Angular velocity vector of a rotating frame

Here is a small scandal at the foundation of rotational mechanics. Take a book, rotate it a quarter turn about a vertical axis, then a quarter turn about a horizontal one. Now start over and do the same two rotations in the opposite order. The book ends up in different orientations. Finite rotations do not commute — so they cannot possibly be vectors, because vector addition never cares about order. And yet every mechanics course asks you to draw the angular velocity $\boldsymbol{\omega}$ as an arrow, add such arrows tip-to-tail, and resolve them into components as though nothing were wrong. The resolution of the scandal is one of the quiet miracles of mechanics: while rotations themselves live in the unruly world of $3\times3$ orthogonal matrices, the rate of rotation is governed by an antisymmetric matrix — and a $3\times3$ antisymmetric matrix has exactly three independent entries, precisely the data of a vector. Differentiation tames rotation. To first order in $\mathrm{d}t$ , rotations do commute, and their generators add like arrows. This article establishes $\boldsymbol{\omega}$ properly: why a single vector captures the turning of an entire frame (Euler's theorem — at each instant there is one axis), why it exists and is unique (the antisymmetry argument), why angular velocities add vectorially when finite rotations refuse to, and why $\dot{\boldsymbol{\omega}}$ is the rare derivative both frames agree on. Everything downstream — the operator identity, the velocity and acceleration transformations, Coriolis, centrifugal and Euler forces, the physics of the spinning Earth — takes the vector built here as its single input.

Key takeaways

The angular velocity vector $\boldsymbol{\omega}$ encodes the instantaneous axis of rotation (its direction) and the rotation rate (its magnitude) of a moving frame relative to an inertial one.
Its defining property is that any vector $\mathbf{e}$ fixed in the rotating frame evolves as $\dot{\mathbf{e}} = \boldsymbol{\omega}\times\mathbf{e}$ — rotation acts as a cross product.
$\boldsymbol{\omega}$ is the axial vector dual to the antisymmetric matrix $\Omega = \dot{R}R^{T}$ , where $R(t)$ is the rotation taking inertial axes to body axes; antisymmetry is what guarantees existence and uniqueness.
Although finite rotations do not commute (they are not vectors), infinitesimal rotations and hence angular velocities add as ordinary vectors: $\boldsymbol{\omega}_{\text{total}}=\boldsymbol{\omega}_1+\boldsymbol{\omega}_2$ .
$\boldsymbol{\omega}$ is generally time-dependent in both magnitude and direction, and $\dot{\boldsymbol{\omega}}$ — the input to the Euler force — is the same in both frames because $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ .
Earth's $\boldsymbol{\omega}$ has magnitude $7.292\times10^{-5}\,\mathrm{rad/s}$ (one turn per sidereal day), pointing along the rotation axis from south to north.
Key idea: The derivative of any rotation matrix has the form $\dot R = \Omega R$ with $\Omega$ antisymmetric; the vector $\boldsymbol{\omega}$ dual to $\Omega$ is the angular velocity, and it acts on every body-fixed vector by $\dot{\mathbf{e}}=\boldsymbol{\omega}\times\mathbf{e}$ .

Where this lives today

Every drone, phone, and spacecraft measures $\boldsymbol{\omega}$ directly. MEMS rate gyroscopes — three per device, one per axis — sense the Coriolis force on a vibrating proof mass and report the components of the body's angular velocity a thousand times per second. The flight controller then integrates orientation using exactly the mathematics of this article: attitude is stored as a quaternion (Hamilton's 1843 invention, chosen because finite rotations are not vectors and Euler angles jam in gimbal lock), and it is updated through the kinematic relation driven by the measured $\boldsymbol{\omega}$ — the matrix statement $\dot{R} = \Omega R$ of Derivation A in quaternion clothing. When a quadcopter snaps level after a gust, it is Derivation A running in firmware. The same structure organizes modern robotics and simulation. The pair $SO(3)$ / $\mathfrak{so}(3)$ — finite rotations versus their additive rates — is now the standard language for legged-robot balance controllers, spacecraft attitude filters, and the geometric integrators used in molecular dynamics and game physics engines, all of which exploit the fact proved in Derivation B: you may add rates freely, but you must compose orientations multiplicatively. Even planetary science leans on the addition law daily, decomposing $\boldsymbol{\omega}_\oplus$ into local components to predict Foucault precession rates and satellite ground tracks.

