Degrees of Freedom

How many numbers does it take to say exactly where something is? For a single speck of dust, three. For the industrial robot arm welding a car body, six — which is precisely why the classic robot arm has six joints. For a spinning satellite, six again. For a pendulum clock's swinging bob: one. And for the $10^{23}$ -atom coffee mug on your desk — a system whose honest Cartesian description needs some $3\times10^{23}$ coordinates — the astonishing answer is still just six, because rigidity has quietly cancelled all the rest. That count — the number of independent quantities that genuinely remain free once every constraint has spoken — is the number of degrees of freedom, and finding it is the mandatory first move of every mechanics problem. The rule itself looks almost too simple: start from $3N$ and subtract one for each independent holonomic constraint, $f = 3N - k$ . The craft is in the subtractions: constraints hide (rigidity is millions of them), constraints overlap (a rigid body's distance relations are massively redundant), and some constraints — the rolling, non-integrable kind classified in the constraints article — restrict motion without removing any coordinate at all, which is why a car needs four numbers to describe but only two to drive. The payoff for counting correctly radiates in every direction. It fixes how many generalized coordinates and equations of motion you will write. It counts a molecule's vibrational modes and thereby predicts heat capacities and infrared spectra — including the single bending mode of CO $_2$ that runs the greenhouse effect. And it tells the roboticist why six is the magic number of joints for placing a tool anywhere, any way up. One integer, computed before any dynamics, that shapes everything downstream.

Key takeaways

The number of degrees of freedom (DOF) is the minimal count of independent coordinates needed to specify a system's configuration completely.
For $N$ particles with $k$ independent holonomic constraints: $f = 3N - k$ . Independence matters — a rigid body's $\sim N^2/2$ distance constraints collapse to leave exactly $f = 6$ , for any $N$ .
The counting ladder: free particle 3 · two particles joined by a rod 5 · rigid triangle 6 · any rigid body 6 · rigid body confined to a plane 3 · double pendulum 2.
Molecules budget their $3N$ as translation + rotation + vibration: $3N-6$ vibrational modes (bent/nonlinear), $3N-5$ (linear) — water has 3, CO $_2$ has 4.
Nonholonomic constraints do NOT reduce the coordinate count: a rolling disk keeps 4 configuration coordinates while having only 2 independent velocity freedoms.
Mechanism mobility is DOF counting with rigid links: the Grübler formula gives the four-bar linkage exactly 1 DOF and a spatial six-revolute arm exactly 6.
Key idea: start from $3N$ Cartesian coordinates and subtract one for each independent holonomic constraint: $f = 3N - k$ . The surviving $f$ numbers are the generalized coordinates, and $f$ is the dimension of configuration space.

Where this lives today

Every robot in every factory is a walking DOF ledger. Six-axis arms remain the industrial default because a rigid tool needs six freedoms; seven-axis collaborative arms buy redundancy for obstacle avoidance; and the humanoid platforms now walking out of half a dozen labs carry 28–40 actuated freedoms whose coordination is learned in physics simulators before touching hardware. The counting shows up before the learning does: the policy network's action space is the mechanism's DOF count, and reinforcement-learning throughput is priced per degree of freedom per timestep. At the other end of the size scale, protein-folding and drug-docking pipelines count conformational freedoms — every rotatable bond is a torsional DOF, and docking cost grows brutally with each one — so "reduce the effective DOF" is as live an engineering strategy in computational chemistry as it is in robot design. The molecular budget of equation (5), meanwhile, runs the planet's energy balance. CO $_2$ 's four modes include a doubly degenerate bend at 667 cm $^{-1}$ — a 15 μm photon, sitting squarely on the peak of Earth's thermal emission spectrum. That one counted mode, times some $10^{44}$ molecules, is the mechanical origin of the greenhouse effect and the reason a linear triatomic trace gas can steer a climate: the freedom budget of a three-atom molecule, set by nothing deeper than $3N - 5$ , decides where the atmosphere is opaque.

