Time derivatives in rotating frames

Chalk an arrow on the floor of a spinning carousel, pointing at one of the horses. To a child riding the platform, that arrow never changes — it is simply there, frozen. To you on the ground, the arrow sweeps around once per revolution, its tip tracing a circle. Both observers are watching the same physical object, and both are differentiating it with respect to the same time. They get different answers. Who is wrong? Neither — and turning that "neither" into an equation is the single most important kinematic step in non-inertial mechanics. In an inertial frame the basis vectors are fixed once and for all, so differentiating a vector means differentiating its components. In a rotating frame the basis vectors themselves turn, and every time derivative silently acquires a second contribution: the rotation of the measuring rods. Disentangling the two gives a one-line operator identity — inertial derivative equals rotating-frame derivative plus $\boldsymbol{\omega}\times$ — that holds for any vector whatsoever: position, velocity, angular momentum, a magnetic field, anything you can point at. The word "operator" is meant literally, and it is where the power lies. Because the rule applies to whatever vector you feed it — including the output of a previous application — it can be iterated. Applied once to position it transforms velocities; applied twice it manufactures every fictitious force of rotating-frame dynamics: centrifugal, Coriolis, and Euler, factor of 2 and all. Weather maps, Foucault's pendulum, the bulge of the Earth — all of it flows downstream from the bookkeeping rule this article builds.

Key takeaways

The central result is the operator identity $\left(\frac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\frac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{rot}} + \boldsymbol{\omega}\times$ , relating the rate of change of any vector in an inertial frame to its rate of change in a frame rotating with angular velocity $\boldsymbol{\omega}$ .
The extra term $\boldsymbol{\omega}\times\mathbf{A}$ arises purely because the rotating frame's basis vectors turn ( $\dot{\hat{\mathbf{e}}}_i = \boldsymbol{\omega}\times\hat{\mathbf{e}}_i$ ); it has nothing to do with how $\mathbf{A}$ is being physically changed.
A vector frozen in the rotating frame still changes in the inertial frame, at rate $\boldsymbol{\omega}\times\mathbf{A}$ — the carousel carrying the arrow around.
The identity holds for any vector and may be iterated: applied once to position it yields the velocity transformation, applied twice the acceleration transformation with its Coriolis, centrifugal, and Euler terms.
$\boldsymbol{\omega}$ itself is the one vector both frames agree about: $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ makes $\dot{\boldsymbol{\omega}}$ frame-independent — which is why the Euler force is unambiguous.
Key idea: For any vector $\mathbf{A}$ , $\left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{rot}} + \boldsymbol{\omega}\times\mathbf{A}$ ; the identity holds as an operator and may be iterated.

Where this lives today

Every spacecraft, drone, and smartphone carries this identity in its firmware. Attitude-control systems work in the vehicle's own body frame — a rotating frame par excellence — and every propagation step of the navigation filter converts rates between body and inertial coordinates using exactly equation (1); the quaternion update equations integrated at kilohertz rates in a flight controller are the operator identity in disguise. When a Mars-bound probe fires a reaction wheel, the guidance computer is differentiating angular momentum in a rotating frame and adding $\boldsymbol{\omega}\times\mathbf{L}$ , precisely the calculation of Worked Example 2. The largest-scale users are the geophysical modelers. Global weather and ocean models solve Newton's second law in the frame of the rotating Earth, where the momentum equation carries the fictitious terms generated by iterating this identity — the Coriolis term above all, whose consequences (from cyclone rotation to the Foucault pendulum) the Coriolis force article unpacks. From a hurricane forecast to the pointing of the James Webb Space Telescope riding Lagrange's L2, the one-line rule $\left(\mathrm{d}/\mathrm{d}t\right)_{\mathrm{in}} = \left(\mathrm{d}/\mathrm{d}t\right)_{\mathrm{rot}} + \boldsymbol{\omega}\times$ is doing the accounting.

The physical picture

Return to the carousel. The chalked arrow is a vector $\mathbf{A}$ ; the child's report is the rotating-frame derivative (zero — the components never change); your report from the ground is the inertial derivative. The entire disagreement is the platform carrying the arrow around, and the velocity of the tip of a frozen vector carried by a rotation is exactly $\boldsymbol{\omega}\times\mathbf{A}$ : perpendicular to both the axis and the arrow, proportional to how far the tip sits from the axis.