The physical picture

Picture a vector $\mathbf{e}$ glued to a turntable. As the table turns, the tip of $\mathbf{e}$ traces a circle whose plane is perpendicular to the axis. The tip's velocity is tangent to that circle, proportional to the distance from the axis, and perpendicular to both the axis and $\mathbf{e}$ — exactly the recipe $\dot{\mathbf{e}}=\boldsymbol{\omega}\times\mathbf{e}$ . The cross product is not an analogy; it is forced by the geometry of rigid rotation.

Figure

The defining picture — a frame-fixed vector sweeps a cone about the instantaneous axis at rate ω

Why is $\boldsymbol{\omega}$ a single vector when orientation needs three angles to specify? Because of Euler's rotation theorem: at any instant, however tangled the motion, a rigid frame is turning about exactly one axis, with every frozen vector momentarily circling it. The angular velocity collapses the turning into one axis plus one rate — an arrow. The orientation itself never simplifies this way; it remains a point in the non-commutative group $SO(3)$ , which is why the scandal of the overview is real:

Figure

Finite rotations do not commute — the same two quarter-turns in opposite order end differently

The rescue happens at first order. Composing two infinitesimal rotations $\mathbf{1}+\Omega_1\,\mathrm{d}t$ and $\mathbf{1}+\Omega_2\,\mathrm{d}t$ gives $\mathbf{1}+(\Omega_1+\Omega_2)\,\mathrm{d}t$ regardless of order — the non-commuting cross terms are second order in $\mathrm{d}t$ and vanish in the limit. This is why you may superpose the Earth's spin and a carousel's spin by simply adding their $\boldsymbol{\omega}$ 's, and it is the precise sense in which differentiation converts the group of rotations into a vector space (the mathematicians' $SO(3) \to \mathfrak{so}(3)$ ).

The core equations

The whole formalism can be hung on one defining property. Glue a unit vector to the rotating frame and watch it from the inertial frame:

\dot{\mathbf{e}}_i = \boldsymbol{\omega}\times\mathbf{e}_i \qquad (1)

where: $\mathbf{e}_i$ — a basis vector fixed in the rotating frame, $i=1,2,3$ ; $\boldsymbol{\omega}$ — angular velocity vector (rad/s); $\dot{\mathbf{e}}_i$ — its rate of change as seen in the inertial frame. This is the Poisson equation of rigid kinematics: one vector $\boldsymbol{\omega}$ drives all three basis vectors (indeed every frame-fixed vector) at once. That a single $\boldsymbol{\omega}$ can do this job — and only one — is precisely what Derivation A establishes, working not with the basis but with the rotation matrix itself. The matrix version packages the same information. If $R(t)$ is the orthogonal matrix carrying inertial axes onto the instantaneous body axes, then differentiating the orthogonality constraint produces

\Omega \equiv \dot{R}\,R^{T},\qquad \Omega^{T}=-\Omega \qquad (2)

where: $R(t)$ — $SO(3)$ matrix mapping inertial axes to body axes; $\Omega$ — angular velocity matrix (antisymmetric). Antisymmetry is the crucial structural fact: it cuts the nine entries of a $3\times3$ matrix down to three independent ones. Those three entries are the components of $\boldsymbol{\omega}$ , arranged so that matrix action reproduces the cross product:

