velocity in Terms of Euler Angles. Leads to: Principal Axes and Principal Moments of Inertia · Inertia Ellipsoid · Euler's Equations of Motion · Free Symmetric Top and Body-Cone/Space-Cone Precession.

Why is there a factor of one half in T equals one half omega dot L, when omega dot L already looks like a complete pairing?

The half is the same one that appears in every quadratic energy — $\tfrac12mv^2$ , $\tfrac12kx^2$ , $\tfrac12CV^2$ — and it has the same origin. Energy is the work done building the state up from rest: as you spin the body up from $0$ to $\boldsymbol\omega$ , the angular momentum grows proportionally ( $\mathbf L=\mathbf I\boldsymbol\omega$ ), so the average momentum over the spin-up is half the final value. Integrating $\mathrm dW = \boldsymbol\omega'\cdot\mathrm d\mathbf L = \boldsymbol\omega'\cdot\mathbf I\,\mathrm d\boldsymbol\omega'$ from $0$ to $\boldsymbol\omega$ gives exactly $\tfrac12\boldsymbol\omega\cdot\mathbf I\boldsymbol\omega$ . Writing $\boldsymbol\omega\cdot\mathbf L$ without the half double-counts, just as $qV$ overstates the energy stored on a capacitor.

Is rotational kinetic energy the same number in the body frame and the space frame?

Yes — and it is worth seeing why, because the ingredients transform while the answer does not. $T_{\mathrm{rot}}=\tfrac12\boldsymbol\omega^{\mathsf T}\mathbf I\,\boldsymbol\omega$ is a scalar contraction of a tensor with two vectors; under any rotation matrix $R$ , $\boldsymbol\omega\to R\boldsymbol\omega$ and $\mathbf I\to R\,\mathbf I\,R^{\mathsf T}$ , and the combination is invariant. In the body frame $\mathbf I$ is constant and $\boldsymbol\omega$ 's components move; in the space frame $\boldsymbol\omega$ 's components may be steadier but $\mathbf I(t)$ rotates with the body. Same number either way. What you may not do is mix frames — body-frame $I_i$ with space-frame $\omega$ components — which is a classic way to manufacture nonsense energies.

How can Explorer 1 lose 99% of its rotational energy with no external torque? Where does conservation of energy go?

Total energy is conserved; mechanical rotational energy is not. The flexing whip antennas acted as internal dampers: as the body wobbled, they bent cyclically and converted ordered rotational energy into heat, which radiated away. Internal forces come in action–reaction pairs, so they cannot change the total angular momentum — $\mathbf L$ was untouchable — but nothing protects $T_{\mathrm{rot}}$ . Equation (7) then forces the outcome: at fixed $L$ , the accessible states run from $T=L^2/2I_{\min}$ down to $T=L^2/2I_{\max}$ , and dissipation walks the body down to the floor, the flat spin. The "missing" 137 J left as infrared radiation from slightly warmed antennas.

When exactly is the scalar shortcut T equals one half I omega squared safe?

Two cases, and only two. First, rotation about a fixed axis $\hat n$ — bearings, hinges, wheels on axles: then $T=\tfrac12 I_{\hat n}\omega^2$ with $I_{\hat n}=\hat n^{\mathsf T}\mathbf I\hat n=\int r_\perp^2\,\mathrm dm$ , always exact, even if $\hat n$ is not principal (though then $\mathbf L$ is not along $\hat n$ and the bearings feel it). Second, free rotation about a principal axis, where $\mathbf L=I\boldsymbol\omega$ genuinely holds with one scalar. In every other situation — a tumbling body, a top mid-nutation — the products of inertia or the multiple principal moments contribute and you need eq. (4) or (5). Ten-second test: is the rotation axis either enforced by hardware or an eigenvector of $\mathbf I$ ? If neither, no scalar shortcut.

Do the energy ellipsoid and the inertia ellipsoid differ, or are they the same surface?

Same family, different normalization — and conflating them costs marks. The inertia ellipsoid is the fixed geometric object $\sum_i I_i x_i^2=1$ attached to the body: pure mass-distribution data, no motion in it. The energy ellipsoid $\sum_i I_i\omega_i^2=2T$ lives in $\boldsymbol\omega$ -space and its overall size depends on how much energy the actual motion has; it is the inertia ellipsoid rescaled by $\sqrt{2T}$ . Poinsot's construction uses the inertia ellipsoid rolling on the invariable plane, with the current $\boldsymbol\omega$ always piercing the contact point. Both have their longest axis along the smallest principal moment — cheap directions permit big $\omega$ .

If rotation about the intermediate axis is an energy-allowed state, why do tumbling bodies avoid it?

Being allowed is not being stable. Exact spin about any principal axis, intermediate included, solves the equations of motion — set $\omega_j=\omega_k=0$ and nothing evolves. But the polhode geometry from the conservation derivation shows the difference: around the largest and smallest axes, neighboring polhodes are small closed loops, so a perturbed spin stays perturbed-but-close. Through the intermediate axis the polhodes form a separatrix — an X-shaped crossing — and any perturbation, however small, rides a polhode that carries $\boldsymbol\omega$ far around the ellipsoid and back: the tennis-racket flip. With dissipation the verdict sharpens further: only the maximum-moment axis survives as the terminal state, since it is the energy minimum at fixed $L$ .

Why does the rolling-race result depend only on the shape factor and not on mass or radius?

Because both the energy supply and the energy price scale identically. The supply is $Mgh\propto M$ ; the cost of speed $v$ is $\tfrac12Mv^2(1+\beta)$ with $\beta=I/MR^2$ , also $\propto M$ — so mass cancels. Radius cancels inside $\beta$ itself: $I$ for any homogeneous shape is (number) $\times MR^2$ , so $\beta$ is a pure geometry label — $\tfrac25$ for a solid sphere, $\tfrac12$ for a solid cylinder, $1$ for a hoop. That is why the race outcome is a classification of shapes, not a contest of sizes: any solid sphere beats any solid cylinder beats any hoop, whether marbles or boulders. What mass and radius do change is the angular speed $\omega=v/R$ and the absolute energies, never the finishing order or the split.

Is there rotational kinetic energy hiding in a body's rest mass relativistically?

Yes, in exactly the sense the correspondence limit suggests. A spinning flywheel is more massive than the same flywheel at rest by $\Delta m = T_{\mathrm{rot}}/c^2$ — the internal kinetic energy of the circulating mass elements contributes to the invariant mass of the composite body, just as thermal motion does. For any laboratory object the correction is fantastically small: Earth's entire $2.1\times10^{29}\ \mathrm J$ of spin energy amounts to about $2.4\times10^{12}\ \mathrm{kg}$ , a few parts in $10^{12}$ of its mass. The Newtonian quadratic form is the leading term of the relativistic energy whenever peripheral speeds satisfy $\omega R\ll c$ — comfortably true even for millisecond pulsars' surfaces, marginally violated only in their cores, where full relativistic treatments take over.