A Mathematical Exploration of Norton's Dome and Determinism in Classical Mechanics

Newton's equations determine acceleration from position and velocity, but a differential equation does not automatically have a unique solution for every initial condition. Norton's dome makes this distinction explicit: an idealized force law admits several motions with the same initial position and velocity.

The model

A point mass rests at the apex of a frictionless, rotationally symmetric dome. Let $r\ge0$ denote arc length along a radial meridian, measured from the apex. With the dome's shape and units chosen appropriately, the equation of motion is

\ddot r=\sqrt r,\qquad r(0)=0,\qquad\dot r(0)=0.

The coefficient has been normalized to one; with physical units retained, a dimensional coefficient multiplies $\sqrt r$ . This is a radial equation, not a vector equation for Cartesian displacement. Norton's original discussion explains the idealized surface that produces it.

One solution is permanent rest, $r(t)=0$ . For every $T\ge0$ , another solution on $t\ge0$ is

r_T(t)=\begin{cases} 0,&0\le t\le T,\\ (t-T)^4/144,&t>T. \end{cases}

For $t>T$ ,

\dot r_T=\frac{(t-T)^3}{36},\qquad \ddot r_T=\frac{(t-T)^2}{12}=\sqrt{r_T}.

Both derivatives approach zero at $T$ , so the joined function is twice continuously differentiable and satisfies the equation there as well. Each solution obeys the same initial conditions. Rotational symmetry also leaves the departure direction undetermined.

Where uniqueness fails

Writing the equation as a first-order system gives

\frac{d}{dt}\begin{pmatrix}r\\v\end{pmatrix} =\begin{pmatrix}v\\\sqrt r\end{pmatrix}.

The force term is continuous at $r=0$ , but it is not locally Lipschitz there:

\frac{|\sqrt r-\sqrt0|}{|r-0|}=\frac1{\sqrt r}\longrightarrow\infty.

Thus the usual local Lipschitz hypothesis in the Picard–Lindelöf uniqueness theorem is unavailable. Failing that hypothesis alone would not prove nonuniqueness; the explicit family $r_T$ supplies the proof. Existence is also established directly by these solutions.

The issue is consequently more precise than “classical mechanics is random.” This particular idealized initial-value problem does not select a unique future. It supplies no probability distribution for the departure time or direction.

Does departure violate Newton's first law?

At $t=T$ , force and acceleration are both zero. At every later time, the moving solution has $r>0$ and therefore positive force and acceleration. There is no first instant strictly after $T$ at which acceleration becomes positive.

Zero force at one instant requires zero acceleration at that instant; it does not, by itself, require rest at all subsequent times. Uniform motion follows when the net force remains zero over an interval. The dome's moving solutions satisfy the stated second-order equation throughout.

What the example establishes

The conclusion concerns the mathematical model. Whether its idealizations should count as physically admissible is a further question; constructing a real surface, particle, and contact interaction with exactly these properties is not established by solving the equation.

A determinism claim must specify its admissible states and force laws, as well as the interval on which solutions exist. Norton's dome shows why uniqueness assumptions belong in that claim. It does not show that ordinary smooth mechanical models generally fail to determine their motion.