The figure shows part of an infinitely large cube lattice.
What percentage of points are visible from a given point in the infinite lattice (to the nearest integer)?
Note: A point is not visible if it is directly behind another point in the line of sight.
This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try
refreshing the page, (b) enabling javascript if it is disabled on your browser and,
finally, (c)
loading the
non-javascript version of this page
. We're sorry about the hassle.
Why can't we just use the two-dimensional case: The probability that two numbers are relative primes is p i 2 6 ≈ 0 . 6 0 7 9 3 .
Then extend to 3 dimensions: Adding a third coordinate, this new number will try to have a common factor with each of the previous two. So the probability all three are relatively prime is the previous number cubed or about . 2 2 4 7
Log in to reply
( x , y , z ) has to be relatively prime. It doesn't have to be pairwise coprime.
For example ( 4 , 6 , 7 ) are relatively prime. But not pairwise coprime since 4 and 6 aren't coprime.
May I know how to evaluate the second last line leading to your answer pls
Log in to reply
ζ ( 3 ) (also known as Apéry's constant ) is equal to n = 1 ∑ ∞ n 3 1
Also the product p prime ∏ 1 − p − 3 1 is equal to ζ ( 3 ) . Here's the proof
Absolutely brilliant work
I solved the question in the same way, but I have a doubt. I can say that for every point (x,y,z) where x,y,z are coprime there are infinite points (kx,ky,kz) not visible to us. Thereby I conclude that we see 0% of the points. I know that this thinking is wrong, but I don't know why it's wrong. Please explain.
Log in to reply
The rigorous way to talk about this would be to say the limit as N goes to infinity of the probability that 3 random numbers less than N are coprime. This is more helpful because the concept of a "randomly chosen number" (with no upper bound) is somewhat nebulous and not really possible to define. Instead we have to think about a "randomly chosen number" in terms of asymptotic density.
Log in to reply
Thanks for the information! It makes more sense now!
Hmmm. In my humble opinion every possible point is blocking infinte other points. So we definately see points, but I could not put a number to it. Please explain where I go wrong.
Martien van Geffen
Log in to reply
What is the probability that a randomly chosen natural number is odd? For every odd number n, there are an infinite number of even numbers (2n, 4n, 8n, 16n, ...) This seems to imply that the probability is 0% which doesn't make any sense. Moral of the story: In order to talk about a randomly chosen natural number we must have an upper bound, because the concept of a "randomly chosen natural number" is impossible to define. When trying to talk about the probability that 3 randomly chosen numbers are coprime, we have to say the limit as N goes to infinity of the probability that 3 randomly chosen numbers less than N are coprime.
Log in to reply
This is wrong. There is no such thing "randomly chosen natural number", so there is no meaningful probability of being odd. If you limit to an interval, it makes sense. Please give me an algorithm that choose a random natural number, using a primitive that can flip a fair coin.
Log in to reply
@Attila Kiss – That is exactly the point I was making. As I mentioned, because a "randomly chosen number" is not well defined, we have to talk about the limit as N goes to infinity of a randomly chosen number less than N.
Log in to reply
@John Ross – Still, for every point i could create a point which is obscured by that point. So the number should be less then 50 %.
Log in to reply
@Martien van Geffen – You must have an upper bound in order to talk about a randomly chosen point. If we use your logic and don't have an upper bound, we could prove that much less than 50% (0% in fact) of the natural numbers are odd. This false result arises because we are trying to compare the sizes of two infinite sets, which can lead to contradictions if we treat them like finite sets. If we are going to talk about the size of a subset of the natural numbers as compared to the natural numbers, we must use the concept of asymptotic density. One example of this would be to ask how common perfect squares are. Although there is a one to one correspondence ((1:1),(2:4),(3,9),(4,16),...), this does not mean that perfect squares are just as common as natural numbers. Instead we have to ask how many perfect squares there are below N, compared to N, and what happens to this as N goes to infinity. If we think in terms of asymptotic density, we can see that perfect squares are extremely rare among the natural numbers, even though there is a one to one correspondence between the two.
Log in to reply
@John Ross – I think that in the end we need some well defined rules for the comparision of infinite sets. Can any of you guys suggest me something to read from?
Log in to reply
@Aatman Supkar – The well defined rules we have for comparing infinite sets are called asymptotic density. You could read some more about it if you are interested.
@John Ross – I agree that we must have un upper bound, but the problem stated un infinite lattice. The problem therefor does not induce a upper bound. That is my problem. Maybe the question is wrong?
Log in to reply
@Martien van Geffen – The reason we must talk about limits as N goes to infinity instead of just N=infinity is because you would be in effect trying to show that (infinity) divided by (2*infinity) is equal to one half or something similar. Probability is usually defined as (favourable outcomes)/(possible outcomes), but if you try to plug in infinity it doesn't work. Mathematicians have found that when trying to compare the sizes of infinite sets, they must use asymptotic density. The question as posed is correct just like to say that half of the natural numbers are odd is correct.