\Omega=\begin{pmatrix}0&-\omega_3&\omega_2\\\omega_3&0&-\omega_1\\-\omega_2&\omega_1&0\end{pmatrix},\quad \Omega\,\mathbf{v}=\boldsymbol{\omega}\times\mathbf{v} \qquad (3)

where: $\omega_i$ — components of $\boldsymbol{\omega}$ in the chosen basis; $\mathbf{v}$ — any vector acted on by the rotation rate. The duality runs both ways: every angular velocity vector defines an antisymmetric matrix and vice versa. (Being built from a cross product, $\boldsymbol{\omega}$ is an axial vector — it flips sign under mirror reflection, a fact with consequences explored in the FAQ.) With $\boldsymbol{\omega}$ in hand, the master tool of rotating-frame mechanics follows — for any vector $\mathbf{A}$ , frame-fixed or not, decompose it on the rotating basis and use (1):

\left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\text{in}}=\left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\text{rot}}+\boldsymbol{\omega}\times\mathbf{A} \qquad (4)

where: $\mathbf{A}$ — any vector quantity (position, velocity, angular momentum, ...); $(\cdot)_{\text{in}}$ — derivative computed in the inertial frame; $(\cdot)_{\text{rot}}$ — derivative computed in the rotating frame (components only). We state it here because it is $\boldsymbol{\omega}$ 's job description; its careful derivation, iteration, and consequences are the subject of the time-derivatives article, and iterating it twice generates every fictitious force of the acceleration-transformation article. Because infinitesimal rotations commute (Derivation B), angular velocities of stacked rotations superpose:

\boldsymbol{\omega}_{\text{total}}=\boldsymbol{\omega}_{1}+\boldsymbol{\omega}_{2} \qquad (5)

where: $\boldsymbol{\omega}_1$ — angular velocity of frame 1 relative to the inertial frame; $\boldsymbol{\omega}_2$ — angular velocity of frame 2 relative to frame 1; $\boldsymbol{\omega}_{\text{total}}$ — angular velocity of frame 2 relative to the inertial frame. This is the rule that lets you compose a spinning wheel on a turning platform on a rotating planet by drawing three arrows and adding them — Worked Example 1 does exactly that. Feed $\boldsymbol{\omega}$ itself into the identity (4) and the cross term self-destructs:

\left(\dot{\boldsymbol{\omega}}\right)_{\text{in}}=\left(\dot{\boldsymbol{\omega}}\right)_{\text{rot}} \qquad (6)

where: $\dot{\boldsymbol{\omega}}$ — angular acceleration vector, the same in both frames since $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ . The angular velocity is the unique vector whose derivative needs no frame label — which is why the Euler force $-m\dot{\boldsymbol{\omega}}\times\mathbf{r}$ is unambiguous. The number that matters most in practice belongs to our own planet:

|\boldsymbol{\omega}_\oplus|=\frac{2\pi}{T_{\text{sidereal}}}=7.292\times10^{-5}\,\mathrm{rad/s} \qquad (7)

where: $T_{\text{sidereal}}$ — sidereal day $\approx 86164\,\mathrm{s}$ (not the $86400\,\mathrm{s}$ solar day); $\boldsymbol{\omega}_\oplus$ — directed along the spin axis from south to north pole. Small as it looks, this vector — resolved into local vertical and horizontal components at your latitude — runs the Foucault pendulum, the trade winds, and the eastward deflection of everything that falls.

Derivation AExistence and uniqueness of the angular velocity vector

We are proving that the turning of any rotating frame, however complicated, is captured at each instant by a single well-defined vector $\boldsymbol{\omega}$ — the claim that underwrites every arrow-drawing habit of rotational mechanics. The strategy is to work with the rotation matrix $R(t)$ itself: differentiate the orthogonality constraint $RR^T = \mathbf{1}$ , discover that the rate matrix $\Omega = \dot{R}R^T$ is forced to be antisymmetric, and then count parameters — three independent entries, exactly a vector's worth, with the cross-product action emerging by direct multiplication. Uniqueness comes free, because the packaging map between antisymmetric matrices and vectors is a bijection. Watch how little input the argument needs: rigidity (lengths and angles preserved) is the only physical assumption.

