Kinematics — position, velocity, acceleration

Intuition [Beginner]

Before you can ask why something moves, you need a precise language for describing how it moves. That language is kinematics. It gives you three quantities — position, velocity, and acceleration — each built from the one before, and together they tell the complete story of an object's motion through space over time.

Position answers "where is it?" You pick a fixed point — the corner of a room, the centre of a racetrack, the origin of a coordinate grid — and measure from there. The measurement has a direction (5 metres north-east, not just 5 metres), so position is a vector: an arrow from the origin to the object's current location.

Velocity answers "how fast, and in which direction, is it going?" If the object was at one point a moment ago and is at a different point now, the velocity points from the old position toward the new one, with a length proportional to how quickly the move happened. A car doing 100 km/h north and a car doing 100 km/h south have the same speed but different velocities — the direction matters.

Acceleration answers "how is the velocity changing?" When you press the gas pedal, your velocity's magnitude grows — positive acceleration. When you brake, the magnitude shrinks — but "shrinking velocity" is itself a kind of acceleration, pointing backward. A ball thrown upward slows down, stops, then speeds up going down; throughout, the acceleration points downward. The acceleration arrow does not need to point in the same direction as the velocity arrow, and in most interesting motions it does not.

These three quantities form a chain. Position tells you where. Velocity is the rate of change of position — how position updates from moment to moment. Acceleration is the rate of change of velocity — how velocity updates from moment to moment. The chain runs one way: you differentiate position to get velocity, differentiate velocity to get acceleration. To go backwards, you accumulate: add up (integrate) acceleration over time to get the change in velocity, add up velocity over time to get the change in position.

A crucial subtlety: velocity and acceleration are independent directions. A car going around a circular track at constant speed has a velocity that always points along the track — but an acceleration that points toward the centre of the circle. The speedometer reads a constant, yet the car is accelerating, because the direction of the velocity is changing. "Acceleration" does not mean "speeding up." It means "velocity is changing," and velocity can change in direction without changing in magnitude.

Why kinematics matters for everything that follows: Newton's second law ($F=ma$) connects force to acceleration. You cannot state or use Newton's laws without knowing what acceleration is. Every branch of physics — orbital mechanics, electromagnetism, statistical mechanics — builds on trajectories of particles through space, and those trajectories are described by position, velocity, and acceleration.

Visual [Beginner]

Figure: A particle tracing a curved path through three-dimensional space. Three arrows are drawn at a single instant. The blue arrow starts at the origin and ends at the particle's current location — that is the position vector r. The green arrow is tangent to the path at the particle's location, pointing in the direction of motion — that is the velocity vector v.

The red arrow does not point along the path at all; instead it curves inward toward the concave side of the trajectory — that is the acceleration vector a. Notice that v and a are not parallel; the acceleration is changing the direction of the velocity even though the speed may be constant.

A second instant is shown a short time later: the particle has moved forward along the curve, the blue arrow (position) has shifted, the green arrow (velocity) has rotated slightly, and the red arrow (acceleration) again points inward.

Worked example [Beginner]

A car accelerates uniformly from rest to 100 km/h in 8.0 seconds along a straight road. Take the starting position as zero.

Step 1. Convert the final speed to metres per second: 100 km/h = (100 × 1000) / 3600 = 27.8 m/s.

Step 2. Compute acceleration. The velocity starts at 0 and reaches 27.8 m/s in 8.0 seconds. For constant acceleration, the rate of change of velocity is the same at every instant, so:

acceleration = (final velocity − initial velocity) / time = (27.8 − 0) / 8.0 = 3.47 m/s².

This is a scalar because the motion is along one straight line; the direction is forward.

Step 3. How far does the car travel during those 8 seconds? For constant acceleration from rest, the distance is the average velocity multiplied by the time. The average of 0 and 27.8 m/s is 13.9 m/s, so:

distance = 13.9 × 8.0 = 111 m.

Alternatively, using the kinematic formula for constant acceleration: distance = half × acceleration × time² = 0.5 × 3.47 × 64 = 111 m.

Step 4. What are the velocity and position at the halfway time t = 4.0 s?

velocity at t = 4.0 = 3.47 × 4.0 = 13.9 m/s (exactly half the final velocity).

position at t = 4.0 = 0.5 × 3.47 × 16 = 27.8 m (only a quarter of the total distance, because the car was going slower in the first half).

What this tells us: constant acceleration produces linearly growing velocity and quadratically growing position. The position grows as the square of time — an object at constant acceleration covers only a quarter of the total distance in the first half of the time interval.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $t\mapsto \mathbf{r}(t)\in {\mathbb{R}}^3$ be a twice-differentiable curve parametrised by time $t\in [t_0,t_1]$, representing the position of a point particle in three-dimensional Euclidean space relative to a fixed origin $O$. The curve $\mathbf{r}(t)$ is called the trajectory or world-line of the particle.

The velocity of the particle at time $t$ is

$$\mathbf{v}(t):=\frac{d\mathbf{r}}{dt}(t)=\dot{\mathbf{r}}(t),$$

a vector tangent to the trajectory at $\mathbf{r}(t)$. The speed (or magnitude of velocity) is the scalar

$$v(t):=|\mathbf{v}(t)|=\sqrt{\dot{x}(t{)}^2+\dot{y}(t{)}^2+\dot{z}(t{)}^2},$$

where $\mathbf{r}(t)=(x(t),y(t),z(t))$ in Cartesian coordinates.

The acceleration of the particle at time $t$ is

$$\mathbf{a}(t):=\frac{d\mathbf{v}}{dt}(t)=\frac{d^2\mathbf{r}}{d{t}^2}(t)=\ddot{\mathbf{r}}(t).$$

In components: $\mathbf{a}(t)=(\ddot{x}(t),\ddot{y}(t),\ddot{z}(t))$.

