The Brachistochrone Problem

The brachistochrone is three centuries old, yet its machinery is exactly that of modern optimal-control and trajectory optimization. Every problem of the form "minimize a travel time or cost subject to dynamics" — a robot arm slewing in minimum time, a spacecraft threading a low-thrust transfer — is a direct descendant, solved by the Euler–Lagrange equation's control-theory generalization, Pontryagin's maximum principle. The cycloid itself resurfaces in the tautochrone: precision clocks and vibration isolators still exploit cycloidal constraints to make a period amplitude-independent. It is even a live demo: departments race beads down machined straight and cycloidal tracks — the cycloid bead reliably wins, and beads dropped from different heights on the same cycloid arrive at the bottom together, a tabletop proof of the tautochrone. The deepest legacy is conceptual: the optical–mechanical analogy Bernoulli used runs straight through Hamilton's principle and the eikonal limit to the path-integral formulation of quantum mechanics, where a particle "tries all paths" and the classical trajectory is the stationary-time one — the brachistochrone idea, promoted to a law of nature.

Race two beads down two frictionless ramps to the same finishing point: one a straight chord, the other a curve that plunges steeply at first and then flattens. The plunging ramp converts height into speed early, so the bead spends most of its trip travelling fast, and the time saved at high speed more than pays for the extra distance. That is the entire physical content of the problem — a competition between a shorter path and an earlier speed-up — and the calculus of variations is simply the bookkeeping that finds the exact winning compromise. The first figure runs the race with real numbers.

Johann Bernoulli found the answer without any calculus of variations at all — he imagined light instead of a bead. By Fermat's principle a light ray takes the path of least time, exactly the quantity we are minimizing. Picture a medium sliced into thin horizontal layers in which the speed of light grows with depth as $v=\sqrt{2gy}$, precisely mimicking the falling bead. Snell's law at each interface keeps the ratio $\sin\phi/v$ constant, where $\phi$ is the angle to the vertical; taking the layers infinitely thin turns that optical condition into an equation identical to the brachistochrone first integral, and the curve the ray sweeps out is the cycloid. This optical–mechanical bridge — a mechanical trajectory extremizes an "action" just as a ray extremizes optical path length — is the seed of Hamilton's principle, and the brachistochrone is where it first flowered. The cycloid answer is geometrically vivid: it is the trace of a chalk mark on a rolling wheel. The bead effectively rolls down an inverted cycloidal track, and — the second miracle — no matter where on that track it starts from rest, it arrives at the lowest point in exactly the same time. The second figure shows how the rolling circle generates the curve and where the two special points $A$ and $B$ sit.

The problem starts from a single sentence made precise: the descent time is the accumulated arc length divided by the local speed, integrated along the curve. Everything else is a computation of, or a consequence of, this one functional.

where: $T$ — total descent time, the functional to be minimized; $\mathrm{d}s$ — arc-length element along the curve; $v$ — the bead's speed at that point. Reading it: to make $T$ small you want $v$ large where $\mathrm{d}s$ is being accumulated, which biases the curve toward gaining speed early. The speed is fixed pointwise by energy conservation. A bead released from rest and having fallen a depth $y$ has converted potential energy into kinetic energy, and the mass cancels.

where: $g$ — gravitational acceleration; $y$ — depth fallen below the start, positive downward. Reading it: the speed depends only on the depth, not on the path taken to get there, and it vanishes at the start where $y=0$. Substituting the arc-length element $\mathrm{d}s=\sqrt{1+y'^2}\,\mathrm{d}x$ and the speed $(2)$ turns the abstract functional into an explicit integral over $x$.

where: $y'=\mathrm{d}y/\mathrm{d}x$ — the local slope; $x_B$ — the horizontal coordinate of $B$. Reading it: the integrand $f(y,y')=\sqrt{(1+y'^2)/y}$ contains $y$ and $y'$ but no explicit $x$ — the structural fact that makes the problem solvable in closed form. Because $f$ has no explicit $x$-dependence, the Euler–Lagrange equation admits a first integral: the combination below is constant along the extremal. This is the Beltrami identity, the "energy" of the variational problem.