Step 1. Encode rigidity. The matrix $R(t)$ mapping inertial axes onto body axes preserves all lengths and angles, which is the statement

R(t)\,R(t)^{T}=\mathbf{1}

Step 2. Differentiate the constraint (the right side is constant):

\dot R R^{T}+R\dot R^{T}=0

Step 3. Define $\Omega\equiv\dot R R^{T}$ and recognize the second term as its transpose, $R\dot R^{T}=(\dot R R^{T})^{T}=\Omega^{T}$ :

\Omega+\Omega^{T}=0

Why antisymmetry is the pivot: it is the infinitesimal shadow of orthogonality — the constraint that rotation may reorient vectors but never stretch them shows up, at the level of rates, as a matrix with zero diagonal and mirror-negated off-diagonal entries.

Step 4. Count parameters. An antisymmetric $3\times3$ matrix has exactly three independent entries, which we name $(\omega_1,\omega_2,\omega_3)$ and arrange as in equation (3):

\Omega=\begin{pmatrix}0&-\omega_3&\omega_2\\\omega_3&0&-\omega_1\\-\omega_2&\omega_1&0\end{pmatrix}

Step 5. Verify the action. Direct multiplication of this $\Omega$ against an arbitrary $\mathbf{v} = (v_1,v_2,v_3)$ reproduces $\boldsymbol{\omega}\times\mathbf{v}$ component by component. Why uniqueness follows: the map $\boldsymbol{\omega}\mapsto\Omega$ is linear and invertible — a bijection between vectors and antisymmetric matrices — so the $\boldsymbol{\omega}$ representing a given rate of turning is the only one.

Step 6. Recover the physical velocity field. A point frozen at body coordinates $\mathbf{r}_0$ sits at $\mathbf{r} = R\,\mathbf{r}_0$ , so

\dot{\mathbf{r}}=\dot R\,\mathbf{r}_0=\Omega R\,\mathbf{r}_0=\boldsymbol{\omega}\times\mathbf{r}

Why this closes the loop: the abstract matrix construction lands exactly on the turntable picture — every frame-fixed point circles the instantaneous axis with the rigid-rotation velocity field, equation (1). *Verified: writing R as an explicit rotation about the z-axis by angle θ(t) gives Ω with ω₃ = θ̇ and ω₁ = ω₂ = 0, i.e. ω = θ̇ ẑ, as it must.*

Go deeper · Derivation B · 2 minWhy angular velocities add although finite rotations do not

Derivation BWhy angular velocities add although finite rotations do not

The goal is the addition law (5) — and an honest account of why it is allowed, given that finite rotations flagrantly fail to commute. The strategy is to expand two rotations executed over the same infinitesimal interval to second order in $\mathrm{d}t$ , compose them in both orders, and locate exactly where the order-dependence lives. It turns out to sit entirely in the second-order commutator term: at first order, generators simply add. Since angular velocity is by definition the first-order rate of rotation, order-independence — and hence vector addition — is exact in the limit. The second-order remainder is not a nuisance to be swept away; it is the fingerprint of the non-abelian group structure of rotations.

Step 1. Expand each infinitesimal rotation to second order:

R_i=\mathbf{1}+\Omega_i\,\mathrm{d}t+\tfrac12\Omega_i^2(\mathrm{d}t)^2+\cdots

Why keep second order: the effect we are hunting — non-commutativity — is invisible at first order, so truncating too early would prove the result by accident rather than by argument.

Step 2. Compose in one order:

R_2R_1=\mathbf{1}+(\Omega_1+\Omega_2)\,\mathrm{d}t+\left(\tfrac12\Omega_1^2+\Omega_2\Omega_1+\tfrac12\Omega_2^2\right)(\mathrm{d}t)^2+\cdots

Step 3. Subtract the reverse order. All first-order terms are symmetric under $1\leftrightarrow2$ and cancel; what survives is

R_2R_1-R_1R_2=(\Omega_2\Omega_1-\Omega_1\Omega_2)\,(\mathrm{d}t)^2+\mathcal{O}((\mathrm{d}t)^3)

Why this quantifies the book experiment: swapping the order of two rotations changes the outcome only at order $(\mathrm{d}t)^2$ — for two 1° turns the discrepancy is of order (1°)² ≈ 3×10⁻⁴ rad, while for two 90° turns it is enormous. Small rotations almost commute; instantaneous ones exactly do.