The arc length traversed from $t_0$ to $t$ is

$$s(t):={\int }_{t_0}^t|\mathbf{v}(\tau )|d\tau ,$$

and $\dot{s}(t)=v(t)$, recovering speed as the rate of change of distance along the path.

Resolution of acceleration into tangential and normal components. The unit tangent vector to the trajectory is $\hat{\mathbf{T}}(t)=\mathbf{v}(t)/v(t)$ (for $v\ne 0$). Differentiating $\mathbf{v}=v\hat{\mathbf{T}}$:

$$\mathbf{a}=\dot{v}\hat{\mathbf{T}}+v\dot{\hat{\mathbf{T}}}.$$

The first term, $a_T:=\dot{v}$, is the tangential acceleration — the rate of change of speed. The second term captures the change in direction. Writing $\dot{\hat{\mathbf{T}}}=v\kappa \hat{\mathbf{N}}$ where $\kappa$ is the curvature of the path and $\hat{\mathbf{N}}$ is the principal normal unit vector (pointing toward the centre of curvature), one obtains

$$\mathbf{a}=\dot{v}\hat{\mathbf{T}}+v^2\kappa \hat{\mathbf{N}}=a_T\hat{\mathbf{T}}+\frac{v^2}{\rho }\hat{\mathbf{N}},$$

where $\rho =1/\kappa$ is the radius of curvature. The centripetal acceleration $v^2/\rho$ points inward toward the centre of curvature and is present whenever the path is not a straight line, even at constant speed.

Constant-acceleration kinematics. When $\mathbf{a}$ is constant (e.g., free fall near Earth's surface with $\mathbf{a}=\mathbf{g}\approx -9.8\hat{\mathbf{z}}$ m/s²), integrating $\dot{\mathbf{v}}=\mathbf{a}$ and $\dot{\mathbf{r}}=\mathbf{v}$ yields

$$\mathbf{v}(t)={\mathbf{v}}_0+\mathbf{a}(t-t_0),\mathbf{r}(t)={\mathbf{r}}_0+{\mathbf{v}}_0(t-t_0)+\frac{1}{2}\mathbf{a}(t-t_0{)}^2.$$

Eliminating time between these gives

$$|\mathbf{v}(t){|}^2=|{\mathbf{v}}_0{|}^2+2\mathbf{a}\cdot (\mathbf{r}(t)-{\mathbf{r}}_0).$$

These three equations are the standard kinematic formulae for constant acceleration, valid for projectile motion on a flat Earth in the absence of air resistance.

Counterexamples to common slips

• "Negative acceleration means deceleration." The sign of acceleration is coordinate-dependent. A particle moving in the $-x$ direction with acceleration $\ddot{x}<0$ is speeding up — both velocity and acceleration point the same way. Deceleration means acceleration antiparallel to velocity, not acceleration with a minus sign.

• Conflating speed and velocity. The speed $v=|\mathbf{v}|$ is a non-negative scalar. The velocity $\mathbf{v}$ is a vector. A particle in uniform circular motion has $dv/dt=0$ (constant speed) but $|d\mathbf{v}/dt|=v^2/\rho \ne 0$ (nonzero acceleration). The kinematic formulae above involve $\mathbf{v}$ and $\mathbf{a}$, not their magnitudes alone.

• Assuming acceleration is caused by velocity. Acceleration is the time derivative of velocity — it measures how velocity is already changing, not what caused the change. The cause is force, which belongs to dynamics (unit 09.01.02 pending). Kinematics describes the geometric consequences of a given trajectory without assigning causes.

Key theorem with proof [Intermediate+]

Theorem (Fundamental theorem of kinematics). Let $\mathbf{r}:[t_0,t_1]\to {\mathbb{R}}^3$ be twice continuously differentiable. Then the velocity and position at any time $t\in [t_0,t_1]$ are determined uniquely by the initial position ${\mathbf{r}}_0=\mathbf{r}(t_0)$, the initial velocity ${\mathbf{v}}_0=\dot{\mathbf{r}}(t_0)$, and the acceleration function $\mathbf{a}(t)=\ddot{\mathbf{r}}(t)$ via

$$\mathbf{v}(t)={\mathbf{v}}_0+{\int }_{t_0}^t\mathbf{a}(\tau )d\tau ,\mathbf{r}(t)={\mathbf{r}}_0+{\int }_{t_0}^t\mathbf{v}(\tau )d\tau .$$

Proof. Since $\mathbf{a}(t)=\dot{\mathbf{v}}(t)$ by definition, integrating from $t_0$ to $t$ and applying the fundamental theorem of calculus to each component:

$$\mathbf{v}(t)-\mathbf{v}(t_0)={\int }_{t_0}^t\dot{\mathbf{v}}(\tau )d\tau ={\int }_{t_0}^t\mathbf{a}(\tau )d\tau .$$

Setting $\mathbf{v}(t_0)={\mathbf{v}}_0$ and rearranging gives the first formula. For the second, $\mathbf{v}(t)=\dot{\mathbf{r}}(t)$, so integrating again:

$$\mathbf{r}(t)-\mathbf{r}(t_0)={\int }_{t_0}^t\dot{\mathbf{r}}(\tau )d\tau ={\int }_{t_0}^t\mathbf{v}(\tau )d\tau .$$

Setting $\mathbf{r}(t_0)={\mathbf{r}}_0$ yields the position formula. Uniqueness follows because a continuously differentiable function is uniquely determined by its derivative and initial value (this is the uniqueness part of the Picard-Lindelof theorem 02.12.01 applied to the first-order system $(\dot{\mathbf{r}},\dot{\mathbf{v}})=(\mathbf{v},\mathbf{a}(t))$). ∎

Corollary. For constant acceleration $\mathbf{a}$, the formulae reduce to $\mathbf{v}(t)={\mathbf{v}}_0+\mathbf{a}(t-t_0)$ and $\mathbf{r}(t)={\mathbf{r}}_0+{\mathbf{v}}_0(t-t_0)+\frac{1}{2}\mathbf{a}(t-t_0{)}^2$.

