Sphere Packing Problem Explained: Kepler to Viazovska
Quick answer: The sphere packing problem asks how densely identical, non-overlapping balls can fill space. In three dimensions the maximum density is π/√18, about 74.05%, achieved by the face-centered cubic (orange-stack) packing. Exact optimal densities are known only in dimensions 1, 2, 3, 8 and 24; the problem is open in all other dimensions. Maryna Viazovska solved dimension 8 in 2016, and a week later she and four collaborators solved dimension 24.
Stack oranges in a market and you will naturally make a pyramid: each layer sits in the hollows of the one below. Is there any cleverer arrangement that fits more oranges in the same box? This question, simple enough to ask a child, took almost 400 years to answer in three dimensions, and in most dimensions it is still open.
What Is Packing Density?
The density of a packing is the fraction of space covered by the spheres. In two dimensions, a quick comparison shows why arrangement matters. Put circles of radius 1 in a square grid: each circle sits in a \(2 \times 2\) square, so the density is
\[ \frac{\pi \cdot 1^2}{4} = \frac{\pi}{4} \approx 0.785. \]
Shift every other row to form a honeycomb (hexagonal) pattern and the density rises to \(\frac{\pi}{\sqrt{12}} \approx 0.907\). That is the best possible in the plane, proved by Axel Thue and, rigorously, by László Fejes Tóth in 1940.
Where Is Sphere Packing Solved?
| Dimension | Best packing | Density | Proved |
|---|---|---|---|
| 1 | Intervals end to end | 1 | Trivial |
| 2 | Hexagonal circles | \(\pi/\sqrt{12} \approx 0.9069\) | Thue, Fejes Tóth (1940) |
| 3 | Face-centered cubic (orange stack) | \(\pi/\sqrt{18} \approx 0.7405\) | Hales and Ferguson (announced 1998); formal proof 2014 |
| 8 | \(E_8\) lattice | \(\pi^4/384 \approx 0.2537\) | Viazovska (2016) |
| 24 | Leech lattice | \(\pi^{12}/12! \approx 0.0019\) | Cohn, Kumar, Miller, Radchenko, Viazovska (2016) |
| All others | Unknown | Only bounds | Open |
The Kepler Conjecture (3 Dimensions)
In 1611, Johannes Kepler claimed that no packing of equal spheres beats the orange stack, about 74% density. It became part of Hilbert’s 18th problem. In 1998, Thomas Hales, with Samuel Ferguson, announced a proof that reduced the question to thousands of inequalities checked by computer. The computer-heavy proof was so hard for referees to check completely that Hales led the Flyspeck project to verify every step by computer. It finished in 2014, and a paper on the formal proof appeared in 2017. You can see why spheres leave gaps by comparing a sphere’s volume with the cube around it: \(\frac{4}{3}\pi r^3\) versus \(8r^3\), only about 52%.
Viazovska’s Magic Functions (8 and 24 Dimensions)
Higher dimensions sound harder, yet dimensions 8 and 24 turned out special, because two extraordinarily symmetric lattices live there: \(E_8\) and the Leech lattice. In 2003, Henry Cohn and Noam Elkies showed that a single well-chosen function, with special Fourier properties, would prove a packing optimal, and computer experiments suggested such functions exist in dimensions 8 and 24.
In March 2016, Maryna Viazovska found the “magic function” for dimension 8, building it from modular forms. Her proof is about 23 pages. Within a week, she and Cohn, Abhinav Kumar, Stephen D. Miller and Danylo Radchenko adapted it to dimension 24. Viazovska received the Fields Medal in 2022. The related kissing numbers in these dimensions, 240 and 196,560, were settled much earlier, in 1979, by Levenshtein and independently by Odlyzko and Sloane, using similar linear programming ideas.
