Arithmetically progressive

Probability Level pending

Find the smallest positive integer N N with the following property. For any subset S S of the set { 1 , 2 , . . . , N } \{1,2,...,N\} , there is an increasing three-term arithmetic progression of numbers from 1 1 to N N such that either all of its terms or none of its terms are in S . S.

Details and assumptions

As an explicit example, if N = 1000 N = 1000 and S S is the subset of all primes less than 1000, then the arithmetic progression 4 6 8 4-6-8 has none of its terms in S S .


The answer is 9.

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.

3 solutions

Ivan Koswara
Sep 22, 2013

If we color the integers in S S red and those not in S S blue, we get a certain case of a famous problem, solved by Van der Waerden . We can simply pull out the number W ( 2 , 3 ) = 9 W(2,3) = \boxed{9} , but that's not very interesting is it? ;)

First, note that N = 8 N = 8 is not valid: We can do the coloring RBBRRBBR to avoid a monochromatic 3-term arithmetic progression (hereby called MAP for Monochromatic (3-term) Arithmetic Progression). Also, N < 8 N < 8 aren't valid either; just truncate the above coloring to N N terms and we still lack a MAP. We will now prove that N = 9 N = 9 suffices. The proof also exists on the net , but what follows is my own solution when I first saw this problem several years ago.

Suppose there exists a coloring that doesn't make a MAP. WLOG let 1 be colored red, otherwise swap red and blue.

  • Case 1: 2 is red

Then 3 is blue, otherwise (1,2,3) forms a red MAP.

  • Case 1.1: 4 is red

Then 6,7 are blue, otherwise (2,4,6) or (1,4,7) forms a red MAP. Then 5,8 are red, otherwise (5,6,7) or (6,7,8) forms a blue MAP. But (2,5,8) forms a red MAP.

  • Case 1.2: 4 is blue

Then 5 is red, otherwise (3,4,5) forms a blue MAP. Then 8,9 are blue, otherwise (2,5,8) or (1,5,9) forms a red MAP. Then 6,7 are red, otherwise (4,6,8), (7,8,9) forms a blue MAP. But (5,6,7) forms a red MAP.

  • Case 2: 2 is blue

  • Case 2.1: 3 is red

Then 5 is blue, otherwise (1,3,5) forms a red MAP. Then 8 is red, otherwise (2,5,8) forms a blue MAP.

  • Case 2.1.1: 4 is red

Then 7 is blue, otherwise (1,4,7) forms a red MAP. Then 6 is red, otherwise (5,6,7) forms a blue MAP. But (4,6,8) forms a red MAP.

  • Case 2.1.2: 4 is blue

Then 6 is red, otherwise (4,5,6) forms a blue MAP. Then 7,9 is blue, otherwise (6,7,8) or (3,6,9) forms a red MAP. But (5,7,9) forms a blue MAP.

  • Case 2.2: 3 is blue

Then 4 is red, otherwise (2,3,4) forms a blue MAP. Then 7 is blue, otherwise (1,4,7) forms a red MAP. Then 5 is red, otherwise (3,5,7) forms a blue MAP. Then 6,9 are blue, otherwise (4,5,6) or (1,5,9) forms a red MAP. But (3,6,9) forms a blue MAP.

The above covers all cases, so it's impossible to have a coloring of 9 numbers without making a MAP. So all colorings of 9 numbers always make a MAP, and so the minimum value of N N is 9 \boxed{9} .

+1 -- I came up with the same case-by-case proof but I wonder if there is a simple conceptual way to show 9 is the answer. I was playing with interpreting the colored sequences as binary numbers but that doesn't seem to lead anywhere. Another idea would be to interpret the binary number as an element of the vector space over the field with two elements and study intersection of planes or some such. Not sure that works either though. Any ideas?

Marek Bernat - 7 years, 8 months ago

Log in to reply

Well, my solution has fewer cases, but I also wish to see a case-less proof. I tried pigeonhole quickly, but didn't find anything.

Daniel Chiu - 7 years, 8 months ago

Log in to reply

There must be a geometrical or graph-theory proof that shows it quickly instead of having to use exhaustive logic like this.

Matt McNabb - 7 years, 8 months ago

Log in to reply

@Matt McNabb If there is one, Van der Waerden's numbers should all be found by using similar logic. The fact that neither all numbers with 2 colors or all numbers with 3-term arithmetic progressions are known suggests that either there is one such method very specific to this case, or there is none at all.

Ivan Koswara - 7 years, 8 months ago

Log in to reply

@Ivan Koswara Now there's a challenge if I ever heard one :)

Or perhaps the solution involves A A dimensions where A = 3 A=3 in this case, the length of the AP.