The corollary is the constant-acceleration case used throughout projectile motion and free-fall problems. The general theorem shows that any acceleration profile — constant, time-varying, or position-dependent — determines the full trajectory once two initial conditions (position and velocity) are specified. This is why Newton's second law ($\mathbf{F}=m\mathbf{a}$) is a second-order differential equation: it specifies the acceleration, and the fundamental theorem of kinematics converts acceleration into the full trajectory given ${\mathbf{r}}_0$ and ${\mathbf{v}}_0$.

Bridge. The fundamental theorem of kinematics builds toward 09.01.02 pending Newton's laws, where force determines acceleration and the theorem above converts that specification into a definite trajectory. The foundational reason the theorem works is the Picard-Lindelof uniqueness result from 02.12.01, and this is exactly the bridge between the ODE existence theory and the deterministic structure of Newtonian mechanics: specifying position and velocity at one instant fixes the entire past and future of the particle. The pattern recurs in 09.02.01 pending, where the action principle produces second-order Euler-Lagrange equations whose solution space is likewise parameterised by initial data on $TQ$.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib has the differential calculus of curves in ${\mathbb{R}}^n$ (Analysis.Calculus.Deriv, Analysis.Calculus.FDeriv) and the definition of smooth manifolds with tangent bundles. It does not have named structures for kinematic quantities — "position curve," "velocity as first derivative," "acceleration as second derivative" — nor a formalisation of Galilean spacetime, inertial frames, or the rotating-frame transformation. These would be straightforward to add using existing derivative machinery but are absent. lean_status: none reflects this gap.

Kinematics on manifolds and in Galilean spacetime [Master]

The intermediate-tier treatment assumes $\mathbf{r}(t)\in {\mathbb{R}}^3$ with a fixed origin. The coordinate-free version replaces ${\mathbb{R}}^3$ by a smooth manifold $Q$ (the configuration manifold), and the trajectory by a smooth curve $\gamma :[t_0,t_1]\to Q$. Velocity is the tangent vector $\dot{\gamma }(t)\in T_{\gamma (t)}Q$, and acceleration requires a connection — either the Levi-Civita connection of a Riemannian metric on $Q$ (giving ${\nabla }_{\dot{\gamma }}\dot{\gamma }$, the covariant acceleration) or, in local coordinates, the geodesic equation with Christoffel symbols.

Definition (Kinematics on a manifold). Let $Q$ be a smooth manifold and $\gamma :I\to Q$ a smooth curve. The velocity of $\gamma$ at time $t$ is the tangent vector $\dot{\gamma }(t)\in T_{\gamma (t)}Q$. Given a connection $\nabla$ on $Q$, the acceleration of $\gamma$ at time $t$ is the covariant derivative ${\nabla }_{\dot{\gamma }}\dot{\gamma }\in T_{\gamma (t)}Q$. A curve satisfying ${\nabla }_{\dot{\gamma }}\dot{\gamma }=0$ is a geodesic — the generalisation of "straight-line motion at constant speed" to curved geometry.

In local coordinates $(q^1,\dots ,q^n)$ on $Q$, a curve $\gamma (t)=(q^1(t),\dots ,q^n(t))$ has velocity components ${\dot{q}}^i(t)$ and acceleration components

$${\ddot{q}}^i+{\Gamma }_{jk}^i{\dot{q}}^j{\dot{q}}^k=0\text{(geodesic equation)},$$

where ${\Gamma }_{jk}^i$ are the Christoffel symbols of the connection. For $Q={\mathbb{R}}^3$ with the flat Euclidean connection, all ${\Gamma }_{jk}^i=0$ and this reduces to $\ddot{\mathbf{r}}=0$ — the law of inertia in its kinematic form.

Galilean spacetime. Kinematics in Newtonian mechanics presupposes a four-dimensional structure $(\mathcal{G},t,d)$ where $\mathcal{G}$ is a four-dimensional affine space, $t:\mathcal{G}\to \mathbb{R}$ is an absolute-time function, and $d$ is a Euclidean inner product on the simultaneous hyperplanes $\ker (dt)$. A particle is a curve $\gamma :\mathbb{R}\to \mathcal{G}$ with $t(\gamma (\tau ))=\tau$ (absolute time parametrisation). Velocity and acceleration are defined using the affine structure of $\mathcal{G}$ restricted to each simultaneity slice. The Galilean group — the automorphisms of $(\mathcal{G},t,d)$ — consists of time translations, spatial rotations, spatial translations, and Galilean boosts $\mathbf{r}\to \mathbf{r}+{\mathbf{v}}_{0}t$.

Proposition (Galilean invariance of acceleration). Under a Galilean boost ${\mathbf{r}}^{\prime }=\mathbf{r}+{\mathbf{v}}_{0}t$ with constant ${\mathbf{v}}_0$, the velocity transforms as ${\mathbf{v}}^{\prime }=\mathbf{v}+{\mathbf{v}}_0$ but the acceleration is invariant: ${\mathbf{a}}^{\prime }=\mathbf{a}$.

Proof. Differentiating ${\mathbf{r}}^{\prime }(t)=\mathbf{r}(t)+{\mathbf{v}}_{0}t$ once gives ${\dot{\mathbf{r}}}^{\prime }=\dot{\mathbf{r}}+{\mathbf{v}}_0$, so ${\mathbf{v}}^{\prime }=\mathbf{v}+{\mathbf{v}}_0$. Differentiating again, ${\ddot{\mathbf{r}}}^{\prime }=\ddot{\mathbf{r}}$ since ${\mathbf{v}}_0$ is constant. Hence ${\mathbf{a}}^{\prime }=\mathbf{a}$. $\square$

This framework, due to Klein (1872) and formalised by Cartan (1923), makes precise the distinction between absolute and relative acceleration: acceleration (unlike velocity) is invariant under Galilean boosts, which is why Newton's second law can single out acceleration as the quantity force produces. The Galilean group is a subgroup of the affine automorphisms of $\mathcal{G}$; its Lie algebra is generated by the translations (spatial and temporal), rotations ($\mathfrak{so}(3)$), and boost generators, forming a ten-dimensional Lie algebra isomorphic to the Bargmann algebra central extension.