The physical picture

Think of degrees of freedom as the number of knobs you can turn independently. A bead threaded on a curved wire has one knob — how far along the wire it sits — no matter how elaborately the wire bends through three-dimensional space. The wire is two constraints pinning the bead to a one-dimensional curve: $f = 3 - 2 = 1$ . A fly in the room has three knobs. A rigid mug has six: three sliders to place its center of mass, three dials to set its orientation. The count is the honest dimension of the space of possible arrangements — the configuration space whose geometry is developed in the generalized coordinates article. The counting ladder below is worth climbing slowly, because it contains the single most important non-obvious fact in the subject: adding matter does not keep adding freedom. One particle: 3. Rigidly join a second: the connecting rod is one constraint, $6 - 1 = 5$ (the pair can point anywhere, but spinning about its own axis changes nothing — a line has no visible roll). Rigidly attach a third, off the line, with two more rods: $9 - 3 = 6$ . And now the ladder tops out: every further particle you weld on brings three new coordinates and three new independent constraints — its distances to three already-placed anchors — for a net gain of zero. Six forever.

Figure

The counting ladder — freedom saturates at six no matter how much matter you add

The same budget-keeping, applied to a molecule, becomes experimental physics. A diatomic molecule is two atoms: $3N = 6$ total freedoms. Spend three on translating the center of mass, two on pointing the axis (rotation about the axis moves nothing, exactly like the rod above), and precisely one remains: the bond-stretching vibration. That $3 + 2 + 1$ budget is directly visible in heat capacities and infrared spectra — and its temperature-dependent surprises (frozen vibrations) were the first crack through which quantum mechanics entered.

Figure

A diatomic molecule spends its six freedoms — three translations, two rotations, one vibration

One caution before the formulas, and it is the deepest one: the subtraction rule counts holonomic constraints only. A rolling wheel's no-slip conditions restrict its velocities without removing a single configuration coordinate — the wheel can still reach every position, heading, and spin. Such systems carry two honest counts at once (coordinates versus controls), and conflating them is the classic blunder the last derivation below untangles.

The counting rule and its fine print

The master formula deserves a precise statement, because all its power is in the adjectives. For $N$ point particles subject to $k$ constraints that are (i) holonomic — equations among positions and possibly time — and (ii) mutually independent:

f = 3N - k \qquad (1)

where: $f$ — the number of degrees of freedom, equal to the dimension of the configuration space; $3N$ — the unconstrained Cartesian count; $k$ — the number of independent holonomic constraints. Each such constraint confines the system to a hypersurface one dimension thinner; $k$ independent ones intersect in a surface of dimension $3N - k$ , and $f$ is simply that dimension. The constraints being subtracted have the standard form

g_\alpha(\mathbf{r}_1,\dots,\mathbf{r}_N,t)=0,\qquad \alpha=1,\dots,k \qquad (2)

where: $g_\alpha$ — the $\alpha$ -th constraint function and $t$ — time, whose presence (a rheonomic constraint, like the rotating wire of the constraints article) changes the energy bookkeeping but not the count: a moving wall still walls off one dimension. What "independent" means is also precise — the constraint gradients must be linearly independent, i.e. the Jacobian of the $g$ 's must have full rank:

\operatorname{rank}\left(\frac{\partial g_\alpha}{\partial x_j}\right)=k \qquad (3)

where: $x_j$ runs over all $3N$ Cartesian coordinates. This rank condition is not lawyer's fine print; it is load-bearing. A rigid body of $N$ particles has $N(N-1)/2$ pairwise-distance constraints — over half a million equations for a mere thousand particles, vastly exceeding $3N$ — and blind subtraction would predict a negative $f$ . The rank of that huge system is only $3N - 6$ : the constraints are massively redundant, and Derivation A shows why exactly six freedoms always survive. Once the count is known, the $f$ surviving freedoms get names — generalized coordinates $q_1,\dots,q_f$ — and the Cartesian positions become functions of them:

