Try this with a trapezium instead

Let S S be the set of lattice points ( x , y ) (x,y) with 1 x 20 1 \leq x \leq 20 and 1 y 30. 1 \leq y \leq 30. Let B S B \subseteq S be a subset of S S such that no set of 4 4 vertices in B B form a (non-degenerate) parallelogram. What is the largest possible number of elements in B ? B?

Details and assumptions

A figure is degenerate if it has 0 area.


The answer is 49.

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.

5 solutions

Jon Haussmann
Oct 28, 2013

We generalize the problem as follows: If counters are placed in the squares of an m × n m \times n grid, where m 2 m \ge 2 and n 2 n \ge 2 , then the largest number of counters that can be placed so that no four counters form a non-degenerate parallelogram is m + n 1 m + n - 1 .

Proof : Consider an m × n m \times n grid that contains at least m + n m + n counters. We call a row crowded if it contains at least two counters. Let r r be the number of crowded rows. There must be at least two crowded rows; otherwise, the number of counters would be at most n + ( m 1 ) = m + n 1 n + (m - 1) = m + n - 1 .

Each of the remaining m r m - r rows that are not crowded contain at most one counter, so the number of counters among all crowded rows is at least m + n ( m r ) = n + r m + n - (m - r) = n + r .

For each crowded row, we record the distance between the leftmost counter and every other counter in the same row. Each such distance is between 1 and n 1 n - 1 .

Let a i a_i be the number of counters in the i th i^{\text{th}} crowded row. Then we record a i 1 a_i - 1 distances for the i th i^{\text{th}} crowded row, so the total number of distances that we record is ( a 1 1 ) + ( a 2 1 ) + + ( a r 1 ) = a 1 + a 2 + + a r r , (a_1 - 1) + (a_2 - 1) + \dots + (a_r - 1) = a_1 + a_2 + \dots + a_r - r, which is at least ( n + r ) r = n (n + r) - r = n .

Hence, by the Pigeonhole Principle, two of the distances that we recorded are equal. The four counters that correspond to these equal distances form a non-degenerate parallelogram.

Furthermore, if we place a counter in each square in the first row and first column, then we obtain a grid with m + n 1 m + n - 1 counters, where no four counters form a non-degenerate parallelogram.

Great generalization :) That was how I set the question initially.

Any thoughts on how to approach this for such that the vertices doesn't form a trapezoid (US definition for a quadrilateral which has 1 set of parallel lines)?

Calvin Lin Staff - 7 years, 7 months ago

Log in to reply

If two pairs of points have the same slope between them (and aren't all collinear), they form a trapezoid. This might work.

Daniel Chiu - 7 years, 7 months ago

the answer such that the vertices does not form trapezoid is 31 for this question

ojas dhiman - 7 years, 7 months ago

Log in to reply

the answer such that the vertices does not form trapezoid is 31 for this question

Don't be so sure.

Peter Byers - 7 years, 7 months ago

Log in to reply

@Peter Byers .

........................

.

Peter Byers - 7 years, 7 months ago

@Peter Byers can you tell me any flaw in my answer

ojas dhiman - 7 years, 7 months ago

Log in to reply

@Ojas Dhiman Peter B. gave an example with 32 points above (although it took me a few seconds to figure out why the pattern of periods had 3 upvotes).

Basically,

Rows 1 and 3 have a point in column 1.

Row 2 has 30 points.

Daniel Chiu - 7 years, 7 months ago

Log in to reply

@Daniel Chiu @Daniel C.

(although it took me a few seconds to figure out why the pattern of periods had 3 upvotes)

:)

Thanks for pointing that out. I hadn't even noticed that it had gotten votes -- actually, I don't think I've even thought about this thread since then.

BTW, as I recall I also tried doing more than 32 (for example, having 2 above, 2 below, and removing only one from the horizontal row) but it seemed to be impossible (for the given dimensions). But I didn't take the time to try and prove it.

Peter Byers - 7 years, 7 months ago

Can you explain your thinking step by step?

Calvin Lin Staff - 7 years, 7 months ago

It works for m , n 1 m, n \geq 1 .