The Bargmann extension adds a central generator corresponding to mass, and the resulting eleven-dimensional algebra has a natural unitary representation on the Hilbert space of a non-relativistic quantum particle. This algebraic structure explains why mass appears as a central parameter in Newtonian mechanics and why Galilean-invariant quantum systems have superselection rules forbidding superpositions of different mass sectors. The Newton-Cartan formalism of Galilean gravity generalises this flat-spacetime picture to curved spatial metrics and non-flat Galilean connections, recovering Newtonian gravity as a geometric effect analogous to general relativity but within the Galilean rather than Lorentzian signature.

Jet-bundle formulation. A trajectory $\gamma :\mathbb{R}\to Q$ has, at each time $t$, a 1-jet $j_t^1\gamma =(\gamma (t),\dot{\gamma }(t))\in TQ$ and a 2-jet $j_t^2\gamma =(\gamma (t),\dot{\gamma }(t),\ddot{\gamma }(t))$ in the second jet bundle $J^2(\mathbb{R},Q)$. The fundamental theorem of kinematics is the statement that a section of $J^2(\mathbb{R},Q)$ that satisfies the contact structure (the 2-jet is genuinely the second derivative of its underlying curve) is uniquely determined by the 1-jet initial data $(q_0,v_0)$ and the second-order ODE $\ddot{q}=f(q,\dot{q},t)$. This jet-bundle language is the geometric underpinning of Lagrangian and Hamiltonian mechanics: the Lagrangian lives on $TQ$ (the 1-jet space), and the Euler-Lagrange equations select those curves whose 2-jets satisfy the variational contact condition.

Non-inertial frames and fictitious forces. The rotating-frame acceleration formula derived in Exercise 8 has a geometric interpretation: the covariant derivative of the velocity field in the rotating frame picks up connection terms from the non-standard frame bundle structure. The Coriolis, centrifugal, and Euler accelerations are not forces — they are kinematic terms arising from the difference between the inertial connection and the rotating-frame connection. The geodesics of the inertial connection are straight lines (free particles); the geodesics of the rotating-frame connection curve, and the "fictitious forces" are precisely the terms needed to explain those curves within the rotating frame.

The Frenet-Serret apparatus and the fundamental theorem of curves [Master]

The tangential-normal decomposition of acceleration introduced in the Intermediate tier generalises to a full differential-geometric apparatus for space curves, due to Frenet (1847) and Serret (1851). Let $\mathbf{r}:I\to {\mathbb{R}}^3$ be a smooth curve parametrised by arc length $s$ (so $|d\mathbf{r}/ds|=1$ throughout). The Frenet-Serret frame at each point consists of three orthonormal vectors:

$$\hat{\mathbf{T}}(s)=\frac{d\mathbf{r}}{ds},\hat{\mathbf{N}}(s)=\frac{1}{\kappa (s)}\frac{d\hat{\mathbf{T}}}{ds},\hat{\mathbf{B}}(s)=\hat{\mathbf{T}}(s)\times \hat{\mathbf{N}}(s),$$

where $\kappa (s)=|d\hat{\mathbf{T}}/ds|$ is the curvature and $\hat{\mathbf{N}}$ is the principal normal. The vector $\hat{\mathbf{B}}$ is the binormal. The three vectors $(\hat{\mathbf{T}},\hat{\mathbf{N}},\hat{\mathbf{B}})$ form a right-handed orthonormal frame at each point of the curve, called the Frenet-Serret frame or moving frame.

The torsion $\tau (s)$ measures the rate at which the curve twists out of the osculating plane (the plane spanned by $\hat{\mathbf{T}}$ and $\hat{\mathbf{N}}$):

$$\tau (s):=-\hat{\mathbf{N}}(s)\cdot \frac{d\hat{\mathbf{B}}}{ds}.$$

The sign convention varies; the convention here is that positive torsion corresponds to a right-handed helix. The Frenet-Serret equations are the system of nine coupled first-order ODEs governing the evolution of the moving frame:

Theorem (Frenet-Serret equations). Let $\mathbf{r}(s)$ be a smooth space curve parametrised by arc length with $\kappa (s)>0$ everywhere. Then the Frenet-Serret frame satisfies

$$\frac{d}{ds}\left(\begin{array}{c}\hat{\mathbf{T}}\\ \hat{\mathbf{N}}\\ \hat{\mathbf{B}}\end{array}\right)=\left(\begin{array}{ccc}0& \kappa & 0\\ -\kappa & 0& \tau \\ 0& -\tau & 0\end{array}\right)\left(\begin{array}{c}\hat{\mathbf{T}}\\ \hat{\mathbf{N}}\\ \hat{\mathbf{B}}\end{array}\right).$$

Proof. By definition $d\hat{\mathbf{T}}/ds=\kappa \hat{\mathbf{N}}$, which is the first row. For the third row, $d\hat{\mathbf{B}}/ds=d(\hat{\mathbf{T}}\times \hat{\mathbf{N}})/ds=\kappa \hat{\mathbf{N}}\times \hat{\mathbf{N}}+\hat{\mathbf{T}}\times d\hat{\mathbf{N}}/ds=\hat{\mathbf{T}}\times d\hat{\mathbf{N}}/ds$. Since $\hat{\mathbf{N}}$ is a unit vector, $d\hat{\mathbf{N}}/ds$ is perpendicular to $\hat{\mathbf{N}}$ and hence lies in the span of $\hat{\mathbf{T}}$ and $\hat{\mathbf{B}}$.

