3

A point $P$ is chosen at random in the coordinate plane. What is the probability that the unit disk with center $P$ contains exactly two lattice points in its interior?

In short I've been trying to wrap my head around this problem but I have not been able to produce anything of value. Below I provide my thinking:

In the Cartesian plane we'll use grid lines as reference points for our coordinate system (each integer is represented by the intersection of two grid lines). enter image description here

As you can see in the image above (I just worked on first quadrant for sake of simplicity), we can achieve our scope if and only if the point $P$ lies exactly on any grid line but without being on any crossroads or exactly strictly below any vertical line or "strictly" to the left or right of any horizontal grid line.

(I know, I am abusing the meaning of the word "strictly".)

Travis Willse
  • 108,056
Mr. Y
  • 2,757
  • I do not see why if you move the point c a little down (thus outside the line, but not to much) it does not contain 2 lattice points? Thus you will get not just a line an thus not 0 probability. – wece Jan 22 '16 at 10:16
  • 1
    $4-\sqrt{3}-\frac{2\pi}{3}$? If you draw the four quarter-circular sectors of unit radius centered one at each corner of a unit square, that is the area in the unit square not covered by the central "X". – Brian Tung Jan 22 '16 at 10:18
  • Whoops.That's right.However,the bounds are really so small there,how would one go to calculate them ?I am editing now btw. – Mr. Y Jan 22 '16 at 10:19
  • It seems like there are some arcs that define those boundaries(for each segment between any two crossroads,I would be thinking of ellipses which encapsulate the correct area in which point $P$ can move). – Mr. Y Jan 22 '16 at 10:24

2 Answers2

5

There is no uniform measure on the plane, so strictly speaking, one cannot make rigorous sense of choosing a random point in the plane. The number of points contained within the unit circle centered at a point $(x, y) \in \Bbb R^2$ is, however, invariant under adding integers to $x, y$, so it preserves the spirit of the problem to reformulate it as follows:

What is the probability $\color{#bf0000}{P_2}$ that the unit circle centered at a point in the unit square $[0, 1) \times [0, 1)$ (randomly chosen with uniform probability) contains exactly two lattice points?

(More symmetrically, one could ask this about a random point in the compact quotient $\Bbb R^2 / \Bbb Z^2$, which is just the torus.) Now, the unit circle centered at a point $P$ containing exactly two lattice points is equivalent to $P$ being with $1$ unit of exactly two lattice points. So, if we draw the unit square and the unit disks centered at the lattice points at the corners of the square, we see that the set of points in the square within $1$ unit of exactly two lattice points is precisely the red region in the following diagram:

enter image description here

Since the unit square has area $1$, the probability that a uniformly selected random point is in the red region is precisely the area of that region, and some elementary geometry gives that it is $$\color{#bf0000}{\boxed{P_2 = 4 - \sqrt{3} - \frac{2 \pi}{3} \approx 0.174.}}$$

Similar, the probability that the unit circle centered at a point contains $3$ lattice points is $$\color{#00bf00}{P_3 = -4 + 2 \sqrt{3} + \frac{\pi}{3} \approx 0.511,}$$ and the probability that the unit circle centered at a point contains $4$ lattice points is $$\color{#0000bf}{P_4 = 1 - \sqrt{3} + \frac{\pi}{3} \approx 0.315.} .$$

This recovers, for example, that the expected number of lattice points is just $$2 \color{#bf0000}{P_2} + 3 \color{#00bf00}{P_3} + 4 \color{#0000bf}{P_4} = \pi .$$

Travis Willse
  • 108,056
3

Consider the square $x_0=(0,0)$, $x_1=(1,0)$, $x_2=(0,1)$ and $x_3=(1,1)$.
If $P$ lies in/on this square, the lattice points in the circle can only be one of the four points above.
So let $P = (a,b)$, then $x_i$ lies in the interior of the circle

  • iff $d(P,x_i) < 1$
  • iff $P$ lies in the interior of the circle around $x_i$.

As this problem is invariant under rotation by $90^\circ$, we can consider the case $x_0$ and $x_2$ and multiply by $4$ in the end.
If $x_0$ and $x_2$ do not lie in the circle around $P$, $P$ lies above the circles around $x_0$ and $x_1$, convince yourself that $x_2$ and $x_3$ then lie in the interior.
So we are interested in the point where $y=\sqrt{1-x^2}$ and $y=\sqrt{1-(x-1)^2}$ intersect, which is at $(x,y)=(\frac{1}{2},\frac{\sqrt{3}}{2})$, then the area we are interested in is $$O = 1 - 2 \int_0^\frac{1}{2} \sqrt{1-x^2} \textrm{d}x = 1 -\frac{ \sqrt{3}} {4}- \frac{\pi}{6}.$$ Now multiply by four, convince yourself that the areas are mutually exclusive for two different pairs $x_i,x_j$ and $x_{i'},x_{j'}$ and that $x_1$ and $x_2$ cannot lie in the same circle.

Then the answer is $4-\sqrt{3} - \frac{2\pi}{3}$.

Please note that convincing yourself is easy by drawing the four unit circles all at once. Furthermore, you know have to translate this probability from one unit square to the infinite two-dimensional plane.

Hetebrij
  • 4,011
  • I accepted Travis's answer as I don't know yet calculus ,but your answer is definetly rigorous so thanks for sharing it. – Mr. Y Jan 22 '16 at 10:52
  • @Mr.Y In my answer I said that one can compute the area of the relevant area using elementary geometry, and this is certainly true, but when preparing my answer I used a CAS to evaluate quickly the same integral that Hetebrij wrote above. Perhaps this answer provides (yet more) incentive to learn some calculus! – Travis Willse Jan 22 '16 at 10:58