The Brachistochrone: The Fastest Path Is Not a Straight Line

The problem, and why it is a paradox

Fix two points, A above and B below and off to one side, and connect them with a smooth wire. Release a bead at A and let it slide without friction under gravity to B. Among all possible wire shapes, which one delivers the bead to B in the shortest time? Intuition screams "the straight line" — it is the shortest distance, after all. Intuition is wrong.

The catch is that travel time depends on both the distance covered and the speed along the way, and gravity gives the bead speed only as it descends. A path that dives steeply near the start sacrifices a little extra length but is rewarded with high speed almost immediately, and it carries that speed through the rest of the journey. The winning curve, the brachistochrone, strikes the exact optimal trade between "drop fast to get speed" and "don't wander too far." For a typical placement of B, the cycloid reaches it about 16% faster than the straight line even though its arc is roughly 7% longer. Galileo had puzzled over this in 1638 and guessed — incorrectly — that the fastest path was an arc of a circle. He was right that the straight line loses, but wrong about the winner.

Speed, time, and the functional to minimize

Set A at the origin and let y point downward, so the bead's depth below A is y ≥ 0. Because the wire is frictionless, mechanical energy is conserved. Starting from rest, kinetic energy equals the gravitational energy released: ½mv² = mgy, so the speed at depth y is simply

v = √(2gy).

Remarkably, this depends only on how far the bead has fallen, not on the shape of the path. Now write the total descent time as the sum of tiny travel times ds / v along the curve. An arc element is ds = √(1 + y′²) dx, where y′ = dy/dx. The time to descend along a curve y(x) is therefore the integral (a functional — a number assigned to each whole curve):

T[y] = ∫ √(1 + y′²) / √(2gy)  dx.

Our task is to choose the function y(x), among all curves joining A to B, that makes T[y] as small as possible. Ordinary calculus minimizes over numbers; here we must minimize over an infinite-dimensional space of functions. That is precisely the domain of the calculus of variations.

The Euler-Lagrange equation and the Beltrami shortcut

For any functional J[y] = ∫ f(x, y, y′) dx, the minimizing curve must satisfy the Euler-Lagrange equation:

∂f/∂y − d/dx (∂f/∂y′) = 0.

This is the variational analogue of "set the derivative to zero." It says that any small wiggle in the curve, vanishing at the fixed endpoints, changes the time only to second order. Here the integrand f = √(1 + y′²) / √(2gy) has a special feature: it contains no explicit x. Whenever that happens, the Euler-Lagrange equation admits a first integral, the Beltrami identity:

f − y′ · ∂f/∂y′ = constant.

Carrying out the algebra, the y′² terms collapse beautifully and the whole expression reduces to 1 / [√(1 + y′²) · √(2gy)] = constant. Squaring and rearranging kills the g and the awkward roots, leaving a clean, purely geometric differential equation:

y · (1 + y′²) = 2r,

where 2r is a positive constant fixed by the requirement that the curve pass through B. Every brachistochrone, whatever the target point, obeys this one equation; the constant r just sets the scale.

Why the answer is a cycloid

The equation y(1 + y′²) = 2r is solved exactly by the cycloid, the path traced by a fixed point on the rim of a circle of radius r as the circle rolls along a line. In parametric form, with rolling angle θ:

x = r(θ − sinθ),    y = r(1 − cosθ).

To verify, differentiate: dy/dθ = r sinθ and dx/dθ = r(1 − cosθ), so y′ = sinθ / (1 − cosθ). A short trig calculation gives 1 + y′² = 2 / (1 − cosθ). Multiplying by y = r(1 − cosθ) leaves exactly 2r — the constant, confirmed. For the brachistochrone we use the cycloid inverted, with its sharp cusp at the starting point A: the curve leaves A vertically (initially near-free-fall, which is why it builds speed so efficiently), sweeps out and down, and arrives at B tangent to its motion. One full arch spans a horizontal distance of 2πr and has total arc length exactly 8r — a striking, r-independent integer multiple. This is a genuine geometric object, not just an abstract solution: the fastest slide is literally the track of a rolling wheel.