Write $d\hat{\mathbf{N}}/ds=\alpha \hat{\mathbf{T}}+\beta \hat{\mathbf{B}}$ for some scalars $\alpha ,\beta$. From $\hat{\mathbf{T}}\cdot \hat{\mathbf{N}}=0$, differentiation gives $(d\hat{\mathbf{T}}/ds)\cdot \hat{\mathbf{N}}+\hat{\mathbf{T}}\cdot (d\hat{\mathbf{N}}/ds)=0$, so $\kappa +\alpha =0$, hence $\alpha =-\kappa$. The coefficient $\beta$ is defined to be the torsion: $\beta =\tau$. Therefore $d\hat{\mathbf{N}}/ds=-\kappa \hat{\mathbf{T}}+\tau \hat{\mathbf{B}}$, which is the second row. For the third row, $d\hat{\mathbf{B}}/ds=\hat{\mathbf{T}}\times (-\kappa \hat{\mathbf{T}}+\tau \hat{\mathbf{B}})=\tau \hat{\mathbf{T}}\times \hat{\mathbf{B}}=-\tau \hat{\mathbf{N}}$. $\square$

The coefficient matrix is skew-symmetric, reflecting the fact that the frame evolves by a rotation in $\mathrm{SO}(3)$: the Frenet-Serret frame is a curve in the rotation group, and the curvature and torsion are its angular velocities.

Theorem (Fundamental theorem of curves). Let $\kappa (s)>0$ and $\tau (s)$ be smooth functions on an interval $[0,L]$. Then there exists a smooth curve $\mathbf{r}:[0,L]\to {\mathbb{R}}^3$, unique up to a rigid motion of ${\mathbb{R}}^3$, whose arc-length parametrisation has curvature $\kappa (s)$ and torsion $\tau (s)$.

Proof sketch. The Frenet-Serret equations form a linear first-order ODE system in nine real variables (the three components of each of $\hat{\mathbf{T}},\hat{\mathbf{N}},\hat{\mathbf{B}}$). Given continuous functions $\kappa (s),\tau (s)$, the Picard-Lindelof theorem 02.12.01 guarantees a unique solution for each initial orthonormal frame at $s=0$. The skew-symmetry of the coefficient matrix ensures that the solution remains orthonormal for all $s$: the matrix exponential $\exp ({\int }_0^{s}A(\sigma )d\sigma )$ preserves the inner product. Integrating $\hat{\mathbf{T}}(s)$ recovers the curve: $\mathbf{r}(s)=\mathbf{r}(0)+{\int }_0^s\hat{\mathbf{T}}(\sigma )d\sigma$. Two curves with the same curvature and torsion have Frenet-Serret frames differing by a constant rotation; integrating produces curves differing by a constant rotation and translation — a rigid motion. $\square$

The curvature and torsion are therefore the differential invariants of a space curve: two curves are congruent (related by a rigid motion) if and only if they have the same curvature and torsion functions. This result is the one-dimensional analogue of the Gauss-Bonnet theorem's statement that Gaussian curvature determines the intrinsic geometry of a surface.

For the kinematic applications of this unit, the Frenet-Serret apparatus connects directly to the tangential-normal decomposition: if a particle moves along a curve with speed $v(t)=\dot{s}(t)$, then its acceleration in the Frenet-Serret frame is $\mathbf{a}=\ddot{s}\hat{\mathbf{T}}+\kappa {\dot{s}}^2\hat{\mathbf{N}}$. The tangential component $\ddot{s}=\dot{v}$ measures the rate of change of speed; the normal component $\kappa v^2=v^2/\rho$ measures the rate of turning. The binormal component is always zero for a smooth curve: acceleration always lies in the osculating plane.

A helix provides the canonical example. The curve $\mathbf{r}(s)=(R\cos (s/c),R\sin (s/c),hs/c)$ with $c=\sqrt{R^2+h^2}$ has constant curvature $\kappa =R/(R^2+h^2)$ and constant torsion $\tau =h/(R^2+h^2)$. The ratio $\tau /\kappa =h/R$ is the pitch parameter. A circle ($h=0$) has zero torsion — the curve lies in a plane. A straight line ($R=0$) has zero curvature — the Frenet-Serret frame degenerates. The helix occupies a privileged position in the theory: it is the only curve (up to rigid motion) with both constant curvature and constant torsion, making it the spatial analogue of uniform circular motion. Charged particles in uniform magnetic fields trace helical paths: the magnetic force provides the centripetal acceleration for the circular component while the velocity component along the field direction is unaffected 10.01.01 pending.

Kinematics of rigid bodies and rotating reference frames [Master]

A rigid body is a system of particles whose mutual distances are constrained to remain constant. Its configuration at any instant is specified by the position of a chosen reference point (say the centre of mass $\mathbf{R}(t)$) and the orientation of a body-fixed frame relative to an inertial frame. The orientation is an element of $\mathrm{SO}(3)$, the rotation group. The configuration space of a rigid body is therefore the six-dimensional manifold ${\mathbb{R}}^3\times \mathrm{SO}(3)$, with three translational degrees of freedom and three rotational degrees of freedom.

The angular velocity $\mathbf{\omega }(t)$ of the rigid body is defined as follows. Let $R(t)\in \mathrm{SO}(3)$ be the rotation matrix carrying body-frame coordinates to inertial coordinates. Then $\dot{R}(t)R(t{)}^{-1}$ is a skew-symmetric matrix (since $R{R}^T=I$ implies $\dot{R}{R}^T+R{\dot{R}}^T=0$). The hat map $\hat{\cdot }:{\mathbb{R}}^3\to \mathfrak{so}(3)$ sends a vector $\mathbf{\omega }=({\omega }_1,{\omega }_2,{\omega }_3)$ to the skew matrix

$$\hat{\mathbf{\omega }}=\left(\begin{array}{ccc}0& -{\omega }_3& {\omega }_2\\ {\omega }_3& 0& -{\omega }_1\\ -{\omega }_2& {\omega }_1& 0\end{array}\right).$$

The angular velocity is defined by $\hat{\mathbf{\omega }}(t)=\dot{R}(t)R(t{)}^{-1}$, or equivalently $\dot{R}=\hat{\mathbf{\omega }}R$. The velocity of a point at body-frame position ${\mathbf{r}}_0$ is then

$$\mathbf{v}(t)=\dot{\mathbf{R}}(t)+\mathbf{\omega }(t)\times (R(t){\mathbf{r}}_0).$$

The first term is the translational velocity of the reference point; the second is the rotational velocity, perpendicular to both the angular velocity vector and the position vector relative to the reference point.