Step 4. Read off the first-order composition:

R_2R_1\big|_{\mathcal{O}(\mathrm{d}t)}=\mathbf{1}+(\Omega_1+\Omega_2)\,\mathrm{d}t

The generators add, and by the duality of equation (3) so do the vectors: $\boldsymbol{\omega}_{\text{total}}=\boldsymbol{\omega}_1+\boldsymbol{\omega}_2$ .

Step 5. Name the structure. Finite rotations live in the non-abelian group $SO(3)$ , where order matters; their derivatives at the identity live in the vector space $\mathfrak{so}(3)$ of antisymmetric matrices, where addition is unconditional. Why this matters beyond vocabulary: it locates $\boldsymbol{\omega}$ 's home — angular velocities are elements of a genuine vector space, which is the licence behind every decomposition of Earth's rotation into vertical and horizontal components at your latitude. *Verified: consistent with the standard commutator result for SO(3); the 90°-about-x versus 90°-about-y book test shows the finite-rotation discrepancy, and the discrepancy scales as the angle squared for small angles, as Step 3 predicts.*

Limiting cases

How $\boldsymbol{\omega}$ behaves in special regimes:

Case	Result	Meaning
Pure spin (fixed axis)	$\boldsymbol{\omega}=\omega\,\hat{\mathbf{n}},\quad \dot{\boldsymbol{\omega}}=0$	Constant rate about a fixed axis: the Euler force vanishes, only Coriolis and centrifugal survive. The turntable / steady-spinning Earth case.
Spin-up / spin-down	$\boldsymbol{\omega}=\omega(t)\,\hat{\mathbf{n}},\quad \dot{\boldsymbol{\omega}}=\dot\omega\,\hat{\mathbf{n}}$	Axis fixed but rate changing: Euler force $-m\,\dot{\boldsymbol{\omega}}\times\mathbf{r}$ appears, directed tangentially.
Precession (axis itself rotating)	$\boldsymbol{\omega}=\boldsymbol{\Omega}_{\text{prec}}+\boldsymbol{\omega}_{\text{spin}}$	Total angular velocity is the vector sum of precession and spin rates; the instantaneous axis sweeps a cone.
Slow rotation (perturbative)	$\vert \boldsymbol{\omega}\vert \,T \ll 1$	When the rotation period $2\pi/\vert \boldsymbol{\omega}\vert$ greatly exceeds the dynamical timescale $T$ , fictitious forces are small corrections (e.g. Coriolis on a thrown ball over seconds).
Non-rotating limit	$\boldsymbol{\omega}\to 0$	Frame becomes (at most) linearly accelerating; all $\boldsymbol{\omega}$ -dependent fictitious forces vanish and only a possible translational term remains.

Common trap“Rotations are vectors — just add the angles”

Finite rotations are NOT vectors — rotating a book $90°$ about $x$ then $90°$ about $y$ leaves it in a different orientation than the reverse order, while vector addition could never care. Only infinitesimal rotations, and hence angular velocities, add vectorially; the order-dependence of finite rotations is a second-order effect that survives composition, which is why orientation must be tracked with matrices, Euler angles, or quaternions rather than with a running “angle vector.” A second trap rides along: $\boldsymbol{\omega}$ specifies the instantaneous axis, and that axis can itself move — do not assume the rotation axis is fixed in space unless $\dot{\boldsymbol{\omega}} = 0$ . A gyroscope's $\boldsymbol{\omega}$ sweeps a cone during precession even while its magnitude barely changes.

