Pack spheres of diameter one in a face-centred cubic lattice and the fraction of space they occupy is π/√18 = 0.74048048969306, on the first link. The second link reaches the same number as π/(3√2) from a different grouping of the same factors, and the two screens are identical — which is the point of putting them side by side. The arithmetic is in the third and fourth links: four whole spheres of diameter one occupy 4 · π/6 = 2.0943951023932, and the conventional cubic cell that holds them has side √2 and therefore volume 2√2 = 2.8284271247462. Divide the third by the fourth and the screen's 0.74048 is what you get.
What makes this hard is not that the lattice is good but that no irregular arrangement is better, and there is no obvious reason the two questions should have the same answer. Fejes Toth showed in 1953 that the check could be made finite — a few thousand configurations rather than every packing in existence — and Hales used that reduction in 1998, with thousands of linear programmes evaluated on borrowed processor time; the referee panel could not verify the computation to its own satisfaction, and the formal proof in Flyspeck was announced only in 2014. Two facts make the number feel less fragile: stacking hexagonal layers in any order at all, not just the alternating ABC order of the cubic lattice, gives exactly the same density, and in dimensions eight and twenty-four the corresponding problems were solved in 2016 by Viazovska and collaborators using the E8 and Leech lattices — where the proof is a line of analysis rather than a search.
Answer
0.74048048969306
How to read it
π over √18, the density. The second link is the same number by another route, the third and fourth the sphere volume and the cell volume that divide to it.
Source
Johannes Kepler, 'Strena seu de Nive Sexangula', 1611; reduced to a finite check by Fejes Toth in 1953, proved by Thomas Hales in 1998, with the Flyspeck formal verification announced in 2014, and the analogous eight- and twenty-four-dimensional cases settled by Viazovska in 2016
Work it in the calculator
These links open the right calculator with this problem's numbers already typed in. The line above is what the screen should show — check it yourself.