Theorem (Rotating-frame kinematics). Let $S_{\mathrm{in}}$ be an inertial frame and $S_{\mathrm{rot}}$ a frame rotating with angular velocity $\mathbf{\omega }(t)$ relative to $S_{\mathrm{in}}$. Denote the time derivative in the inertial frame by $d/dt$ and in the rotating frame by $d_0/dt$. For any vector-valued function $\mathbf{Q}(t)$,

$$\frac{d\mathbf{Q}}{dt}=\frac{d_0\mathbf{Q}}{dt}+\mathbf{\omega }\times \mathbf{Q}.$$

Applying this to the position $\mathbf{r}$ twice yields the acceleration relation

$${\mathbf{a}}_{\mathrm{in}}={\mathbf{a}}_{\mathrm{rot}}+2\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{rot}}+\mathbf{\omega }\times (\mathbf{\omega }\times \mathbf{r})+\dot{\mathbf{\omega }}\times \mathbf{r}.$$

Proof. A vector $\mathbf{Q}$ fixed in the rotating frame (so $d_0\mathbf{Q}/dt=0$) still rotates relative to the inertial frame. In an infinitesimal time $dt$, the change in $\mathbf{Q}$ is $d\mathbf{Q}=\mathbf{\omega }dt\times \mathbf{Q}$, so $d\mathbf{Q}/dt=\mathbf{\omega }\times \mathbf{Q}$. If $\mathbf{Q}$ also changes in the rotating frame, the changes add: $d\mathbf{Q}/dt=d_0\mathbf{Q}/dt+\mathbf{\omega }\times \mathbf{Q}$.

For velocity, apply the operator to $\mathbf{r}$:

$${\mathbf{v}}_{\mathrm{in}}=\frac{d\mathbf{r}}{dt}=\frac{d_0\mathbf{r}}{dt}+\mathbf{\omega }\times \mathbf{r}={\mathbf{v}}_{\mathrm{rot}}+\mathbf{\omega }\times \mathbf{r}.$$

For acceleration, apply the operator to ${\mathbf{v}}_{\mathrm{in}}$:

$${\mathbf{a}}_{\mathrm{in}}=\frac{d{\mathbf{v}}_{\mathrm{in}}}{dt}=\frac{d_0{\mathbf{v}}_{\mathrm{in}}}{dt}+\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{in}}.$$

Substituting ${\mathbf{v}}_{\mathrm{in}}={\mathbf{v}}_{\mathrm{rot}}+\mathbf{\omega }\times \mathbf{r}$:

$$\frac{d_0{\mathbf{v}}_{\mathrm{in}}}{dt}={\mathbf{a}}_{\mathrm{rot}}+\dot{\mathbf{\omega }}\times \mathbf{r}+\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{rot}},$$

using the product rule for $d_0(\mathbf{\omega }\times \mathbf{r})/dt$. The second term is:

$$\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{in}}=\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{rot}}+\mathbf{\omega }\times (\mathbf{\omega }\times \mathbf{r}).$$

Combining and collecting terms yields the four-term acceleration formula stated above. $\square$

The four terms in the acceleration relation have distinct physical roles. The Coriolis acceleration $2\mathbf{\omega }\times {\mathbf{v}}_{\mathrm{rot}}$ is present only when the particle moves in the rotating frame; it deflects the trajectory sideways, producing the characteristic clockwise spiral of large-scale weather systems in the northern hemisphere. The centrifugal acceleration $\mathbf{\omega }\times (\mathbf{\omega }\times \mathbf{r})$ points radially outward from the rotation axis; its magnitude is ${\omega }^2{r}_{\perp }$ where $r_{\perp }$ is the perpendicular distance from the axis. The Euler acceleration $\dot{\mathbf{\omega }}\times \mathbf{r}$ is nonzero only when the angular velocity is changing, as during spin-up or spin-down of a rotating platform.

Euler angles $(\varphi ,\theta ,\psi )$ parametrise $\mathrm{SO}(3)$ by three successive rotations: first about the $z$-axis by $\varphi$, then about the new $x$-axis by $\theta$, then about the new $z$-axis by $\psi$. The angular velocity in terms of Euler angle rates is

$$\mathbf{\omega }=\dot{\varphi }\hat{\mathbf{z}}+\dot{\theta }{\hat{\mathbf{x}}}^{\prime }+\dot{\psi }{\hat{\mathbf{z}}}^{\prime \prime },$$

where hatted vectors refer to the successive rotated axes. This parametrisation exhibits gimbal lock at $\theta =0$ or $\theta =\pi$, where the first and third rotation axes coincide and one degree of freedom is lost. The singularity is topological: $\mathrm{SO}(3)$ has no global coordinate chart (it is diffeomorphic to ${\mathbb{RP}}^3$, which is not parallelisable). Unit quaternions (equivalently, elements of $\mathrm{SU}(2)$) provide a singularity-free parametrisation at the cost of a double cover 07.07.01.

Geometric structure of configuration space and phase space [Master]

The configuration manifold $Q$ of a mechanical system carries a natural geometric structure that makes the kinematic chain $\text{position}\to \text{velocity}\to \text{acceleration}$ into a statement about the geometry of its tangent bundle $TQ$. This perspective, due to Lagrange (1788) ^{[Lagrange 1788]} and formalised geometrically by Arnold (1974), unifies the kinematics of all mechanical systems — particles, rigid bodies, linked mechanisms — under a single differential-geometric framework.