\mathbf{r}_i = \mathbf{r}_i(q_1,\dots,q_f,\,t),\qquad i=1,\dots,N \qquad (4)

where: the parametrization satisfies every constraint identically, which is what banishes constraint forces from the Lagrangian formalism. The count $f$ is exactly the number of Euler–Lagrange equations you will solve. Choosing the $q$ 's well is its own craft, developed in the generalized coordinates article; this article's job is only the integer $f$ . For molecules the count is traditionally quoted as a budget: from the $3N$ total, remove rigid-whole motions and what remains vibrates,

f_{\text{vib}} = 3N - 6 \ \ (\text{nonlinear}),\qquad f_{\text{vib}} = 3N - 5 \ \ (\text{linear}) \qquad (5)

where: the 6 (or 5) removed are 3 translations plus 3 rotations — only 2 rotations for a linear molecule, whose configuration is genuinely unchanged by spinning about its own axis. Water ( $N=3$ , bent): $9 - 6 = 3$ vibrational modes. Carbon dioxide ( $N=3$ , linear): $9 - 5 = 4$ modes. That single unit of difference between the two most important greenhouse molecules is pure geometry, and it is measurable in any infrared spectrometer. Engineers run the same arithmetic on machines, where the "particles" are rigid links (6 DOF each in space, 3 in the plane) and the constraints are joints. For a planar mechanism of $n$ links (counting the fixed ground link) connected by $j$ one-freedom joints (hinges or sliders), the mobility is

F = 3(n-1) - 2j \qquad (6)

where: $3(n-1)$ — the planar freedoms of the moving links; each hinge welds two links together at a point and removes 2 of the 3 relative planar freedoms. This is the Grübler criterion, derived and exercised below; its spatial sibling replaces 3 by 6 and 2 by $(6 - f_{\text{joint}})$ . A four-bar linkage scores exactly 1 — one crank angle drives the whole machine — and a six-revolute robot arm scores exactly 6.

Which count answers which question

you want	count	notes
how many generalized coordinates / equations of motion	$f = 3N - k$ , eq. (1)	independent holonomic constraints only
dimension of configuration space	same $f$	by definition
a molecule's vibrational modes	$3N-6$ or $3N-5$ , eq. (5)	linear molecules keep only 2 rotations
a mechanism's mobility	Grübler, eq. (6)	links and joints instead of particles
independent velocity directions	$f$ minus the number of nonholonomic constraints	coordinates are NOT reduced — see Derivation C
equipartition energy per mode	$\tfrac12 k_B T$ per quadratic term	vibrations carry two quadratic terms (KE + PE)

Derivation AWhy every rigid body has exactly six degrees of freedom

We are proving the headline of the counting ladder: the $N(N-1)/2$ pairwise-distance constraints of a rigid body — however astronomically many — always leave exactly $f = 6$ , independent of $N$ . The naive subtraction $3N - N(N-1)/2$ fails spectacularly (it goes negative for $N \geq 8$ ), so the strategy is to count constructively instead: place the particles one at a time, tallying the freedoms each placement genuinely retains, and watch the tally saturate. The step to watch is the fourth particle — the first one whose every coordinate is already spoken for — because it exposes exactly which distance constraints are independent and which are redundant.

Step 1. Place particle 1: nothing constrains it, so it contributes 3 freedoms. Running total: 3.

Step 2. Place particle 2: it must sit at fixed distance $c_{12}$ from particle 1 — one constraint — so it roams a sphere: 2 freedoms. Running total: 5. Why this matches the rod: two particles joined rigidly form the dumbbell of the ladder, and 5 is its count — a line segment has no distinguishable spin about its own axis.

Step 3. Place particle 3 (non-collinear): distances to particles 1 and 2 are both fixed — two constraints — leaving it a circle about the 1–2 axis: 1 freedom. Running total: $3 + 2 + 1 = 6$ .

Step 4. Place particle 4, and every particle after it: its distances to the three non-collinear anchors 1, 2, 3 are fixed — three independent constraints on its three coordinates — pinning it completely (up to a reflection, fixed once and for all by the body's handedness). It adds $3 - 3 = 0$ freedoms. Why the remaining distance constraints are redundant: with particle 4 already pinned by three distances, its distances to particles 5, 6, … are automatic consequences of rigidity — equations already implied, contributing nothing to the Jacobian's rank. This is where the $N(N-1)/2$ collapses.

Step 5. Read off the total and its anatomy. The tally is 6 for every $N \geq 3$ non-collinear particles: $f_{\text{rigid body}} = 6$ . The six split naturally as 3 + 3 — place one marked point (3 translations), then orient the body about it (3 rotations, e.g. Euler angles). Why the split is legitimate: the constructive count placed particle 1 freely (translation) and then distributed the remaining $2 + 1$ freedoms among directions about it, which is exactly an orientation.