Figure

One arrow, two derivatives — the rotating observer sees the sliding, the inertial observer sees sliding plus sweep

Now let the child also slide the chalk arrow outward as she rides. The ground observer sees both motions added vectorially: the sliding (which the rotating observer measures) and the carrying-around (which she does not). That picture is the identity — everything else in this article is making it precise and exploring what it implies. The identity is best read as a bookkeeping rule for basis vectors. Any vector is $\mathbf{A}=A_i\hat{\mathbf{e}}_i$ (sum over $i$ ). Differentiating, $\dot{\mathbf{A}} = \dot{A}_i\hat{\mathbf{e}}_i + A_i\dot{\hat{\mathbf{e}}}_i$ . The first piece is what the rotating observer measures — components changing against her own axes. The second is what the rotation of the axes contributes, and since each rigid basis vector obeys $\dot{\hat{\mathbf{e}}}_i=\boldsymbol{\omega}\times\hat{\mathbf{e}}_i$ , it collapses to $\boldsymbol{\omega}\times\mathbf{A}$ . There is no physics in the second term beyond geometry — it is the rotation of the ruler, not of the thing measured. Two consequences deserve flagging before the algebra. First, because the rule is an operator, you can chain it: $D_{\mathrm{in}}=D_{\mathrm{rot}}+\boldsymbol{\omega}\times$ applied to its own output is what manufactures the quadratic-in- $\omega$ centrifugal term and the factor-2 Coriolis term automatically — no extra postulates. Second, the angular velocity itself is special: feed $\mathbf{A} = \boldsymbol{\omega}$ into the identity and the cross term $\boldsymbol{\omega}\times\boldsymbol{\omega}$ vanishes, so both frames agree on $\dot{\boldsymbol{\omega}}$ . The Euler force, which is built from $\dot{\boldsymbol{\omega}}$ , is unambiguous for exactly this reason.

The core equations

Fix an inertial frame and a frame rotating with angular velocity $\boldsymbol{\omega}(t)$ about a common origin. Any vector $\mathbf{A}(t)$ changes for two reasons: because its components change within the rotating frame, and because the frame's basis vectors sweep around. Writing that sentence in symbols gives the master key of the whole subject:

\left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{rot}} + \boldsymbol{\omega}\times\mathbf{A} \qquad (1)

where: $\mathbf{A}$ — any vector quantity (position, velocity, angular momentum, field, ...); $(\mathrm{d}/\mathrm{d}t)_{\mathrm{in}}$ — time derivative measured in the inertial frame; $(\mathrm{d}/\mathrm{d}t)_{\mathrm{rot}}$ — time derivative measured in the rotating frame; $\boldsymbol{\omega}$ — angular velocity of the rotating frame relative to the inertial frame. Both derivatives act on the same physical vector; they differ only in which set of axes is held fixed while differentiating. Stripping away the operand exposes the identity as a relation between derivative operators — the form in which it is iterated:

\left(\frac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\frac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{rot}} + \boldsymbol{\omega}\times \qquad (2)

where: $\times$ — left-multiplication by $\boldsymbol{\omega}$ via the cross product; the operator awaits a vector argument. Reading it as an operator is not pedantry: it licenses applying the rule to the output of a previous application, which is precisely how the acceleration transformation is built. The geometric engine underneath is the motion of the rotating triad itself. Each basis vector is a rigid unit vector carried by the rotation, so its tip rides a circle about the axis:

\dot{\hat{\mathbf{e}}}_i = \boldsymbol{\omega}\times\hat{\mathbf{e}}_i \qquad (3)

where: $\hat{\mathbf{e}}_i$ — orthonormal basis vectors fixed in the rotating frame, $i=1,2,3$ ; $\dot{\hat{\mathbf{e}}}_i$ — their inertial-frame time derivative. This is the kinematic origin of the cross-product term, and it is derived — not assumed — in Derivation A below. (Why the cross-product form is forced on a rigid triad, and why $\boldsymbol{\omega}$ deserves to be called a vector at all, is the business of the companion article on the angular velocity vector.) Feeding the position vector through the identity once gives the velocity transformation,