The tangent bundle as velocity phase space. For a configuration manifold $Q$ of dimension $n$, the tangent bundle $TQ$ is a $2n$-dimensional manifold whose points are pairs $(q,v)$ with $q\in Q$ and $v\in T_{q}Q$. A trajectory $\gamma :I\to Q$ lifts to a curve $\dot{\gamma }:I\to TQ$ by $\dot{\gamma }(t)=(\gamma (t),\dot{\gamma }(t))$. The lifted curve lives in the velocity phase space and encodes the complete kinematic state (position and velocity) at each instant.

The fundamental theorem of kinematics on manifolds is the statement that the acceleration ${\nabla }_{\dot{\gamma }}\dot{\gamma }$ is the vertical component of the derivative of the lifted curve $\dot{\gamma }$ in $TQ$ with respect to the connection induced by $\nabla$ on $TTQ$. Concretely, a curve $\Gamma :I\to TQ$ with $\Gamma (t)=(\gamma (t),v(t))$ is the velocity lift of some configuration-space curve if and only if the contact condition $v(t)=\dot{\gamma }(t)$ is satisfied. The set of all such pairs $(\gamma ,\dot{\gamma })$ is the first jet prolongation $J^1(\mathbb{R},Q)\cong \mathbb{R}\times TQ$, sitting inside $\mathbb{R}\times TQ$ as a submanifold defined by the contact structure.

The Legendre transform and momentum phase space. Given a Lagrangian function $L:TQ\to \mathbb{R}$ (typically $L=T-V$ with $T$ the kinetic energy and $V$ the potential), the Legendre transform $\mathbb{F}L:TQ\to T^{\ast }Q$ sends $(q,v)$ to $(q,p)$ where $p_i=\partial L/\partial {\dot{q}}^i$ are the conjugate momenta. The codomain $T^{\ast }Q$ is the momentum phase space — a $2n$-dimensional manifold with a canonical symplectic structure $\omega =d{p}_i\wedge d{q}^i$ that makes Hamiltonian mechanics possible.

For the simplest case of a single particle in ${\mathbb{R}}^3$ with $L=\frac{1}{2}m|\dot{\mathbf{r}}{|}^2-V(\mathbf{r})$, the Legendre transform is $\mathbf{p}=m\dot{\mathbf{r}}$ — the familiar momentum equals mass times velocity. The phase space is $T^{\ast }{\mathbb{R}}^3\cong {\mathbb{R}}^6$ with coordinates $(\mathbf{r},\mathbf{p})$, and the symplectic form is $\omega =d{p}_x\wedge dx+d{p}_y\wedge dy+d{p}_z\wedge dz$. The kinematic trajectory $\mathbf{r}(t)$ lifts to a phase-space trajectory $(\mathbf{r}(t),\mathbf{p}(t))$ that is a flow of the Hamiltonian vector field $X_H$ defined by ${\iota }_{X_H}\omega =dH$, where $H(\mathbf{r},\mathbf{p})=|\mathbf{p}{|}^2/(2m)+V(\mathbf{r})$ is the Hamiltonian.

Theorem (Liouville's theorem — kinematic prelude). The flow ${\Phi }_t$ of a Hamiltonian vector field on $(T^{\ast }Q,\omega )$ preserves the symplectic form: ${\Phi }_t^{\ast }\omega =\omega$ for all $t$. In particular, the flow preserves the phase-space volume element ${\omega }^n/n!$.

This is a kinematic theorem in the sense that it constrains the geometry of trajectories in phase space without reference to the specific form of the forces. The content is that phase-space volume is incompressible: a region of initial conditions may deform as it evolves, but its volume remains constant. This constrains the long-time behaviour of kinematic trajectories and is the foundation of statistical mechanics 11.01.01 pending.

The symplectic structure and kinematics. The relationship between $TQ$ and $T^{\ast }Q$ is the geometric backbone of mechanics. The Lagrangian formulation works on $TQ$ with the Euler-Lagrange equations as second-order ODEs; the Hamiltonian formulation works on $T^{\ast }Q$ with Hamilton's equations as first-order ODEs. The kinematic chain is:

$$Q\stackrel{\text{lift}}{\to }TQ\stackrel{\text{Legendre}}{\to }{T}^{\ast }Q\stackrel{\text{flow}}{\to }{T}^{\ast }Q.$$

The first arrow is the kinematic lift (position to velocity). The second is the Legendre transform (velocity to momentum). The third is the Hamiltonian flow (time evolution). The acceleration does not appear explicitly in the Hamiltonian picture — it has been absorbed into the first-order structure via the doubling of dimension from $Q$ to $T^{\ast }Q$. This is the central insight behind the geometric formulation of mechanics: the second-order nature of Newton's equation $\mathbf{F}=m\mathbf{a}$ is an artifact of working on $Q$ instead of $TQ$ or $T^{\ast }Q$.

Synthesis. The four Master-tier sections form a coherent arc. Kinematics on manifolds generalises the position-velocity-acceleration chain to arbitrary configuration spaces, and this is exactly the bridge to Lagrangian mechanics 09.02.01 pending, where the configuration manifold and its tangent bundle become the primary objects. The foundational reason the Frenet-Serret apparatus works is that curvature and torsion are the differential invariants of $\mathrm{SO}(3)$ acting on curves, identifying the geometry of trajectories with the representation theory of the rotation group. The rigid-body and rotating-frame kinematics generalise the single-particle picture and build toward 09.01.03 conservation laws, where the angular velocity $\mathbf{\omega }$ and the moment of inertia combine to give angular momentum. The symplectic structure of $T^{\ast }Q$ appears again in 09.02.01 pending as the geometric substrate for Hamilton's equations, and the pattern generalises to field theory, where the configuration space becomes infinite-dimensional and the tangent bundle becomes a bundle over the space of fields.