Step 6. State the rank result the construction proved: of the $N(N-1)/2$ distance constraints, exactly $3N - 6$ are independent — the rank of the constraint Jacobian — and $f = 3N - (3N - 6) = 6$ . Equation (1) survives, provided $k$ counts rank, not equations. Verified: reproduces Goldstein §1.3 and Landau–Lifshitz §1 (rigid body: six degrees of freedom, three translational plus three rotational); the constructive 3+2+1 count is the standard argument.

Go deeper · Derivation B · 2 minThe Gruebler count for planar mechanisms

Derivation BThe Gruebler count for planar mechanisms

The goal is the engineer's version of equation (1): the mobility formula $F = 3(n-1) - 2j$ for planar linkages, followed immediately by its two most famous verdicts — the four-bar linkage has exactly one degree of freedom, and a triangle of links has none. The strategy is the same subtraction as always, but with rigid links as the primitive objects (3 planar freedoms each, courtesy of Derivation A restricted to the plane) and joints as the constraints. Watch what a single hinge actually removes: not three freedoms but two — the two relative displacement components — because the relative rotation stays free.

Step 1. Count the raw freedoms. A mechanism has $n$ rigid links, one of which is the fixed ground (the machine's frame). Each moving link, as a rigid body confined to the plane, carries 3 freedoms ( $x, y, \varphi$ ), giving $3(n-1)$ in total. Why the plane gives 3: of the six rigid-body freedoms, confinement to a plane removes one translation and two rotations — tilting out of the plane is forbidden.

Step 2. Count what a joint removes. A revolute joint (hinge) forces a point of link A to coincide with a point of link B: two scalar equations ( $x$ -match and $y$ -match), removing 2 freedoms while leaving the relative angle free. A prismatic joint (slider) equally removes 2 (one translation direction and the relative rotation survive as a single freedom). Why not 3: a joint that removed all three relative freedoms would be a weld, and the two links would count as one.

Step 3. Subtract. With $j$ such one-freedom joints:

F = 3(n-1) - 2j

Step 4. Try the triangle: $n = 3$ links pinned in a loop by $j = 3$ hinges gives $F = 3(2) - 6 = 0$ . Zero freedoms — a structure, not a mechanism. This is precisely why triangles brace bridges: the shape is already fully determined. Why this is the planar shadow of rigidity: three links hinged in a triangle is the two-dimensional analogue of Derivation A's rigid triangle of particles.

Step 5. Try the four-bar: $n = 4$ (ground, crank, coupler, rocker), $j = 4$ hinges: $F = 3(3) - 8 = 1$ . One single degree of freedom: turn the crank and every other link's motion is completely determined. This is the reason the four-bar is the workhorse of machine design — windshield wipers, bicycle brakes, piston engines (as the slider-crank variant with $n = 4$ , $j = 4$ : also $F = 1$ ) — one motor, one motion. Adding one more link and hinge (five-bar: $n=5$ , $j=5$ ) gives $F = 12 - 10 = 2$ : two motors needed, the architecture of many pick-and-place robots.

Step 6. State the spatial sibling used for robot arms: $F = 6(n-1) - \sum_i (6 - f_i)$ over joints permitting $f_i$ relative freedoms. A serial arm with six revolute joints ( $f_i = 1$ ) and seven links counting ground: $F = 6(6) - 6\times 5 = 6$ — exactly the six freedoms a rigid tool needs. Why the formula can lie: Grübler counting assumes generic geometry; deliberately special proportions (parallel mechanisms, aligned axes) can hide redundant constraints and move with $F \leq 0$ on paper. The count is a rank condition in disguise, and special geometry can lower the rank. *Verified: Grübler mobility $F = 3(n-1)-2j$ and the four-bar/slider-crank result $F = 1$ are standard results of kinematics of machinery; the spatial 6R arm count $F=6$ matches standard robotics references.*

Go deeper · Derivation C · 2 minCoordinates versus controls for a rolling wheel

Derivation CCoordinates versus controls for a rolling wheel

This derivation untangles the one place where naive DOF counting reliably fails: rolling. The claim to establish — an upright disk rolling on a plane has a full four-dimensional configuration space yet only two independent velocity freedoms, and both numbers are physically meaningful. The strategy is to count configurations and velocities separately and then show, by exhibiting an explicit maneuver, that the velocity restrictions do not shrink the reachable configurations. The step to watch is the integrability question: only an integrable (holonomic) velocity constraint may be subtracted from the coordinate count, and rolling fails that test.