Worked examples

1 · A tilted turntable on a carousel (confidence builder)

A carousel turns about the vertical at $\Omega = 0.5\ \mathrm{rad/s}$ . Mounted on it is a turntable spinning at $\omega_s = 2.0\ \mathrm{rad/s}$ about an axis tilted $60°$ from the vertical. Find the magnitude and direction of the total angular velocity of the turntable relative to the ground.

Step 1. Invoke the addition law (5): $\boldsymbol{\omega}_{\text{total}} = \boldsymbol{\Omega} + \boldsymbol{\omega}_s$ . Angular velocities add as vectors even though the two rotations are about different, non-parallel axes — that is the entire content of Derivation B.

Step 2. Resolve into vertical and horizontal components. Vertical: $\Omega + \omega_s\cos 60° = 0.5 + 2.0\times0.5 = 1.50\ \mathrm{rad/s}$ . Horizontal: $\omega_s\sin 60° = 2.0\times0.866 = 1.73\ \mathrm{rad/s}$ .

Step 3. Recombine: $|\boldsymbol{\omega}_{\text{total}}| = \sqrt{1.50^2 + 1.73^2} = \boxed{2.29\ \mathrm{rad/s}}$ , tilted $\arctan(1.73/1.50) = \boxed{49.1°}$ from the vertical.

Step 4. Interpret with care: this is the instantaneous axis. Because the turntable's tilted axis is itself carried around by the carousel, the horizontal component of $\boldsymbol{\omega}_{\text{total}}$ rotates at rate $\Omega$ — the total vector sweeps a cone, so $\dot{\boldsymbol{\omega}} \neq 0$ here even though both rates are individually constant. (Riders of this contraption feel an Euler force accordingly.)

2 · The rolling wheel's secret axis (exam level)

A wheel of radius $a = 5.0\ \mathrm{cm}$ rolls without slipping around a circle of radius $R = 30\ \mathrm{cm}$ , its axle horizontal, its centre moving at $v = 0.60\ \mathrm{m/s}$ . Find the total angular velocity vector and show that the instantaneous axis passes through the contact point.

Step 1. Identify the two stacked rotations. The axle assembly precesses about the vertical through the circle's centre at $\Omega = v/R = 0.60/0.30 = 2.0\ \mathrm{rad/s}$ ; the wheel spins about its (horizontal, radial) axle at $\omega_s = v/a = 0.60/0.05 = 12\ \mathrm{rad/s}$ .

Step 2. Add the vectors: they are perpendicular (vertical + horizontal), so $|\boldsymbol{\omega}_{\text{total}}| = \sqrt{\Omega^2 + \omega_s^2} = \sqrt{2.0^2 + 12^2} = \boxed{12.2\ \mathrm{rad/s}}$ , tilted below the horizontal axle direction by $\arctan(\Omega/\omega_s) = \arctan(2/12) = \boxed{9.5°}$ .

Step 3. Locate the axis geometrically. The resultant is tilted downward from the axle by an angle whose tangent is $\Omega/\omega_s = a/R$ (since $v = \Omega R = \omega_s a$ ). A line through the wheel's centre with slope $a/R$ toward the circle's centre drops a height $a$ over the horizontal run $R$ — it passes exactly through the contact point on the floor.

Step 4. Confirm with the no-slip condition. Rolling without slipping means the material point touching the floor is momentarily at rest; a rigid body's zero-velocity points always lie on the instantaneous axis, so the axis must pass through the contact point — the vector addition and the kinematic constraint agree. Points on this line, and only these, are momentarily stationary.

Step 5. Sanity check the limits: $R \to \infty$ (straight-line rolling) tilts the axis to horizontal through the contact point, the familiar result that a rolling wheel momentarily pivots about its contact line; $a \to R$ steepens the tilt to $45°$ .