Johann Bernoulli's optical trick: Snell's law in disguise

Bernoulli's own solution is one of the most admired arguments in mathematics because it barely computes at all. He noticed that light, by Fermat's principle, also travels the path of least time. In a medium where light's speed varies from layer to layer, Snell's law of refraction holds: sinθ / v = constant, where θ is the angle between the ray and the vertical.

So Bernoulli imagined the falling bead as a ray of light passing through a stack of infinitely thin horizontal layers, in each of which the "speed of light" happens to equal the bead's gravitational speed v = √(2gy). Applying Snell's law across every layer forces sinθ / √(2gy) = constant along the whole path. Since the geometry gives sinθ = 1/√(1 + y′²), this is exactly the Beltrami relation 1 / [√(1 + y′²) · √(2gy)] = constant — the same cycloid, obtained with no variational machinery. It was a stunning demonstration that the mechanics of a falling bead and the optics of bending light are the same optimization problem wearing two costumes.

The tautochrone bonus and Huygens' pendulum

The cycloid carries a second miracle, discovered by Christiaan Huygens decades earlier (1659). It is also the tautochrone ("same time"): a bead released from rest at any point on an inverted cycloid reaches the bottom in the identical time, regardless of how high or low it started. High releases travel farther but gather more speed; low releases travel a shorter, gentler arc — and the two effects cancel perfectly.

The reason is that, measured along the cycloid's arc length s from the lowest point, the equation of motion becomes that of a simple harmonic oscillator: s″ = −(g/4r) s. The descent time to the bottom is one quarter of the period, π√(r/g), containing no amplitude — hence independent of the starting height. This is what makes a cycloidal pendulum exactly isochronous, unlike an ordinary pendulum, whose period drifts with swing size. Huygens exploited it in clock designs by hanging the bob between cycloidal cheeks. The same π√(r/g) also governs the brachistochrone: a bead sliding the half-arch from the cusp down to the lowest point arrives in precisely that time.

History: a 1696 duel and the birth of a new calculus

In June 1696 Johann Bernoulli published the problem in the journal Acta Eruditorum as an open challenge "to the most acute mathematicians of the entire world," allowing six months. The list of solvers reads like a roll call of the era's giants: Isaac Newton, Gottfried Leibniz, Guillaume de l'Hôpital, Johann himself, and his brother Jakob Bernoulli. Newton, then Warden of the Mint, is said to have received the problem in the late afternoon and solved it by 4 a.m., publishing his answer anonymously in the Philosophical Transactions. Bernoulli reportedly recognized the author instantly, remarking that one knows "the lion by his claw" (tanquam ex ungue leonem).

The deepest legacy came from Jakob Bernoulli's more laborious solution, whose general method of comparing a curve with nearby curves was the seed of the calculus of variations. Leonhard Euler and Joseph-Louis Lagrange later systematized it into the Euler-Lagrange equation and Lagrange's δ-variation notation, giving physics one of its master tools. The same framework yields the shortest paths on curved surfaces (geodesics), the shape of hanging chains, the principle of least action underpinning classical and quantum mechanics, and modern optimal-control theory. A puzzle about a sliding bead turned out to be a doorway to how nature optimizes almost everything.

Candidate descent paths from (0,0) to a point at (πr, 2r), with y measured downward. Times are in units of √(r/g); shorter is faster.

Path	Key behavior	Path length	Descent time
Straight line (chord)	Shortest distance, constant acceleration	3.72 r	3.72 (slow)
Circular arc	Galileo's 1638 guess; beats the chord but not optimal	~3.9 r	~3.2
Cycloid (brachistochrone)	Drops steeply first to gain speed early	4.00 r	π ≈ 3.14 (fastest)
Parabola / other smooth arcs	Curved, but wrong curvature profile	varies	> 3.14