I have a donut. Its boundary is a two-dimensional surface embedded in three-dimensional space, and surely is homeomorphic to a torus. If we fix a Riemannian metric on the space, it induces a two-dimensional Riemannian metric on the donut. Forgetting the scaling, we get a conformal structure on the surface. If we fix an orientation, this defines a complex structure.

Complex surfaces homeomorphic to a torus are classified by their $j$-invariant. Thus, I would like to calculate the $j$-invariant of my donut. There are several variants of this question:

  1. Suppose my donut is modeled in cylindrical coordinates $(r,z,\theta)$ by the equation $(r-1)^2 + z^2=\rho^2$. What is the $j$-invariant as a function of $\rho$?

  2. Find an actual, real donut, and approximate it by a surface of this form. Estimate $\rho$ and thereby estimate its $j$-invariant.

  3. Determine how your donut differs from this model, and estimate in which direction that might change its $j$-invariant. Or solve a different model that better fits your donut.

I would be happy to see progress on any of these vitally important problems, which you will no doubt agree have many fruitful applications to mathematics and our daily life.

  • 2
    I tried 2, but ended up eating the doughnut! – Lord Shark the Unknown May 1 '17 at 16:30
  • @LordSharktheUnknown is woke af for eating the 🍩 – Dave huff May 1 '17 at 17:19
  • 2
    If your donut is a jelly donut then it will not be homeomorphic to a torus and will thus not be able to be classified by a j-invariant. You will have to use a related invariant called a jelly-invariant. – ei pi May 1 '17 at 17:20
up vote 6 down vote accepted

For $\rho\in (0,1)$, write $D_{\rho}$ for your donut $(r-1)^2 + z^2 = \rho^2$. For $x\in\mathbb{R}$, define $$ q_\rho(x):=2\sqrt{1-\rho^2}\tan\left(\frac{x\sqrt{1-\rho^2}}{2\rho}\right)-\rho. $$ We define a parametrization of $D_\rho$: $$ \begin{align*} f_{\rho}:\mathbb{C}&\to D_{\rho}\\ x+iy&\mapsto \left(1+\rho\frac{2 q_\rho(x)}{1+q_\rho(x)^2},\rho\frac{1-q_\rho(x)^2}{1+q_\rho(x)^2},y\right). \end{align*} $$ Here the image is written in $(r,z,\theta)$ coordinates. Note that this function is defined even where $q_\rho(y)$ has poles. It is straightforward but tedious to verify that $f_\rho$ is conformal. If we write $r(x)$, $z(x)$ for the $r$- and $z$-coordinates, this amounts to the identity $$ r'(x)^2+z'(x)^2=r(x)^2. $$ I used a CAS to find the formula for $f_\rho$.

From the formula, we see that $f_{\rho}$ is doubly periodic, with minimal periods $2\pi i$ and $2\pi\rho/\sqrt{1-\rho^2}$. Thus $D_\rho$ is conformally equivalent to the torus $$ \frac{\mathbb{C}}{\mathbb{Z}+i\frac{\sqrt{1-\rho^2}}{\rho}\mathbb{Z}}. $$ Write $j(\tau)$ for the $j$-function, which takes $\tau$ in the upper half plane to the $j$-invariant of $\mathbb{C}/(\mathbb{Z}+\tau\mathbb{Z})$. The $j$-invariant of $D_{\rho}$ is then $$ j\left(i\frac{\sqrt{1-\rho^2}}{\rho}\right). $$

I don't have a donut in front of me, but the internet suggests that a Krispy Kream original glazed has outter diameter $3.5''$ and width from edge to hole $1.4''$. Scaling, this gives the value $\rho=\frac{2}{3}$. Sage then gives the estimate $$ j\approx 2060.94946856905. $$

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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