Connections [Master]

• Phase space and integral curves 02.12.01 is the ODE-theoretic foundation: a trajectory $\mathbf{r}(t)$ with its velocity $\mathbf{v}(t)$ defines an integral curve of a vector field on the phase space $(\mathbf{r},\mathbf{v})$.

• Newton's laws of motion 09.01.02 pending take kinematics as input and supply dynamics: force determines acceleration, and the fundamental theorem of kinematics converts acceleration into trajectory.

• Conservation laws 09.01.03 follow from Newton's laws applied to kinematic quantities: conservation of momentum from $\dot{\mathbf{p}}=0$, conservation of energy from the work-energy theorem that relates force dot displacement to changes in kinetic energy $\frac{1}{2}m{v}^2$.

• The action principle 09.02.01 pending reformulates the trajectory $\mathbf{r}(t)$ as the curve that extremises the action functional, subsuming the kinematic description into a variational principle on the tangent bundle $TQ$.

• Coulomb's law 10.01.01 pending uses kinematic trajectories for charged particles in electric fields — the circular orbits of the Bohr model are constant-acceleration (centripetal) motion at the kinematic level.

• Biomechanics 18.04.01 describes limb motion using joint-angle kinematics; the tangential-normal decomposition of acceleration is the standard tool for analysing joint torques.

Historical & philosophical context [Master]

The study of motion as a quantitative subject begins with Galileo Galilei's experiments on inclined planes, published in Discorsi e dimostrazioni matematiche intorno a due nuove scienze (1638) ^{[Galilei 1638]}. Galileo established that the distance fallen under constant acceleration grows as the square of the elapsed time — the kinematic relation $s=\frac{1}{2}a{t}^2$ — and that the horizontal and vertical components of projectile motion are independent. The conceptual move from Aristotelian "natural motion requires a cause" to Galilean "motion persists unless interfered with" is the intellectual breakpoint that makes kinematics a subject in its own right.

Rene Descartes' La Geometrie (1637, appendix to Discours de la methode) introduced coordinate geometry, giving algebraic meaning to curves in space and making it possible to write $\mathbf{r}(t)=(x(t),y(t),z(t))$ as a set of coordinate functions. Newton and Leibniz independently developed the calculus (Newton 1665-66, De methodis serierum et fluxionum; Leibniz 1675, published 1684), providing the formal tools — the derivative as instantaneous rate of change and the integral as accumulation — that make velocity and acceleration precise rather than merely intuitive.

The vector treatment of kinematics — position, velocity, and acceleration as vectors in three-dimensional Euclidean space — was standardised by Gibbs and Heaviside in the 1880s-90s, extracting the vector algebra from Grassmann and Hamilton's quaternionic frameworks. The decomposition of acceleration into tangential and normal components appears in the differential geometry of curves, formalised by Frenet and Serret in the 1840s-50s (the Frenet-Serret frame). The Galilean-spacetime viewpoint is due to Cartan (Sur les varietes a connexion affine et la theorie de la relativite generalisee, 1923-25) ^{[Cartan 1923]}, who observed that Newtonian mechanics has a natural geometric formulation as a curved affine connection on a flat Galilean manifold. The symplectic-geometric reformulation of configuration space and phase space is due to Arnold (1974) ^{[Arnold 1974]}, building on Poincare's qualitative dynamics and the earlier Hamilton-Jacobi theory.

Philosophically, kinematics raises a question that persists through all of physics: is motion in space and time, or are space and time constituted by the relations among material bodies? The Newtonian (substantivalist) answer — space and time are real entities in which bodies are placed — is built into the language of position vectors and time parameters. The Leibnizian (relationist) alternative — only the distances and temporal intervals between bodies are real — resurfaces in Mach's principle and in the relational dynamics of Barbour and Bertotti (1982) ^{[Barbour 1982]}. This unit adopts the Newtonian convention without taking a stand; the philosophical questions belong to unit 20.04.01 pending (pending) on the philosophy of space and time.

Bibliography [Master]

• Galilei, G., Discorsi e dimostrazioni matematiche intorno a due nuove scienze (1638). [Originator of quantitative kinematics.]
• Descartes, R., La Geometrie (1637).
• Newton, I., Philosophiae Naturalis Principia Mathematica (1687), Book I, Definitions and Scholium.
• Leibniz, G. W., "Nova methodus pro maximis et minimis" (1684).
• Frenet, F., "Sur les courbes a double courbure" (1847); Serret, J.-A., "Sur quelques formules relatives a la theorie des courbes a double courbure" (1851).
• Gibbs, J. W., Vector Analysis (1901, posthumous, ed. E. B. Wilson).
• Cartan, E., "Sur les varietes a connexion affine et la theorie de la relativite generalisee", Ann. Ec. Norm. 40 (1923), 325-412; 41 (1924), 1-25.
• Taylor, J. R., Classical Mechanics (University Science Books, 2005), Ch. 1.
• Susskind, L. & Hrabovsky, G., The Theoretical Minimum: Classical Mechanics (Basic Books, 2014), Lecture 1.
• Arnold, V. I., Mathematical Methods of Classical Mechanics, 2nd ed. (Springer GTM 60, 1989).
• Landau, L. D. & Lifshitz, E. M., Mechanics, 3rd ed. (Course of Theoretical Physics Vol. 1, Pergamon, 1976).
• Tong, D., Classical Dynamics (DAMTP Cambridge lecture notes), §1.
• Barbour, J. & Bertotti, B., "Mach's principle and the structure of dynamical theories", Proc. Roy. Soc. Lond. A 382 (1982), 295-306.
• Lagrange, J.-L., Mechanique analytique (1788). [Originator of the configuration-space formulation.]
• Arnold, V. I., "Sur la geometrie differentielle des groupes de Lie de dimension infinie et ses applications a l'hydrodynamique des fluides parfaits", Ann. Inst. Fourier 16 (1966), 319-361.