Inspired by the delicious buns and Siu Mai in bamboo steamers I saw tonight in a food show about Cantonese Dim Sum, here is a natural question. It probably has been well studied in the literature, but I cannot find related reference.

Given a real number $r \le 1$, let $f(r)$ be the maximum number of radius-$r$ disks that can be packed into a unit disk. For example, $f(1)=1$ for $r \in (1/2, 1]$, $f(r)=2$ for $r \in (2\sqrt{3}-3, 1/2]$, etc.

Question: Is it true that $\{f(r): r \in (0, 1]\}=\mathbb{N}$?

Noam Elkies immediately pointed out that the maximum radius allowed for packing $6$ and $7$ balls are the same (both $1/3$). Now let me change the problem to just make it harder :)

Modified question: Is $\mathbb{N} \setminus \{f(r): r \in (0, 1]\}$ a finite set?

  • 2
    $\begingroup$ A good resource for this kind of question is Erich's Packing Center, erich-friedman.github.io/packing/index.html $\endgroup$ Apr 2, 2021 at 11:55
  • 4
    $\begingroup$ Changing the question just because it was answered quickly is a bit unfriendly towards the answerer, and confusing for later readers (who will find answers to the two different questions mixed together) — it’s much better to accept the answer, and then ask the follow-up as a new question. Of course, it’s different if the quick answer is trivial, e.g. pointing out a typo, or an obvious degenerate solution the question forgot to rule out. But here, the original question was interesting and non-trivial, and the answer answered it very well. $\endgroup$ Apr 2, 2021 at 13:42
  • $\begingroup$ Peter: Thanks for the comment. I am relatively new to mathoverflow. Which of the following ways are considered acceptable in this case: (a) posting an entire new question including a link to this one, (b) posting the new question as a comment on the original question, or (c) commenting under Noam's answer? I thought the purpose of having websites like mathoverflow is to facilitate the discussions on interesting math problems. But if there are certain community policies users should stick to, I would be happy to learn about them. $\endgroup$
    – Hao
    Apr 2, 2021 at 19:44
  • $\begingroup$ @Hao Supporting Peter's proposal which btw is an agreed procedure in these not so seldom cases, so please go for (a) ! All the more your refined follow-up question merits a new question posting, whereas your (b) & (c) generate comments, being more fugacious in character. $\endgroup$
    – Hanno
    Apr 3, 2021 at 20:18

1 Answer 1


No: according to the pictures in https://en.wikipedia.org/wiki/Circle_packing_in_a_circle, $f(r)$ is never $6$, with $f(\frac13) = 7$ but $f(\frac13 + \epsilon) = 5$. (This result is attributed to the late R.L.Graham's solution in 1968 of a problem in the American Math. Monthly.) It is also known that $f(r_0) = 19$ for $r_0 = 1 / (1+\sqrt2+\sqrt6)$, and conjectured that $f(r_0 + \epsilon) = 17$, which would mean that $f(r)$ is never 18 either. The conjecture is in a 1998 article co-authored by the same R.L.Graham:

Graham RL, Lubachevsky BD, Nurmela KJ, Ostergard PRJ. Dense packings of congruent circles in a circle. Discrete Math 1998;181:139–154.

  • 2
    $\begingroup$ Thanks for the quick answer! I checked the the next interesting case (36 balls) on hydra.nat.uni-magdeburg.de/packing/cci. And somewhat surprisingly the best known radius beats the 37 ball case. $\endgroup$
    – Hao
    Apr 2, 2021 at 1:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.