For simplicity, consider a two-dimensional onion.
Insight: The depth to which you have to aim your knife for radial cuts depends on the number of layers.
So, we might as well consider the limiting case as the number of layers approaches infinity.
So, we might as well consider the limiting case as the number of layers approaches infinity.
Similarly, the number of cuts being made has an effect on the answer. So, for simplicity, we can think of making infinitely many cuts as well.
Rectangular
Problem: With infinitely many layers and cuts, the area of each piece of onion is zero. So, it is hard to measure variance.
Solution: Recognize that the Jacobian
Fact from Integral Calculus: the average value,
Here, over a quarter onion of radius 1, the average "relative area",
To generalize the variance we saw earlier, we recall the variance is the average of the square deviations from the mean! So, the variance of our relative area is
All of this is great, but it doesn't answer the question!
What allowed all this to work was a coordinate system whose axes cut the onion.
Can we find a coordinate system that cuts the onion in the way described by Chef Kenji Lopez-Alt?
We make a coordinate system for cutting towards a point a distance
This coordinate system only works for the upper half plane, as there are now technically two points in the plane for a given point
In order to mimic our computation for polar coordinates, we need to
Fix
So,
Also,
In order to compute the Jacobian
We need to know what is the relationship between
Define
Using the law of cosines, we can calculate
With
From this, for a given depth
Yikes!
Despite how difficult the Jacobian looks, we can proceed as we did before. Putting the pieces our plan together, we need to first compute
Amazingly, this is not as bad as it looks (even though Mathematica cannot do it).
Let's look at the numerator
Let's look at the numerator
If we go back to the rectangular coordinate system, this would simply be the integral of
Now let's look at the denominator
Now let's look at the denominator
This is fairly easy to do as well. We can see the denominator is equal to
In all, the average ``relative area'' of each piece is
So, finding
This integral cannot be evaluated by Mathematica. We can be clever, though.
We can go back to the Cartesian coordinate system using
But, here we are integrating over a quarter disc. Wouldn't polar coordinates be easier?
Let
So, we now have a closed form for
Whew! We have done a lot of integration work with three different coordinate systems. What do we have at the end of it?
We have an expression for the variance of the ``relative area'' as a function of the depth
Our goal is to minimize variance. What should we do?
Take the derivative and set it equal to zero! (then test).
where
define ס to be the unique root of
We can also do a similar analysis in three dimensions. We model the onion as nested spheres. We first cut the onion in half, as before.
Usually, we cut an onion into slices perpendicular to the root. If we add a
If we do the same analysis as before, we find the three-dimensional onion constant is approximately 0.484457.
This is awfully close to 1/2 (an easy number for humans to estimate. What if we add in the fact that usually we cut off the ends of the onion?
Why do this?
What now?
What questions do you have?