Step 1. Count the configuration. To state completely "where the disk is" requires the contact point $(x, y)$ , the heading angle $\theta$ of the disk's plane, and the spin angle $\phi$ marking which rim point touches the ground: 4 coordinates. No equation among these four is imposed by rolling — any $(x, y, \theta, \phi)$ is a legitimate snapshot. Configuration DOF: 4.

Step 2. Count the velocities. Rolling without slipping ties the contact-point velocity to the rim: $\dot x = R\dot\phi\cos\theta$ and $\dot y = R\dot\phi\sin\theta$ — two conditions on four velocities, leaving 2 independent ones ( $\dot\phi$ : roll, and $\dot\theta$ : pivot). Velocity DOF: 2. Why these cannot be subtracted from the 4: the <a class="wiki-link" href="/wiki/aq-mech-constraints-classification" data-track="wiki_article.body.wiki_link">constraints article</a> proves (by the Frobenius wedge test) that neither condition is integrable — there is no function $g(x,y,\theta,\phi) = \text{const}$ they are equivalent to, hence no coordinate they eliminate.

Step 3. Show the four coordinates are honestly independent — every configuration is reachable. The parallel-parking cycle does it constructively: roll by $\Delta\phi$ at heading 0, pivot by $\beta$ , roll back by $-\Delta\phi$ , pivot back. Every leg respects the rolling conditions, $\theta$ and $\phi$ return to their start values, yet the position nets a displacement (sideways component $\approx R\,\Delta\phi\,\beta$ for small $\beta$ ). Iterating reaches any $(x,y)$ ; a final roll-and-pivot sets any $(\theta, \phi)$ . Reachable set: all of the 4-dimensional space. Why this settles it: a constraint that removed a coordinate would confine the disk to a 3-dimensional surface, and no maneuver could leave it. The maneuver leaves it.

Step 4. Reconcile the two counts. Both are real and answer different questions. Four is the dimension of configuration space: four generalized coordinates, four Lagrange equations (plus multiplier terms), four numbers to record a snapshot. Two is the number of controls: independent ways to move at each instant — what a driver actually holds (gas and steering). The gap between them, $4 - 2 = 2$ , is the count of nonholonomic constraints, and it measures precisely how much maneuvering (how many bracket-generated directions) the system must exploit to roam its full configuration space. A car with a trailer runs the same ledger: 5 coordinates, still 2 controls, gap 3 — which is why reversing a trailer is hard. Verified: the 4-coordinate, 2-control accounting of the rolling disk matches Goldstein §1.3's discussion of the rolling coin as the standard nonholonomic example; the reachability construction is the standard Chow-theorem illustration.

Limiting cases

Case	Result	Meaning
free particle	$f = 3$	no constraints: $3\cdot1 - 0$
$N$ unconstrained particles	$f = 3N$	a gas before any rigidity is imposed
two particles + rigid rod	$f = 5$	a line has no distinguishable roll about its own axis
any rigid body	$f = 6$	the ladder saturates: 3 translations + 3 rotations, for any $N$
rigid body confined to a plane	$f = 3$	$(x, y, \varphi)$ — the link count used by Grübler
fully braced structure (rigid triangle of links)	$F = 0$	a structure, not a mechanism: no motion possible
rolling disk	4 coordinates, 2 controls	nonholonomic constraints reduce velocities, never coordinates

Common trapJust subtract one for every constraint equation you can write down

Three separate failures hide in that instinct. First, redundancy: a rigid body's $N(N-1)/2$ distance equations would give a negative count for $N \geq 8$ — only the rank of the constraint Jacobian may be subtracted, and it is $3N - 6$ , never more. Second, nonholonomy: rolling conditions are velocity constraints that fail the integrability test, so they remove exactly zero coordinates — the rolling disk keeps all 4 despite obeying 2 constraints. Subtracting them (getting 2) miscounts the configuration space and produces equations of motion for a different, fictitious system. Third, stiffness is not rigidity: a very stiff spring between two atoms is no constraint at all — the vibrational coordinate still exists, merely expensive to excite. Whether you may treat it as frozen depends on the energy scale of the question ( $k_B T$ versus $\hbar\omega$ ), not on the spring constant alone. Count rank, count only holonomic, and count honestly-rigid.