\mathbf{v}_{\mathrm{in}} = \mathbf{v}_{\mathrm{rot}} + \boldsymbol{\omega}\times\mathbf{r} \qquad (4)

where: $\mathbf{v}_{\mathrm{in}}=(\mathrm{d}\mathbf{r}/\mathrm{d}t)_{\mathrm{in}}$ — velocity in the inertial frame; $\mathbf{v}_{\mathrm{rot}}=(\mathrm{d}\mathbf{r}/\mathrm{d}t)_{\mathrm{rot}}$ — velocity as measured in the rotating frame; $\mathbf{r}$ — position vector (common origin on the rotation axis). The $\boldsymbol{\omega}\times\mathbf{r}$ term is the sweep velocity of the platform point the particle currently occupies. The full anatomy of this equation — when the two speeds differ, what it says on a turntable and on the rotating Earth — lives in the velocity-transformation article. Feeding the identity through twice — apply it to $\mathbf{r}$ , then to $\mathbf{v}_{\mathrm{in}}$ — produces the acceleration transformation:

\mathbf{a}_{\mathrm{in}} = \mathbf{a}_{\mathrm{rot}} + 2\,\boldsymbol{\omega}\times\mathbf{v}_{\mathrm{rot}} + \boldsymbol{\omega}\times(\boldsymbol{\omega}\times\mathbf{r}) + \dot{\boldsymbol{\omega}}\times\mathbf{r} \qquad (5)

where: $\mathbf{a}_{\mathrm{in}},\mathbf{a}_{\mathrm{rot}}$ — acceleration in inertial and rotating frames; $2\boldsymbol{\omega}\times\mathbf{v}_{\mathrm{rot}}$ — Coriolis term; $\boldsymbol{\omega}\times(\boldsymbol{\omega}\times\mathbf{r})$ — centripetal term (gives centrifugal force when moved to the force side); $\dot{\boldsymbol{\omega}}\times\mathbf{r}$ — Euler (azimuthal) term, nonzero only for variable $\boldsymbol{\omega}$ . We state it here to show where the iteration leads; the full derivation — including the bookkeeping that produces the famous factor of 2 — is the core of the acceleration-transformation article, and the sorted physical consequences are the Coriolis, centrifugal, and Euler force articles. One vector is exempt from all this machinery. Set $\mathbf{A} = \boldsymbol{\omega}$ in (1): the cross term is $\boldsymbol{\omega}\times\boldsymbol{\omega} = \mathbf{0}$ , so

\left(\frac{\mathrm{d}\boldsymbol{\omega}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\frac{\mathrm{d}\boldsymbol{\omega}}{\mathrm{d}t}\right)_{\mathrm{rot}} \equiv \dot{\boldsymbol{\omega}} \qquad (6)

where: $\dot{\boldsymbol{\omega}}$ — the angular acceleration, the same vector in both frames. This is why the Euler term in (5) needs no frame subscript — a small but essential consistency of the formalism.

Derivation AThe operator identity from basis-vector differentiation

We are deriving the master identity (1) from nothing but the product rule and the rigid rotation of the basis triad. The strategy: expand an arbitrary vector on the rotating basis, differentiate in the inertial frame — where both the components and the basis vectors change — and identify the two resulting pieces as the rotating-frame derivative and the cross-product term. Along the way we prove the basis kinematics (3) rather than assuming it, using only the constancy of the dot products $\hat{\mathbf{e}}_i\cdot\hat{\mathbf{e}}_j$ . Watch the logic of the final step: because nothing whatsoever is assumed about $\mathbf{A}$ , the result holds as an operator on any vector — the property every downstream derivation relies on.