2026: Sphere Packing Checked by Computer
A Lean project led by Sidharth Hariharan, Chris Birkbeck, Bhavik Mehta, Seewoo Lee, Gareth Ma and Viazovska has been formalizing the 8-dimensional proof since 2024. In February 2026 the team announced a complete, “sorry-free” Lean proof for dimension 8, after Gauss, an AI formalization agent from the startup Math, Inc., finished the remaining goals. Gauss also produced a formalization of the 24-dimensional case. The verified proof artifact is available, while work continues on folding reusable parts into the maintained community project. Along the way, the formalization caught three small oversights in the supporting computer code of the 24-dimensional paper, which the authors corrected in September 2026; the result itself stands. See our guide to the Lean theorem prover for what such a check guarantees.
What About High Dimensions?
In most dimensions, nobody knows the best packing. Researchers work on bounds instead. In 1947, Claude Rogers proved that lattice packings with density about \(c\,d \cdot 2^{-d}\) exist in dimension \(d\). In 2025, Boaz Klartag improved this to about \(c\,d^2 \cdot 2^{-d}\), the biggest gain in decades, using a randomly growing ellipsoid. Several 2026 preprints have pushed both the lower and the upper bounds a little further. Still, upper bounds remain exponentially far from lower bounds, so the true answer in high dimensions is wide open. High-dimensional packings matter in practice: error-correcting codes for phones and the internet are packings in disguise.
Common Misconceptions
“Random packing is just as good.” No. Pouring and shaking balls in a jar typically gives around 64%, depending on the method, well below the orange stack’s 74%.
“Higher dimensions are always harder.” Usually, but dimensions 8 and 24 are solved while dimensions 4 through 7 remain open, thanks to the special lattices \(E_8\) and Leech.
Watch: Sphere Packing in Video
If you learn better by watching, these videos are worth your time.
Check Your Understanding
Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.
Answer each question and press Submit to see your result right away.
Frequently Asked Questions
What is the sphere packing problem?
It asks for the arrangement of equal, non-overlapping spheres that fills the largest fraction of space, in a given number of dimensions.
What is the densest sphere packing in 3D?
The face-centered cubic packing, the familiar orange stack, with density π/√18, about 74.05%. Hales and Ferguson announced the proof in 1998; it was formally verified in 2014.
Why did Maryna Viazovska win the Fields Medal?
For proving that the E8 lattice gives the densest sphere packing in 8 dimensions (2016) and, with coauthors, that the Leech lattice does in 24 dimensions.
Is sphere packing solved in all dimensions?
No. It is solved only in dimensions 1, 2, 3, 8 and 24. In other dimensions only upper and lower bounds are known.
Keep Exploring
- Kissing Number Problem: Exact Values and AI Records
- Fourth Dimension Explained: Tesseracts and Time
- Lean Theorem Prover: Lean 4 and Mathlib Explained
Sources and Further Reading
- Maryna Viazovska, The sphere packing problem in dimension 8 (arXiv)
- Cohn, Kumar, Miller, Radchenko, Viazovska, The sphere packing problem in dimension 24 (arXiv)
- Formalising Sphere Packing in Lean (project page)
- EPFL: Prof. Viazovska’s proofs of sphere packing formalized with AI (2026)
- Boaz Klartag, Lattice packing of spheres in high dimensions using a stochastically evolving ellipsoid (arXiv)
Related to This Article
More math articles
- Kansas KAP Grade 6 Math Free Worksheets: 72 Free Printable Worksheets with Step-by-Step Keys
- Free Michigan M-STEP Grade 5 Math Practice: 49 Printable PDFs for Classrooms, Tutors, and Parents
- Women’s Work, Gender and Household Economies Before the Civil War
- How to Select the Right Tool for the Job
- List Of the Best Middle School Math Supply for Learning
- Common Resources: Overuse, Shared Costs and Rules
- Full-Length 6th Grade PARCC Math Practice Test
- Count Vertices, Edges, and Faces
- Kennedy’s New Frontier and Cold War Development Policy
- Cashing In on Math: How to Master the Addition of Money Amounts


What people say about "Sphere Packing Problem Explained: Kepler to Viazovska - Effortless Math"?
No one replied yet.