Worked examples

1 · The molecular freedom budget, checked against the lab (confidence builder)

Count the degrees of freedom and vibrational modes of N $_2$ , H $_2$ O, and CO $_2$ ; then predict the room-temperature molar heat capacity of nitrogen and compare with the measured $c_V \approx 20.8\ \mathrm{J/(mol\,K)}$ .

Step 1. N $_2$ ( $N = 2$ , linear): total $3N = 6$ . Budget: 3 translations + 2 rotations + $6 - 5 = 1$ vibration (the bond stretch).

Step 2. H $_2$ O ( $N = 3$ , bent): total 9. Budget: 3 + 3 rotations (bent molecules rotate about three axes) + $9 - 6 = \boxed{3\ \text{vibrational modes}}$ — the symmetric stretch, the bend, and the asymmetric stretch of every infrared textbook.

Step 3. CO $_2$ ( $N = 3$ , linear): total 9. Budget: 3 + 2 + $9 - 5 = \boxed{4\ \text{vibrational modes}}$ — one symmetric stretch, one asymmetric stretch, and a doubly degenerate bend (the molecule can buckle in two perpendicular planes; that pair is the geometric origin of the extra mode a linear molecule carries).

Step 4. Heat capacity of N $_2$ at 300 K. Equipartition pays $\tfrac12 R$ per quadratic energy term per mole. Translations: 3 kinetic terms. Rotations: 2 kinetic terms. Vibration would pay $2\times\tfrac12 R$ (kinetic plus potential) — but the stretch quantum for N $_2$ is $\hbar\omega/k_B \approx 3390\ \mathrm{K}$ , so at 300 K that mode is frozen out. Prediction: $c_V = \tfrac52 R = 2.5 \times 8.314 = \boxed{20.8\ \mathrm{J/(mol\,K)}}$ — matching the measured value. The "missing" vibrational $R$ that classical counting demanded was, historically, a first glimpse of quantization.

Step 5. Sanity-check the counting the other way: at 3000 K the vibration thaws and $c_V$ climbs toward $\tfrac72 R \approx 29.1\ \mathrm{J/(mol\,K)}$ — the freedom budget is temperature-revealed, not temperature-dependent: the mode always existed; only its price changed.

2 · Mobility audit of three machines (exam level)

Apply DOF counting to (a) a windshield-wiper four-bar linkage, (b) a five-bar linkage with two motors, and (c) a spatial six-revolute welding arm. Then explain why the arm needs exactly six joints to weld at an arbitrary point with arbitrary torch orientation, and what a seventh joint would buy.

Step 1. (a) Four-bar: $n = 4$ links (ground, crank, coupler, rocker), $j = 4$ hinges. Grübler, eq. (6): $F = 3(4-1) - 2(4) = 9 - 8 = \boxed{1}$ . One motor at the crank determines the entire motion — which is the point: a wiper must sweep one prescribed arc, repeatably, from one input.

Step 2. (b) Five-bar: $n = 5$ , $j = 5$ : $F = 3(4) - 10 = \boxed{2}$ . Two independent inputs are required — and sufficient: driving the two ground-adjacent cranks steers the coupler point anywhere in its planar workspace. This two-motor architecture is exactly a planar parallel pick-and-place robot.

Step 3. (c) Spatial arm: 6 revolute joints, $n = 7$ links counting ground, each joint permitting $f_i = 1$ freedom. Spatial Grübler: $F = 6(7-1) - \sum(6-1) = 36 - 30 = \boxed{6}$ .

Step 4. Why six is the magic number: the welding torch is a rigid body, and Derivation A says placing a rigid body takes exactly 6 numbers — 3 for the weld point, 3 for the torch orientation. A task that prescribes all six needs $F \geq 6$ ; a serial arm with $F = 6$ can meet it (within reach limits) with, generically, a finite set of joint solutions.