Step 1. Establish the basis kinematics. The rotating triad stays orthonormal at all times, $\hat{\mathbf{e}}_i\cdot\hat{\mathbf{e}}_j=\delta_{ij}$ , so differentiating gives $\dot{\hat{\mathbf{e}}}_i\cdot\hat{\mathbf{e}}_j+\hat{\mathbf{e}}_i\cdot\dot{\hat{\mathbf{e}}}_j=0$ : the matrix $\Omega_{ij}\equiv\dot{\hat{\mathbf{e}}}_i\cdot\hat{\mathbf{e}}_j$ is antisymmetric. Why antisymmetry is decisive: a $3\times3$ antisymmetric matrix has exactly three independent entries, which package into a pseudovector $\omega_k=-\tfrac12\varepsilon_{ijk}\Omega_{ij}$ , and contracting back gives

\dot{\hat{\mathbf{e}}}_i = \boldsymbol{\omega}\times\hat{\mathbf{e}}_i

— rigid rotation forces the cross-product form; no other motion preserves all lengths and angles of the triad.

Step 2. Expand the arbitrary vector on the rotating basis:

\mathbf{A} = \sum_i A_i\,\hat{\mathbf{e}}_i

Why this basis: the components $A_i$ are precisely the numbers a rotating observer reads off; choosing her axes makes her measurement explicit in the algebra.

Step 3. Differentiate in the inertial frame, where the basis vectors are moving objects, and apply the product rule:

\left(\frac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{in}} = \sum_i \dot{A}_i\,\hat{\mathbf{e}}_i + \sum_i A_i\,\dot{\hat{\mathbf{e}}}_i

Step 4. Identify the first sum. To the rotating observer the basis is fixed, so only component changes register: $\sum_i \dot{A}_i\hat{\mathbf{e}}_i = (\mathrm{d}\mathbf{A}/\mathrm{d}t)_{\mathrm{rot}}$ — by definition, her derivative.

Step 5. Insert the basis kinematics from Step 1 into the second sum and use linearity of the cross product:

\sum_i A_i\,\dot{\hat{\mathbf{e}}}_i = \sum_i A_i\,(\boldsymbol{\omega}\times\hat{\mathbf{e}}_i) = \boldsymbol{\omega}\times\mathbf{A}

Adding the two pieces gives the boxed identity (1). Why it is an operator: no property of $\mathbf{A}$ was used beyond having components — the relation therefore holds for any vector, including one produced by an earlier application of the same rule. *Verified: matches Goldstein §4.9 Eq. (4.86) and Morin Ch. 10; setting ω = 0 collapses both derivatives into one, and a frame-frozen vector recovers Step 1 exactly.*

Go deeper · Derivation B · 1 minConsistency check — scalars must not care about the frame

Derivation BConsistency check — scalars must not care about the frame

The goal is a sanity theorem: the derivative of any scalar built from vectors — a dot product $\mathbf{A}\cdot\mathbf{B}$ , a length, an angle — must come out the same in both frames, because a number has no direction for the rotating basis to confuse. If the identity (1) failed this test it would be internally inconsistent. The strategy is direct: differentiate $\mathbf{A}\cdot\mathbf{B}$ in the inertial frame, insert the identity for each factor, and watch the rotation terms annihilate via the antisymmetry of the scalar triple product. The cancellation mechanism is worth internalizing — it is the same algebra that makes the Coriolis force do no work.

Step 1. Apply the product rule in the inertial frame:

\left(\frac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{in}}(\mathbf{A}\cdot\mathbf{B}) = \dot{\mathbf{A}}_{\mathrm{in}}\cdot\mathbf{B} + \mathbf{A}\cdot\dot{\mathbf{B}}_{\mathrm{in}}

Step 2. Insert the identity (1) for each vector:

= \left(\dot{\mathbf{A}}_{\mathrm{rot}}+\boldsymbol{\omega}\times\mathbf{A}\right)\cdot\mathbf{B} + \mathbf{A}\cdot\left(\dot{\mathbf{B}}_{\mathrm{rot}}+\boldsymbol{\omega}\times\mathbf{B}\right)

Step 3. Collect the two rotation terms and recognize scalar triple products:

(\boldsymbol{\omega}\times\mathbf{A})\cdot\mathbf{B}+\mathbf{A}\cdot(\boldsymbol{\omega}\times\mathbf{B}) = \boldsymbol{\omega}\cdot(\mathbf{A}\times\mathbf{B}) + \boldsymbol{\omega}\cdot(\mathbf{B}\times\mathbf{A}) = 0

Why they cancel: swapping the factors of a cross product flips its sign, so the two triple products are exact negatives. Geometrically, a rigid rotation can change no length and no angle — the algebra is honoring the geometry.