Origins · 1 minLeonhard Euler and how this was discovered

Leonhard Euler

1707–1783 · Basel → St Petersburg → Berlin

Rotational kinematics is Euler's private estate. In the 1750s and 1760s, working between the academies of Berlin and St Petersburg, he built the entire apparatus this article uses: body-fixed rotating axes, the decomposition of any orientation into three angles (the Euler angles), and in 1775 the theorem that guarantees $\boldsymbol{\omega}$ exists — any displacement of a rigid body about a fixed point is a rotation about a single axis. That instantaneous-axis theorem is why the turning of a frame, which naively needs a full matrix, collapses at each moment to one arrow. Nearly blind for his last decade, Euler dictated close to half of his life's work — a corpus so vast the Petersburg Academy was still publishing his backlog forty-eight years after his death. When mechanics later needed the angular velocity of a frame that is speeding up its spin, the resulting fictitious force took his name almost by default: the Euler force, the subject of its own article downstream.

History

1750

— Euler publishes his new principle of mechanics, extending Newton's law to rigid bodies referred to body-fixed rotating axes.

1775

— Euler's rotation theorem: any reorientation of a rigid body about a point is a single rotation about one axis — the guarantee behind the instantaneous ω.

1840

— Rodrigues parametrizes finite rotations and their composition, exposing precisely the structure that makes them non-additive.

1843

— Hamilton carves the quaternion equations into Brougham Bridge, giving rotation composition its most durable computational form.

1901

— Gibbs and Wilson's Vector Analysis standardizes the modern notation in which ω is drawn, added, and resolved as an ordinary arrow.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 9 (Angular Momentum) and Ch. 10 (Accelerating Frames) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §4.8–4.9 (rate of change of a vector, Coriolis effect) · Landau & Lifshitz, Mechanics, Vol. 1, §39 (motion in a non-inertial frame) · Taylor, Classical Mechanics, Ch. 9 (Rotating frames)

Provenance. Equations cross-checked against Goldstein §4.8–4.9 (rate-of-change operator, antisymmetric $\Omega$ ), Morin Ch. 9–10 (Poisson equation, fictitious forces), Landau-Lifshitz §39, and Taylor Ch. 9. Verifiers should recompute: (1) $\dot R R^T$ antisymmetry from $RR^T=\mathbf 1$ ; (2) the explicit $\Omega$ matrix gives $\Omega\mathbf v=\boldsymbol\omega\times\mathbf v$ ; (3) Earth's $|\boldsymbol\omega|=2\pi/86164\,\mathrm{s}=7.292\times10^{-5}\,\mathrm{rad/s}$ (standard value, e.g. IAU); (4) the first-order cancellation in the rotation commutator. No fabricated constants used.

Builds on: Non-inertial reference frames · Vector cross product and antisymmetric matrices · Orthogonal rotation matrices $SO(3)$ Leads to: Time derivatives in rotating frames · Transformation of velocity between rotating frames · Transformation of acceleration in rotating frames · Coriolis force · Euler force · Effective gravity on the rotating Earth

If finite rotations are not vectors, what exactly am I adding when I add angular velocities?

You are adding generators — the first-order rates of rotation — not the rotations themselves. Over an infinitesimal interval $\mathrm{d}t$ , each rotation is $\mathbf{1}+\Omega\,\mathrm{d}t$ plus second-order terms, and composition in either order gives $\mathbf{1}+(\Omega_1+\Omega_2)\mathrm{d}t$ : the order-dependence lives entirely at order $(\mathrm{d}t)^2$ and dies in the limit. So $\boldsymbol{\omega}$ -arrows genuinely inhabit a vector space (the Lie algebra $\mathfrak{so}(3)$ ), while orientations inhabit the group $SO(3)$ where you must multiply, never add. Practical rule: rates — add freely, decompose into components, average. Orientations — compose with matrices or quaternions, and never try to define a running "total angle vector."

Is ω attached to a particular point of the body, like the centre or the axis?