Step 5. The seventh joint: $F = 7$ gives a one-parameter family of joint configurations for every torch pose — redundancy. The extra freedom changes nothing about the task count and everything about the execution: the arm can slide its elbow around obstacles while the torch holds still. Counting says when redundancy exists ( $F$ minus task DOF $= 1$ ); using it well is the control problem. Modern seven-axis collaborative arms are precisely this trade.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831–1879 · Cambridge and King's College London

Maxwell counted degrees of freedom for a living twice over. His kinetic theory hinged on the equipartition theorem — energy shared at $\tfrac12 k_B T$ per quadratic term — which made a gas's heat capacity a pure exercise in counting molecular freedoms: so many translations, so many rotations, so many vibrations. And there Maxwell hit the cleanest anomaly in nineteenth-century physics: diatomic gases stubbornly paid for five freedoms, not the seven his counting demanded. He called the discrepancy the greatest difficulty the molecular theory had yet encountered, and he was right to take it seriously — the missing freedoms were frozen by quantization, invisible until Planck and Einstein, and the humble act of counting DOF had exposed the first crack in classical physics. Maxwell also counted machines: his 1864 rule for the rigidity of frames ( $b \geq 3j - 6$ bars to brace $j$ joints in space) is the structural twin of the Grübler criterion, still taught to civil engineers.

History

1775

— Euler establishes the kinematics of rigid bodies: any displacement is a translation plus a rotation, cementing the 3 + 3 anatomy of the six freedoms.

1788

— Lagrange's Mécanique analytique makes the count of independent coordinates the organizing first step of mechanics.

1864

— Maxwell gives the counting rule for the rigidity of pin-jointed frames, the structural ancestor of mobility criteria.

1871

— Boltzmann generalizes equipartition: each quadratic term in the energy carries the same average share, turning DOF counting into thermodynamic prediction.

1883

— Grübler publishes the mobility criterion for mechanisms, DOF counting for machines with links and joints.

1978

— The PUMA robot arm standardizes the six-revolute architecture — exactly the six freedoms a rigid tool requires.

Going further

References. Goldstein, Poole & Safko, Classical Mechanics (3rd ed.), §1.3 · Morin, Introduction to Classical Mechanics, Ch. 6 · Landau & Lifshitz, Mechanics (Vol. 1), §1 · Taylor, Classical Mechanics, Ch. 7

Provenance. Counting rule $f=3N-k$ and rigid-body result $f=6$ from Goldstein §1.3 and Landau-Lifshitz Mechanics §1. Molecular mode counts $3N-6$ / $3N-5$ are standard (verify: water = 3 modes, CO2 = 4 modes). A verifier should recompute each worked example's constraint count and confirm the listed $f$ values against Morin Ch. 6 and Taylor Ch. 7; all numbers are standard textbook values. Additional counts on this page (Grübler $F = 3(n-1)-2j$ : four-bar 1, five-bar 2, triangle 0, spatial 6R arm 6; N $_2$ $c_V = \tfrac52 R \approx 20.8\ \mathrm{J/(mol\,K)}$ with $\hbar\omega/k_B \approx 3390$ K; CO $_2$ bend at 667 cm $^{-1}$ ≈ 15 μm) are standard kinematics and spectroscopy values, recomputed from the stated formulas.

Builds on: Generalized Coordinates and Configuration Space · Constraints and Their Classification Leads to: The Lagrangian Function · The Euler-Lagrange Equations · Virtual Displacements and the Principle of Virtual Work

Why exactly does a linear molecule get only 2 rotational degrees of freedom instead of 3?

Because rotation about the molecular axis changes nothing about the configuration: every atom sits on the axis, so no atom moves. A degree of freedom must label distinguishable arrangements, and there is nothing to distinguish. (Classically you can also say its moment of inertia about the axis is essentially zero — the electrons' contribution is negligible — so no meaningful energy can be stored there; quantum mechanically, the level spacing for that rotation is so enormous it is never excited.) The freedom does not vanish from the books, though: the total is still $3N$ , so the "lost" rotation reappears in the vibrational budget — which is precisely why linear CO $_2$ has $3N-5 = 4$ modes while bent water has $3N - 6 = 3$ . Geometry moves the freedom from one column of the ledger to another.

Why does a vibration contribute R to the heat capacity but a translation only R over 2?