Step 4. What remains is the rotating-frame product rule: scalars evolve identically in both frames, as they must. Why this matters downstream: with $\mathbf{B} = \mathbf{A}$ it says rotation never changes $|\mathbf{A}|$ — a frame-frozen vector keeps its length while sweeping, and forces of the form $\boldsymbol{\omega}\times\mathbf{v}$ (Coriolis) can steer a velocity but never change a speed. *Verified: consistent with Goldstein §4.9 and Morin Ch. 10; the special case B = A gives d|A|²/dt frame-independent, confirming rigid rotation preserves lengths.*

Limiting cases

The identity simplifies in several instructive regimes:

Case	Result	Meaning
Vector fixed in rotating frame	$\left(\tfrac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{in}} = \boldsymbol{\omega}\times\mathbf{A}$	When $(\mathrm{d}\mathbf{A}/\mathrm{d}t)_{\mathrm{rot}}=0$ , the inertial change is pure rotation at rate $\omega$ , perpendicular to both $\boldsymbol{\omega}$ and $\mathbf{A}$ .
Vector parallel to $\boldsymbol{\omega}$	$\boldsymbol{\omega}\times\mathbf{A} = \mathbf{0} \implies \left(\tfrac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\tfrac{\mathrm{d}\mathbf{A}}{\mathrm{d}t}\right)_{\mathrm{rot}}$	Components along the rotation axis are unaffected by the spin; both observers agree on their rate of change.
Non-rotating frame	$\boldsymbol{\omega} = \mathbf{0} \implies \left(\tfrac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\tfrac{\mathrm{d}}{\mathrm{d}t}\right)_{\mathrm{rot}}$	For a purely translating (possibly accelerating) frame the identity is trivial — derivatives coincide. See linearly accelerating frames.
The angular velocity itself	$\left(\tfrac{\mathrm{d}\boldsymbol{\omega}}{\mathrm{d}t}\right)_{\mathrm{in}} = \left(\tfrac{\mathrm{d}\boldsymbol{\omega}}{\mathrm{d}t}\right)_{\mathrm{rot}} = \dot{\boldsymbol{\omega}}$	Because $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ , the angular acceleration is frame-independent — crucial for the Euler force.
Scalars (dot products)	$\tfrac{\mathrm{d}}{\mathrm{d}t}(\mathbf{A}\cdot\mathbf{B})$ same in both frames	Numbers have no direction to rotate; Derivation B is the proof.

Common trap“The rotating derivative is just the projection onto rotating axes”

No — and this misreading is the most common source of wrong signs in the whole subject. $(\mathrm{d}\mathbf{A}/\mathrm{d}t)_{\mathrm{rot}}$ is not a projection of the inertial derivative; it is a different derivative of the same vector: the rate of change a rotating observer measures, obtained by differentiating the components while holding her basis fixed. The two derivatives differ by the full $\boldsymbol{\omega}\times\mathbf{A}$ , which can be enormous even when the rotating observer sees nothing change at all (the chalked arrow). Two adjacent traps: the identity needs the angular velocity of the frame, not of the object being watched — a satellite tracked from a rotating station uses the station's $\boldsymbol{\omega}$ ; and $\boldsymbol{\omega}$ itself is exempt, since $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ makes $\dot{\boldsymbol{\omega}}$ frame-independent.

Worked examples

1 · The chalked arrow, quantified (confidence builder)

A carousel turns at $\omega = 0.5\ \mathrm{rad/s}$ . A chalk arrow of length $A = 2.0\ \mathrm{m}$ is drawn on the floor pointing radially outward. (a) Find the inertial rate of change of the arrow. (b) A child now redraws it continuously so that it also lengthens at $0.30\ \mathrm{m/s}$ . Find the new inertial rate of change.