From how the question was worded, I also thought the non hidden points were just those visible on the surface of the cube that was visible to us. That is, like an opaque cube. Since the surface points are infinitesimally small compared to the rest, I got 0
Yes, I remember the Euler Product Formula from the increasingly dim and distant past. However I solved this by a simpler and less sophisticated method: initial value of p (the probability that a point is obscured by another) is p(0) = 0. Then use the iteration p(n) = p(n-1) + {1-p(n-1)} / {r(n))^3}, where r(n) is the nth prime. This converges pretty quickly to 0.168..., so required answer is 83%
Log in to reply
Patrick, I'm not seeing how you came up with your iteration formula. Would you add a bit more detail, or explanation? Thanks
What is the ratio of the size of two infinite sets? In this case it is possible to set up a one-to-one correspondence between the visible and not-visible points. These two infinities are the same. It makes no sense to take a fraction of an infinity, as infinity isn't a number and mathematical operations are meaningless. John Ross' comment points to the proper way to ask the question: what is the ratio of hidden points as the distance from the origin approached infinity?
How do things change if we are not at the origin? Do we see all points if our location is irrational? If we are at a rational location, say (1/4, 1/3, 5/7) will there be blockage?
Related numberphile video:
https://www.youtube.com/watch?v=p-xa-3V5KO8
you can start at 9:35 to skip to the relevant section
First, considering the 2-D case. Fix a point e consider this as the origin O of a cartesian reference frame. A point P = ( x , y ) is visible from the origin O only if the G C D ( x , y ) = 1 or equivalent x and y are co-prime. If the G C D ( x , y ) = m > 1 then there are other m-1 points on the same direction of P that are less distant from O than P (see figure below)
In the 2-D case the frequency of 2 coprime numbers can be calculated in the following way
Given a number x , the probability that can be divisible by a prime number p is 1 / p . Since x and y are chosen indipendetly, the combined probability that x and y are divisible for p is the product of the single probability. Then
P ( x ∣ p ∩ y ∣ p ) = P ( x ∣ p ) P ( y ∣ p ) = 1 / p 2
so the probability that x and y are not divisible by p is 1 − 1 / p 2 .
The probability that x,y are coprime is the probability that both x and y are not divisible by any prime number p then
P ( x and y are coprime ) = ∏ prime p ( 1 − p 2 1 )
This infinite production is equal to 1 / ζ ( 2 ) = 6 / π 2 where ζ ( z ) is the Riemann zeta function (see Coprime Numbers )
Since for a infinite number of trials, the probability of a event tend to be the frequency that this event happen (see [Law of large numbers] )(https://en.wikipedia.org/wiki/Law of large_numbers)
In the 3-D case the condition that given a point O = ( 0 , 0 , 0 ) , another point P = ( x , y , z ) is visible from O is that the x,y and z are coprime. So generaling the formula above the probability that x,y and z are coprime is
P ( x , y and z are coprime ) = ∏ prime p ( 1 − p 3 1 )
This infinite production is equal to
ζ ( 3 ) 1 = 1.202056.... 1 = 0.83 *
Then the percentage of points that are visible from a given point in the infinite lattice is 83%
paragraph 1
see Apery constant for more details about ζ ( 3 )
Problem Loading...
Note Loading...
Set Loading...
Relevant wiki: Riemann Zeta Function
First imagine that we are standing at the origin, and that the grid is that of the lattice (integer) points.
The point ( 4 , 6 , 2 ) is not visible because the point ( 2 , 3 , 1 ) is "in the way", and the point ( 6 , 3 , 9 ) is not visible because the point ( 2 , 1 , 3 ) obscures it.
We see that a point ( x , y , z ) is not visible precisely when there is another lattice point blocking it, which is when there is an integer triplet ( x ′ , y ′ , z ′ ) such that ( x , y , z ) = ( k x ′ , k y ′ , k z ′ ) .
Equivalently, ( x , y , z ) is not visible precisely when there is an integer k dividing the trio x , y and z , and is visible when there is no such integer, i.e. when x , y and z have no common factor.
We can now state our question differently: For three random integers x , y and z , what is the probability that they have no common factor?
For this to happen, it must be the case that no prime number p divides x , y and z . For a particular prime p , the probability that x is divisible by p is p 1 , and the same for y and z , so the probability that the trio are divisible by p is p 3 1 , and so the probability that p does not divide the trio is 1 − p 3 1 .
The probability that no prime divides the trio x , y , z is therefore
= ( 1 − 2 3 1 ) × ( 1 − 3 3 1 ) × ( 1 − 5 3 1 ) × ( 1 − 7 3 1 ) × ( 1 − 1 1 3 1 ) × ⋯ = p prime ∏ ( 1 − p 3 1 ) = ⎝ ⎛ p prime ∏ 1 − p − 3 1 ⎠ ⎞ − 1 = ( n = 1 ∑ ∞ n 3 1 ) − 1 = ζ ( 3 ) 1 ≈ 8 3 %
Click here to see the proof that ζ ( s ) = n = 1 ∑ ∞ n s 1 = p prime ∏ 1 − p − s 1