✧Evanescentia
Something Fun/Sphere packing
A little mathematical curiosity · 01

How many spheres
can fit around one?

One big sphere. A collection of smaller ones. A surprisingly difficult question about touching, space, and the limits of “one more”.

Explore the arrangement

The sphere playground

INTERACTIVE · DRAG TO EXPLORE
Central sphereAttached spheres
3D VIEW
Enable JavaScript to explore this arrangement.
Drag the model to look around

Spheres in this arrangement
28
A constructive lower bound
Upper bound on the maximum
34
Minimum centre angle: 38.94°
N(x) lies between 28 and 34

The true maximum is one integer. A range expresses what these bounds establish.

A simple question, a rich problem.

01 / The question

Take a sphere of radius m cm. How many equal spheres of radius m/x cm, with x > 1, can touch its outside at once without overlapping each other?

Call the greatest possible count N(x). Is there a pattern? Does changing the size of the central sphere change the answer?

m > 0Central sphere radius
r = m/xSmall sphere radius
x > 1Ratio of the two radii
N(x)Maximum possible count

The working out, in three moves.

02 / Follow the geometry
01

Follow the centres

A small sphere touches the large one when their centres are m + r apart. So all small-sphere centres sit on a larger, imaginary sphere.

R = m + r = m(1 + 1/x)
02

Keep neighbours apart

Two small spheres cannot overlap, so their centres must be at least 2r apart. A chord between the centres turns this into an angle.

2R sin(θ/2) ≥ 2r
03

Let the scale cancel

Substitute r = m/x. The m disappears! Only the ratio x determines the required separation—and therefore the possible count.

θ ≥ 2 sin−1(1/(x + 1))
Where does the upper bound come from?

On a unit sphere, draw a cap of angular radius α = sin⁻¹(1/(x + 1)) around each centre direction. The interiors of these caps do not overlap. Each cap has area 2π(1 − cos α); together they cannot exceed the sphere’s area, 4π. Hence:

N(x) ≤ ⌊ 2 / (1 − √(1 − 1/(x + 1)²)) ⌋

The area bound leaves gaps between caps, so it usually overestimates the true maximum. A stored, non-overlapping arrangement provides the lower bound. This explorer displays the largest feasible arrangement in its dataset; it does not solve the global optimisation problem live.

Near equal sizes, the proven 13-sphere result gives N(x) = 12 up to approximately x = 1.091145 (excluding the threshold). The explorer conservatively labels x < 1.09114 as exact. Elsewhere, “maximum unproved” means unproved by this explorer, not necessarily unknown to mathematics.

“Fits” is not the same as “most”.

03 / An important distinction
Valid

Every small sphere touches the central one, and no two overlap. Even a single sphere is a valid arrangement.

Maximal

You cannot add another sphere while keeping all current positions fixed. A different arrangement might still fit more.

Maximum

No valid arrangement anywhere can contain more spheres. This is the global best, and it needs a proof.

Our x = 2 example: the model shows 28 spheres and an upper bound of 34. It does not certify that the displayed arrangement is maximal or maximum. If one sphere is removed from a valid 28-sphere arrangement, the remaining 27 are valid but not maximal: the removed sphere still fits in its old position.

Built from our sphere-packing exploration. Mathematical context: Musin & Tarasov, The strong thirteen spheres problem ↗.