Step 1. Frozen arrow: the rotating-frame derivative vanishes, so the identity (1) leaves only the sweep term, $|\dot{\mathbf{A}}_{\mathrm{in}}| = |\boldsymbol{\omega}\times\mathbf{A}| = \omega A \sin 90° = 0.5 \times 2.0 = \boxed{1.0\ \mathrm{m/s}}$ , directed tangentially — perpendicular to both the vertical axis and the arrow.

Step 2. Lengthening arrow: now $(\mathrm{d}\mathbf{A}/\mathrm{d}t)_{\mathrm{rot}} = 0.30\ \mathrm{m/s}$ radially, while the sweep term is unchanged at $1.0\ \mathrm{m/s}$ tangentially. The two contributions are perpendicular.

Step 3. Add vectorially: $|\dot{\mathbf{A}}_{\mathrm{in}}| = \sqrt{0.30^2 + 1.0^2} = \boxed{1.04\ \mathrm{m/s}}$ , tilted $\arctan(1.0/0.30) = 73°$ from the radial direction toward the direction of rotation.

Step 4. Check the limits: stop the carousel ( $\omega \to 0$ ) and only the $0.30\ \mathrm{m/s}$ sliding survives; freeze the chalk ( $\dot A \to 0$ ) and the pure sweep $1.0\ \mathrm{m/s}$ returns. The identity interpolates between the two observers' stories.

Figure

Steering an angular momentum in a circle — the identity applied to L gives the gyroscopic torque

2 · The torque that holds a spinning wheel in a turning frame (exam level)

A bicycle wheel with moment of inertia $I = 0.15\ \mathrm{kg\,m^2}$ spins at $80\ \mathrm{rad/s}$ about its axle, which is held horizontal on a platform rotating at $\Omega = 0.8\ \mathrm{rad/s}$ about the vertical. The axle is clamped so the spin magnitude and its orientation relative to the platform never change. What torque must the clamp supply?

Step 1. Identify the vector to differentiate: the spin angular momentum $\mathbf{L} = I\omega_{\mathrm{spin}} = 0.15\times80 = 12\ \mathrm{kg\,m^2/s}$ , horizontal, frozen in the platform frame. This is the identity applied to a vector that is neither a position nor a velocity — the point of the exercise.

Step 2. Apply (1) with $(\mathrm{d}\mathbf{L}/\mathrm{d}t)_{\mathrm{rot}} = 0$ :

\left(\frac{\mathrm{d}\mathbf{L}}{\mathrm{d}t}\right)_{\mathrm{in}} = \boldsymbol{\Omega}\times\mathbf{L}

Step 3. Newton's rotational law holds in the inertial frame: $\boldsymbol{\tau} = (\mathrm{d}\mathbf{L}/\mathrm{d}t)_{\mathrm{in}}$ . With $\boldsymbol{\Omega}$ vertical and $\mathbf{L}$ horizontal, $\sin\theta = 1$ : $\tau = \Omega L = 0.8 \times 12 = \boxed{9.6\ \mathrm{N\,m}}$ , horizontal, perpendicular to the axle, rotating with the platform.

Step 4. Read it physically: the clamp must continuously steer the angular momentum around a circle. A rider turning a corner feels exactly this gyroscopic torque through the handlebars; the same calculation with $\mathbf{L}$ and $\boldsymbol{\Omega}$ swapped in roles underlies gyrocompasses and control-moment gyros on spacecraft.

Step 5. Order-of-magnitude check: $9.6\ \mathrm{N\,m}$ is what a $1\ \mathrm{kg}$ mass exerts at the end of a metre lever — very noticeable to a hand, exactly matching the classroom spinning-wheel demonstration.

Origins · 1 minJoseph-Louis Lagrange and how this was discovered

Joseph-Louis Lagrange

1736–1813 · Turin → Berlin → Paris

Lagrange was the great systematizer of mechanics — his Mécanique analytique (1788) famously contains not a single diagram — and rotating frames were one of his sharpest tools. In his prize-winning 1772 essay on the three-body problem he transformed the Sun–Jupiter system into a frame rotating with the planets, where the bookkeeping of this article turns a hopeless dynamical problem into a search for equilibrium points. Out fell the five now-famous Lagrange points, including the two triangular ones where the fictitious forces of the rotating frame conspire with gravity to hold a third body in place — points that today host the James Webb Space Telescope (L2) and Jupiter's Trojan asteroid swarms (L4, L5). Born in Turin, he succeeded Euler in Berlin, survived the French Revolution in Paris as the Republic's favorite mathematician, and is buried in the Panthéon — under the same dome where Foucault would later hang his pendulum.