where: $f$ — the integrand $\sqrt{(1+y'^2)/y}$; $C$ — a constant set by the geometry. Reading it: this replaces a second-order differential equation with a first-order one, exactly as conservation of energy replaces Newton's second law with a first-order relation in one-dimensional mechanics. Evaluating $(4)$ for the brachistochrone integrand collapses to a strikingly simple first-order ODE.

where: $2a=1/C^2$ — a constant with the dimensions of length; $a$ turns out to be the radius of the generating circle. Reading it: this separable equation is the whole problem in one line, and every curve satisfying it is a cycloid. Integrating $(5)$ and introducing the rolling angle $\theta$ as parameter yields the cycloid in its cleanest form.

where: $a$ — radius of the generating circle; $\theta$ — rolling angle, $0\le\theta\le\theta_B$. Reading it: at $\theta=0$ the curve is at the cusp $A$; the two unknowns $a$ and $\theta_B$ are pinned down by demanding that the curve pass through $B$. A remarkable dividend falls out of the cycloid geometry: the descent time from the cusp, worked out in Derivation 3, is linear in the rolling angle and independent of the release point at the bottom.

where: $\theta_B$ — rolling angle at the endpoint; the right-hand form is the time to reach the lowest point, the same for any starting height on the arc. Reading it: this is the tautochrone (equal-time) property, the fact Huygens exploited for the pendulum clock. Finally, Bernoulli's optical route encodes the same physics as a Snell's-law condition, revealing the mechanics–optics analogy directly.

where: $\phi$ — the angle the path makes with the vertical; $v=\sqrt{2gy}$ — the local speed. Reading it: this is Snell's law for a continuously stratified medium, and squaring it reproduces the reduced ODE $(5)$ exactly.

1 · Cycloid versus straight line, with numbers (fully worked)

Take the generating radius $a=1\ \mathrm{m}$ and $g=9.8\ \mathrm{m/s^2}$, and let $B$ be the lowest point of the cycloid, $B=(\pi a,\,2a)=(3.14,\,2.00)\ \mathrm{m}$. Compute the descent time along the cycloid and along the straight chord, and find the ratio.

Step 1. From Derivation 3 the cycloid descent time from the cusp to the lowest point ($\theta_B=\pi$) is $T_\text{cyc}=\pi\sqrt{a/g}=\pi\sqrt{1/9.8}=\pi\times0.3194=\mathbf{1.00\ s}$. This used no numerical integration — the tautochrone cancellation made the integrand constant.

Step 2. For the straight chord, the bead undergoes uniform acceleration $g_\text{eff}=g\sin\alpha=g\,(y_B/L)$ along an incline of length $L=\sqrt{x_B^2+y_B^2}=\sqrt{\pi^2+4}=3.724\ \mathrm{m}$. So $g_\text{eff}=9.8\times(2/3.724)=5.263\ \mathrm{m/s^2}$, and from rest $L=\tfrac12 g_\text{eff}t^2$ gives $T_\text{line}=\sqrt{2L/g_\text{eff}}=\sqrt{2\times3.724/5.263}=\sqrt{1.415}=\mathbf{1.19\ s}$.

Step 3. The ratio is $T_\text{line}/T_\text{cyc}=1.19/1.00=1.186$, which is exactly the closed form $\sqrt{\pi^2+4}/\pi$. Notice both times scale as $\sqrt{a/g}$, so this ratio is independent of $a$ and $g$: to this endpoint the straight chord is always about 19% slower. Cycloid: 1.00 s; straight line: 1.19 s; the cycloid wins by the factor √(π²+4)/π ≈ 1.19.

2 · Huygens's cycloidal pendulum as a seconds clock (fully worked)

Design a cycloidal pendulum whose period is exactly $2\ \mathrm{s}$ (a "seconds pendulum," one tick per second), and show it beats an ordinary circular pendulum of the same length at large swing.