Matt McNabb - 7 years, 8 months ago

@Ivan Koswara Indeed, the proof must be such that it does not generalize. So it wouldn't be worth very much in the context of Ramsey theory. But I'd still like to see it -- I am always disatisfied by case-by-case analysis :)

Marek Bernat - 7 years, 8 months ago
Vikram Waradpande
Sep 25, 2013

(Note: The solution can be better understood if you keep drawing the things as you read)

We want to find the smallest N N for which the property holds good. Instead, we will find the largest N N for which the property doesn't hold good.

Suppose we have the set { 1 , 2 , 3 , 4 , 5... , N } \{1,2,3,4,5...,N\} and you choose a random element. Let the elements be { i n , i n + 1 , . . . i 3 , i 2 , i 1 , i , i + 1 , i + 2 , i + 3 , i + 4 , . . . i + n } \{ i-n , i-n+1 , ... i-3 , i-2 , i-1, i , i+1 , i+2 , i+3, i+4 , ... i+n\} If you choose the element i i there arise some cases. We will do this case by case.

_ Case 1 _ : After selecting i i , you select i + 1 i+1 , so now your next selection cannot be i + 2 i+2 or i 1 i-1 as the three elements will form an A.P. Now again there are sub-cases after selecting i + 1 i+1 SUB- CASE 1 : (You select ( i + 2 ) (i+2) ) Now after selecting i + 2 i+2 , we are bound to have the selection of elements lying between ( i ) (i) and ( i + 3 ) (i+3) otherwise we will get an A.P. So we see that we are limited to a 5 5 element selection. SUB-CASE 2 : ( You select ( i + 3 ) (i+3) ) By similar arguments, we are now bound to select the elements lying between ( i 2 ) (i-2) and ( i + 5 ) (i+5) , a 8 8 -element selection. Since we want the worst case scenario, we observe that for the selection { i 2 , i , i + 3 , i + 5 } \{ i-2 , i , i+3 , i+5 \} , there exists no such A.P. satisfying the conditions in the question.

Now with similar arguments used above we can explore the cases when Case 2 After selecting After selecting i i , you select i + 2 i+2 , and Case 3 After selecting i i , you select i + 3 i+3 . We have to keep in mind that more cases will not arise, like we cannot select i + 4 i+4 since then ( i + 1 ) , ( i + 2 ) , ( i + 3 ) (i+1) , (i+2) , (i+3) will form an A.P. After combining the results of all the cases, we find that N = 8 N=8 is the largest N N so that the conditions in the questions do not hold good. Thus we conclude that the smallest value of N N is N = 9 \boxed{N=9}

That's how I initially approached the problem. I was looking for a shorter (maybe more rigorous) approach which does not use brute-forcing or casework. But seeing the solutions posted, I don't think such a solution exists. You could have made it better by showing that N = 9 N=9 indeed satisfies the provided conditions though.

Sreejato Bhattacharya - 7 years, 8 months ago
Daniel Chiu
Sep 23, 2013

8 doesn't work since the subset {2,4,5,7} allows no progression. I like to visualize it as

_ O _ O O _ O _

Where the O's are in S S and the _'s are not. Every number below 8 can be shown not to work by removing spots from the right of my visualization.

Now, I will show 9 works. WLOG, say 5 is in S S . If it isn't, we can simply flip which elements are in S S . Notice that if (1,2,3,4) is in S S , then the corresponding element of (9,8,7,6) isn't.

? ? ? ? O ? ? ? ?

Now, we have two cases: Neither of 1 and 3 are in S S , or exactly one is (or we are done).

If neither are in, then 7 and 9 both are in, and we are done since 5,7,9 are in S S .

If one, say 3, is in, then we have

_ ? O ? O ? _ ? O

Now we see this is symmetrical to the case when 1 is in S S (reflect horizontally). Now, 4 cannot be in S S .

_ ? O _ O O _ ? O

Now we have 3,6,9 in S S , and so the answer is 9 \boxed{9} .

Nice use of symmetry. However, I believe there's a slight flaw that needs to be worked out: You're right to say that "if (1,2,3,4) is in S, then the corresponding element of (9,8,7,6) isn't" ... but the problem that you seem to use this as if it is an if-and-only-if. (For example, when you say "If neither are in, then 7 and 9 both are in".) But having said, I believe your basic plan is workable without too much more trouble.

Peter Byers - 7 years, 8 months ago

Log in to reply

Here's my suggestion: start with 2 and 8. They can't both be in S, and WLOG 2 is out. Then you can say that exactly one of 1 and 3 is in S. That gives two possibilities:

Case 1: _ _ O ? O ? ? ? ? Then 4 is out, but that means that either 1,4,7 are all out or else 3,5,7 are all in. Contradiction.

Case 2: O _ _ ? O ? ? ? ? Then 4 is in. So then 6 and 9 are both out, but so is 3: contradiction.

Peter Byers - 7 years, 8 months ago

Log in to reply

Yes, this works.

Daniel Chiu - 7 years, 8 months ago

Oh yes true enough I overlooked that.

Daniel Chiu - 7 years, 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...