History

1743

— d'Alembert's Traité de dynamique recasts accelerated motion as a statics problem, planting the seed of treating frame effects as forces.

1765

— Euler's rigid-body theory introduces body-fixed rotating axes and the equations that govern vectors referred to them.

1772

— Lagrange works the restricted three-body problem in a rotating frame, discovering the equilibrium points that bear his name.

1835

— Coriolis publishes the full acceleration transformation for rotating machines, giving the velocity-dependent term its modern form.

1851

— Foucault's pendulum at the Panthéon makes the rotating-frame bookkeeping of the Earth visible to a crowd.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 10 (Accelerated frames of reference) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §4.9–4.10 · Landau & Lifshitz, Mechanics, 3rd ed., §39 · Taylor, Classical Mechanics, Ch. 9 · Kleppner & Kolenkow, An Introduction to Mechanics, Ch. 9

Provenance. Operator identity and basis-vector derivation follow Goldstein §4.9–4.10, Morin Ch. 10, and Taylor Ch. 9. A verifier should: (1) reconstruct $\dot{\hat{\mathbf{e}}}_i=\boldsymbol{\omega}\times\hat{\mathbf{e}}_i$ from antisymmetry of $\mathrm{d}(\hat{\mathbf{e}}_i\cdot\hat{\mathbf{e}}_j)/\mathrm{d}t=0$ ; (2) re-derive the acceleration transformation by iterating the operator and confirm the factor-of-2 Coriolis term; (3) recompute the equatorial centrifugal acceleration using sidereal $\omega=2\pi/86164\,\text{s}=7.292\times10^{-5}\,\mathrm{rad/s}$ and $R=6.378\times10^6\,\text{m}$ , obtaining $0.0339\ \mathrm{m/s^2}$ ( $\approx0.35\%\,g$ ). All standard textbook values.

Builds on: Angular velocity vector of a rotating frame · Non-inertial reference frames · Vector cross product and basis vectors Leads to: Transformation of velocity between rotating frames · Transformation of acceleration in rotating frames · Coriolis force · Centrifugal force · Euler force · Effective gravity on the rotating Earth

If both observers watch the same vector, how can they honestly measure different derivatives?

Because “derivative” secretly means “rate of change of components against a chosen set of axes,” and the two observers hold different axes fixed. The rotating observer differentiates the components $A_i$ measured against her turning triad; the inertial observer differentiates components against fixed axes. Same vector, two legitimate derivatives — related exactly by $\boldsymbol{\omega}\times\mathbf{A}$ , which is the motion of one set of axes relative to the other. Nothing is subjective about either answer: each is a well-defined vector, and the identity is the dictionary between them. The confusion only arises when an equation silently mixes derivatives from the two dictionaries — which is precisely the error the subscripts prevent.

Is the ω in the identity the angular velocity of the object I am watching, or of my frame?

Of the frame — always. The identity knows nothing about what the vector $\mathbf{A}$ belongs to; it only corrects for the turning of the measuring axes. If you track a satellite from a rotating space station, you use the station's $\boldsymbol{\omega}$ , whatever the satellite is doing. A special case makes this vivid: watch a vector that is itself rotating with some other angular velocity $\boldsymbol{\omega}_2$ from a frame rotating at $\boldsymbol{\omega}_1$ — the identity still inserts $\boldsymbol{\omega}_1\times$ , and the mismatch $(\boldsymbol{\omega}_2-\boldsymbol{\omega}_1)$ shows up inside $(\mathrm{d}\mathbf{A}/\mathrm{d}t)_{\mathrm{rot}}$ , where it belongs.

Why is there no ω × term for scalars like temperature or speed?