Equipartition pays $\tfrac12 k_B T$ per quadratic term in the energy, and a vibration owns two: kinetic ( $\tfrac12\mu\dot s^2$ ) and potential ( $\tfrac12 k s^2$ ). A translation has kinetic energy only — free motion stores nothing. So the honest currency of equipartition is quadratic terms, not degrees of freedom; each vibrational DOF is worth two terms, each translational or rotational DOF one. That is how a diatomic's budget of $3+2+1$ freedoms prices out at $\tfrac32 R + R + R = \tfrac72 R$ when everything is active — and why the measured $\tfrac52 R$ at room temperature was such a scandal: a whole $R$ 's worth of vibrational payment was missing, frozen by a quantum the nineteenth century didn't have.

Does a time-dependent constraint change the degree-of-freedom count?

No. A rheonomic constraint like "the wire rotates at fixed rate $\omega$ " is still one equation among coordinates and time ( $\theta = \omega t$ ), so it still removes exactly one coordinate — the bead on the rotating wire is a genuinely one-DOF problem in $r$ . What explicit time-dependence changes is the energy bookkeeping (the conserved Jacobi function parts company with $T + V$ , as the constraints article develops) and the embedding $\mathbf{r}(q, t)$ of equation (4), which picks up a $t$ slot. Time itself is never a degree of freedom in this counting: DOF counts the dimension of configuration space — the possible arrangements at one instant — not the parametrization of histories.

So what IS the degree-of-freedom count of a car — four or two?

Both, for two different questions, and being able to say which is which is the whole lesson of Derivation C. Configuration: four (position $x, y$ , heading $\theta$ , and wheel rotation — five with a trailer angle). That is how many numbers a parking app would need to record where your car is, and no rolling constraint reduces it, because rolling constraints are nonholonomic. Controls: two (drive and steer) — how many independent ways you can move right now. The gap between the counts is the number of nonholonomic constraints and is the exact measure of why parking requires maneuvering: the missing directions must be generated by composing the two controls back and forth. When someone quotes "a car has 3 DOF," ask which ledger they are reading — it is usually the planar rigid-body count $(x, y, \theta)$ with the wheels ignored.

If I replace a rigid rod with an extremely stiff spring, do I get the degree of freedom back?

Formally, yes, instantly: a spring — however stiff — imposes no equation among coordinates, so the count jumps from 5 back to 6 for a diatomic dumbbell. Physically, the recovered freedom is expensive: exciting it costs energy of order the vibrational quantum $\hbar\omega$ , and if $k_B T \ll \hbar\omega$ the mode simply never activates — frozen, exactly like N $_2$ 's stretch at room temperature. So the practical modeling rule is energetic, not mechanical: treat a stiff connection as a rigid constraint whenever the energies in play are far below the mode's quantum (or, classically, when its frequency is far above every frequency you care about), and remember the freedom is dormant, not deleted — raise the temperature and it will demand its equipartition share. Rigidity is always an approximation with a price tag attached.

The pairwise-distance constraints of a rigid body outnumber the coordinates for large N. Why doesn't the formula give negative degrees of freedom?

Because equation (1) subtracts the number of independent constraints — the rank of the constraint Jacobian, condition (3) — and never the raw equation count. Of the $N(N-1)/2$ distance equations, only $3N - 6$ are independent: once each particle is pinned by its distances to three non-collinear anchors (Derivation A's construction), every remaining pairwise distance is automatically implied and adds nothing to the rank. Redundant equations are harmless to the physics but fatal to naive subtraction. The same discipline saves Grübler counting from special geometries: a mechanism with deliberately parallel axes can carry constraints that look independent but aren't, which is exactly how "overconstrained" linkages move despite a mobility formula that says $F \leq 0$ .

How do I actually find the number of independent constraints in a messy system?

Strategy one — construct, don't subtract: place the parts one at a time, tallying the freedoms each placement genuinely retains (this is how Derivation A tames the rigid body, and how you should handle chains, linkages, and trees of bodies). Strategy two — count knobs directly: ask how many numbers you must dial to freeze the configuration completely; for a double pendulum you will find yourself dialing two angles, so $f = 2$ without ever listing the four constraints. Strategy three — when the constraints are given as equations, compute the Jacobian's rank at a generic configuration, per condition (3). Cross-check at least two strategies on anything unfamiliar: agreement between the constructive count and the knob count is the fastest error-detector in setting up mechanics problems, and disagreement always means a hidden, redundant, or nonholonomic constraint.