No — $\boldsymbol{\omega}$ is a property of the body (or frame) as a whole, a "free vector." Ask any two frame-fixed vectors how fast they are turning and both answer with the same $\boldsymbol{\omega}$ ; the velocity formula $\mathbf{v} = \boldsymbol{\omega}\times\mathbf{r}$ assigns different velocities to different points only because their positions $\mathbf{r}$ differ. What does depend on a reference choice is the split of a general rigid motion into translation-plus-rotation: change the reference point and the translational velocity changes, but $\boldsymbol{\omega}$ itself is invariant. That is why one can speak of "the" angular velocity of the Earth without specifying a point in it.

What does it mean that ω is a pseudovector — is it less real than a force?

It is exactly as real, but it transforms differently under reflection. A true (polar) vector like velocity flips its component perpendicular to a mirror; $\boldsymbol{\omega}$ , built from a cross product (equation 3), does the opposite — watch a clockwise turntable in a mirror and it turns counterclockwise, so the mirrored $\boldsymbol{\omega}$ points the other way along the axis. The deeper point: the fundamental object is the antisymmetric matrix $\Omega$ (a rotation plane plus a rate); packaging it as an axis-vector is a convenience special to three dimensions, where every plane has a unique perpendicular. In two dimensions the same physics is a single number; in four dimensions it refuses to be a vector at all.

My drone's gyro outputs three numbers. Are those the Euler-angle rates?

No — and confusing the two is a classic firmware bug. The gyro measures the components of $\boldsymbol{\omega}$ along the drone's body axes: honest components of a genuine vector. Euler-angle rates $(\dot\phi, \dot\theta, \dot\psi)$ are the speeds of three nested, differently-oriented rotations, and they relate to $\boldsymbol{\omega}$ through an angle-dependent linear map — one that becomes singular when the middle angle reaches its gimbal-lock value (pitch $\pm90°$ in the aerospace convention), where finite Euler rates cannot represent a finite $\boldsymbol{\omega}$ . Flight controllers therefore integrate the quaternion or matrix kinematics driven directly by the measured $\boldsymbol{\omega}$ , and extract Euler angles only for display.

If the instantaneous axis can move, in what sense is it the axis of rotation?

In the strict but momentary sense of Euler's theorem: right now, every point of the body is moving as if circling that line — points on it have zero velocity (if the axis passes through fixed points, as in rolling) and everything else moves perpendicular to its separation from it. A moment later the axis may lie elsewhere, both in space and in the body; its migrating locus traces the "space cone" and "body cone" of rigid-body dynamics, which roll on each other without slipping. The rolling wheel of Worked Example 2 makes it concrete: the axis passes through the contact point now, but through a different material point of the wheel at each instant.

Do the angular velocities of orbit and spin add too — is Earth's total ω the sum of its daily spin and its yearly revolution?

Careful — the addition law (5) stacks the rotations of frames, and it applies only when you genuinely compose them. Earth's orbital revolution and its axial spin are two separate motions of different things: the orbit moves the planet's centre (a translation of the body through space), while $\boldsymbol{\omega}_\oplus$ describes the turning of the body about that centre. You would add them only for a frame rigidly locked to the Earth–Sun line (a frame that rotates once per year) carrying a sub-frame spinning relative to it — which is precisely the accounting behind the sidereal/solar day difference: the Earth must rotate about $1/365.24$ of a turn extra per solar day, making the solar day 236 seconds longer than the sidereal one.

Why do textbooks bother with the matrix Ω when the vector ω does everything?

Three reasons. Conceptually, $\Omega = \dot{R}R^T$ is where $\boldsymbol{\omega}$ comes from — antisymmetry is a derived fact, not an assumption, and the derivation is the proof that a single instantaneous axis exists. Computationally, the matrix form generalizes: quaternion attitude propagation, Lie-group integrators, and every 6-DOF simulator implement $\dot{R} = \Omega R$ directly, because it has no coordinate singularities. And structurally, the matrix reveals what a rotation rate fundamentally is — a rate of turning in a plane — of which "a vector along the axis" is a three-dimensions-only shorthand. When you later meet angular momentum as an antisymmetric tensor in relativity, it is the same packaging decision being unwound.