A scalar has no components to re-express when the axes turn — it is a single number, identical in every frame at each instant, so its time derivative is frame-independent automatically. Derivation B makes this a theorem for scalars built from vectors: in $\mathrm{d}(\mathbf{A}\cdot\mathbf{B})/\mathrm{d}t$ the two rotation terms cancel through the antisymmetry of the scalar triple product. That cancellation carries real physics: with $\mathbf{B}=\mathbf{A}$ it says rotation alone never changes a vector's length, which is why a frame-frozen vector sweeps a circle rather than spiraling, and why the Coriolis force — built from a cross product with velocity — can change a projectile's direction but never its speed.

Does the identity still hold when ω itself changes with time?

Yes, with no modification — this trips many people up. The derivation used only the instantaneous statement $\dot{\hat{\mathbf{e}}}_i = \boldsymbol{\omega}(t)\times\hat{\mathbf{e}}_i$ , which is valid moment by moment however wildly $\boldsymbol{\omega}$ wanders in magnitude or direction. What changing $\boldsymbol{\omega}$ does affect is the second application: iterating the identity on the velocity produces an extra term $\dot{\boldsymbol{\omega}}\times\mathbf{r}$ — the Euler term of equation (5) — which vanishes for steady rotation. And conveniently, $\dot{\boldsymbol{\omega}}$ needs no frame label, because applying the identity to $\boldsymbol{\omega}$ itself gives $\boldsymbol{\omega}\times\boldsymbol{\omega}=0$ : both observers agree on the angular acceleration.

Where exactly does the factor of 2 in the Coriolis term come from — the identity has only one ω × in it?

From applying the single- $\boldsymbol{\omega}\times$ rule twice, to a quantity that contains $\boldsymbol{\omega}\times$ once already. The first application turns position into $\mathbf{v}_{\mathrm{in}} = \mathbf{v}_{\mathrm{rot}} + \boldsymbol{\omega}\times\mathbf{r}$ . Differentiating that again, one $\boldsymbol{\omega}\times\mathbf{v}_{\mathrm{rot}}$ arises from the rotating-frame derivative of the sweep term $\boldsymbol{\omega}\times\mathbf{r}$ (the particle keeps moving to platform points with different sweep velocities), and a second, equal one from the operator's own $\boldsymbol{\omega}\times$ acting on $\mathbf{v}_{\mathrm{rot}}$ (the velocity's basis vectors turn). Two physically distinct effects, identical algebraic form. The full bookkeeping — and the classic mistake of feeding the operator $\mathbf{v}_{\mathrm{rot}}$ instead of $\mathbf{v}_{\mathrm{in}}$ , which loses the 2 — is laid out in the acceleration-transformation article.

Both frames share an origin here. What changes if the rotating frame's origin also translates?

Then the transformations pick up the translational terms as an independent, additive layer. For an origin at inertial position $\mathbf{R}(t)$ , positions obey $\mathbf{r}_{\mathrm{in}} = \mathbf{R} + \mathbf{r}$ , and differentiating adds $\dot{\mathbf{R}}$ to the velocity transformation and $\ddot{\mathbf{R}}$ to the acceleration transformation — the uniform fictitious force $-m\ddot{\mathbf{R}}$ of linearly accelerating frames, the subject of the equivalence-of-acceleration-and-gravity article. The rotational structure of the identity is untouched, because it concerns the turning of axes, not the motion of the origin. The rotating Earth needs both layers: the frame's origin (Earth's center) is itself in free fall around the Sun, and its axes turn once per sidereal day.

Is the operator identity exact, or an approximation for slow rotation?

Exact — it is pure kinematics, a statement about how components re-express between two sets of axes, with no expansion in $\omega$ anywhere. Feed it a millisecond pulsar's rotation or a lazy carousel's; it does not care. Every approximation in rotating-frame practice enters afterwards and separately: dropping the Euler term because $\dot{\boldsymbol{\omega}} \approx 0$ , dropping centrifugal terms because $\omega^2 r$ is small, linearizing the Coriolis deflection for short flights. Keeping the exactness in mind protects you in strong-rotation problems (centrifuges at $10^4$ rad/s, rotating neutron stars) where those habitual small- $\omega$ discards silently fail.