Step 1. A bob constrained to a cycloid of generating radius $a$ has period $P=4\pi\sqrt{a/g}$ — four times the quarter-period $\pi\sqrt{a/g}$ of Derivation 3, and amplitude-independent because the cycloid is a tautochrone. Setting $P=2\ \mathrm{s}$ gives $\sqrt{a/g}=1/(2\pi)$, so $a=g/(4\pi^2)=9.8/39.48=\mathbf{0.248\ m}$.

Step 2. The cycloidal pendulum swings as if it were a simple pendulum of length $L=4a$ (from matching $4\pi\sqrt{a/g}=2\pi\sqrt{L/g}$), giving $L=4\times0.248=\mathbf{0.993\ m}$ — essentially the textbook seconds-pendulum length. But unlike the simple pendulum this $2.000\ \mathrm{s}$ holds at every amplitude, small or large.

Step 3. Compare a circular pendulum of the same length $0.993\ \mathrm{m}$. At vanishing amplitude it also gives $2\pi\sqrt{0.993/9.8}=2.000\ \mathrm{s}$, but at a $30^\circ$ swing the true period is longer by the factor $1+\theta_0^2/16=1+(0.524)^2/16=1.0171$, i.e. $2.034\ \mathrm{s}$ — a 34-millisecond error every swing that accumulates to about $25$ minutes a day. Huygens's cycloidal cheeks bend the suspension so the bob follows a cycloid, killing this error exactly. a = 0.248 m, equivalent length L = 4a = 0.993 m, exact 2.000 s period at any amplitude — versus a 34 ms per-swing error for the circular pendulum at 30°.

The brachistochrone is inseparable from the man who threw it down as a gauntlet to Europe. Johann Bernoulli did not merely pose the problem — he solved it by a stroke of physical insight that reads today like a preview of the next century of theoretical physics.

Texts. Stone & Goldbart, Mathematics for Physics, §1.1 (the brachistochrone as the opening example of variational calculus) · Arfken & Weber, Mathematical Methods for Physicists, Ch. 22 · Riley, Hobson & Bence, Mathematical Methods for Physics and Engineering, Ch. 22 · Lanczos, The Variational Principles of Mechanics, Ch. II (the tautochrone and Huygens) · Goldstein, Classical Mechanics, §2.2 · Hassani, Mathematical Physics, Ch. on the calculus of variations.

Provenance. The equations are standard textbook results, re-verified here. A checker should recompute: (1) the functional $T=\frac{1}{\sqrt{2g}}\int\sqrt{(1+y'^2)/y}\,\mathrm{d}x$ from $\mathrm{d}s/v$ with $v=\sqrt{2gy}$; (2) the Beltrami reduction to $y(1+y'^2)=2a$; (3) that the cycloid $(6)$ satisfies this ODE; (4) the descent time $T=\theta_B\sqrt{a/g}$ and the tautochrone $T=\pi\sqrt{a/g}$; (5) the worked-example numbers $T_\text{cyc}=1.00\ \mathrm{s}$, $T_\text{line}=1.19\ \mathrm{s}$, ratio $\sqrt{\pi^2+4}/\pi=1.186$, and the seconds-pendulum $a=0.248\ \mathrm{m}$, $L=4a=0.993\ \mathrm{m}$. Cross-check against Stone & Goldbart §1.1, Arfken & Weber Ch. 22, Riley–Hobson–Bence Ch. 22, Goldstein §2.2, and Lanczos Ch. II.

Prerequisites. Functionals and Function Spaces · The Euler–Lagrange Equation · The Beltrami Identity and First Integrals · The First Variation and Stationarity. Leads to. Hamilton's Principle of Stationary Action · Geodesics as Variational Problems · Isoperimetric Problems · Lagrangian Mechanics.

Ask the tutor: Ask me to walk through Bernoulli's Snell's-law derivation line by line, or to solve the cycloid endpoint equations for a specific $B$ and compare the cycloid time against the straight chord.