Michael Tong - 7 years, 7 months ago

Log in to reply

i say 31 for this question not in general

ojas dhiman - 7 years, 7 months ago
Mark Hennings
Oct 29, 2013

For any 1 y 30 1 \le y \le 30 let B y = { 1 x 20 ( x , y ) B } B_y = \{1 \le x \le 20 | (x,y) \in B\} and let C j = { 1 y 30 B y = j } 1 j 20 C_j \; = \; \{1 \le y \le 30 | |B_y| = j\} \qquad 1 \le j \le 20 noting that j = 1 20 C j 30 j = 1 20 j C j = B \sum_{j=1}^{20}|C_j| \le 30 \qquad \sum_{j=1}^{20}j|C_j| = |B| For any 2 j 20 2 \le j \le 20 and y C j y \in C_j , let B y = { x 1 , x 2 , , x j } B_y = \{x_1,x_2,\ldots,x_j\} where x 1 < x 2 < < x j x_1 < x_2 < \cdots < x_j . Then D y = { x k x 1 2 k j } D_y = \{x_k - x_1 | 2 \le k \le j\} is a set of j 1 j-1 distinct numbers between 1 1 and 19 19 . The sets D y D_y (for y C j y \in C_j and 2 j 20 2 \le j \le 20 ) must be disjoint, since otherwise we could find distinct y 1 , y 2 y_1,y_2 and distinct x 11 , x 12 B y 1 x_{11},x_{12} \in B_{y_1} and distinct x 21 , x 22 B y 2 x_{21},x_{22} \in B_{y_2} with the same nonzero difference x 11 x 12 = x 21 x 22 |x_{11} - x_{12}| = |x_{21} - x_{22}| in D y 1 D y 2 D_{y_1} \cap D_{y_2} , and hence obtain a nondegenerate parallelogram in B B (with one horizontal pair of sides). Thus we deduce that j = 2 20 ( j 1 ) C j 19 \sum_{j=2}^{20}(j-1)|C_j| \le 19 and so B = j = 1 20 j C j = j = 1 20 C j + j = 2 20 ( j 1 ) C j 30 + 19 = 49 |B| = \sum_{j=1}^{20}j|C_j| = \sum_{j=1}^{20}|C_j| + \sum_{j=2}^{20}(j-1)|C_j| \le 30 + 19 = 49

The set B = { ( 1 , y ) 1 y 30 } { ( x , 1 ) 2 x 20 } B = \big\{(1,y) \big| 1 \le y \le 30\big\} \cup \big\{(x,1) \big| 2 \le x \le 20\big\} is a concrete example of a set B B with 49 49 elements.

This argument can be generalised to m m rows and n n columns as well!

Mark Hennings - 7 years, 7 months ago

Note that the set S S can be interpreted as a 20 × 30 20 \times 30 rectangular grid. We shall prove by induction that for all a , b N > 1 a,b \in \mathbb{N}_{>1} , the largest possible number of lattice points inside any rectangular a × b a \times b grid such that no four of them form a parallelogram is a + b 1 a+b-1 . For the base case, we have to show that in a 2 × 2 2 \times 2 rectangular grid, the largest possible number of such points is 2 + 2 1 = 3 2+2-1= 3 . This is trivial, since if we choose 4 4 points, we get the 2 × 2 2 \times 2 rectangular grid itself, which is a parallelogram.

We shall now show that if our assertion is true for rectangular a × b a \times b grid and a rectangular ( a 1 ) × b (a-1) \times b grid, it will also be true for a rectangular ( a + 1 ) × b (a+1) \times b grid. Note that if we can prove this, we will be done, since by symmetry our claim will also hold true for a rectangular a × ( b + 1 ) a \times (b+1) grid, which will complete the induction step on a a and b b .

We have assumed that for a rectangular a × b a \times b grid, the largest number of such points is a + b 1 a+b-1 . Suppose we increase a a by 1 1 . We have to show that we can incorporate at most 1 1 lattice point in our set. On the contrary, assume we can take 2 2 more points in our selection. We divide the ( a + 1 ) × b (a+1) \times b grid into two rectangular grids: the first grid G 1 G_1 consisting of the first a a columns, and the second grid G 2 G_2 consisting of the last a a columns. Note that the dimensions of each these grids is a × b a \times b , and hence can accommodate at most a + b 1 a+b-1 lattice points. Also note that G 1 G 2 G_1 \cap G_2 is a ( a 1 ) × b (a-1) \times b rectangular grid, and can hence accommodate at most a + b 2 a+b-2 lattice points. Thus, atleast 3 3 points must lie within either the first column or the last column. Either way, the column with the greater number of points must have at least 3 2 = 2 \left \lceil \frac{3}{2} \right \rceil = 2 lattice points. WLOG assume the first column has more points than the last column. It is then easy to see that G 1 G_1 has more than a + b 1 a+b-1 lattice points, a contradiction. This completes our induction step.

It is easy to see that equality holds when we take the a + b 1 a+b-1 points in a L \text{L} or I \text{I} -shaped figure at the edge of the rectangular grid. Our desired answer will then be 20 + 30 1 = 49 20+30-1= \boxed{49} .

... WLOG assume the first column has more points than the last column. It is then easy to see that G 1 G_1 has more than a + b 1 a+b-1 lattice points, a contradiction.

The problem with this logic is that G 1 G 2 G_1\cap G_2 might contain fewer than a + b 2 a+b-2 lattice points.

Peter Byers - 7 years, 7 months ago

Log in to reply

Yes, sorry for that. My logic is incomplete.

Sreejato Bhattacharya - 7 years, 7 months ago
Marek Bernat
Nov 2, 2013

We can easily exhibit a set with 49 49 elements by simply picking points having either x = 1 x=1 or y = 1 y=1 . To show that this is the best possible, suppose B B is maximal and has 50 50 elements. We can assume that B B contains at least one element from each column (otherwise it's easy to add some point from that column obtaining a bigger set). Subtracting these 30 30 points (one per column), there must be 20 20 more points somewhere.

The idea now is that putting a point in any column will form a line with the point that we have already selected in the previous step. The length of such a line is between 1 1 and 19 19 . Therefore, by pigeonhole principle, there are two columns containing lines of the same length. But that's the same as saying that the four points from those lines form a parallelogram, in contradiction with B B being allowed.

Interesting problem. My first guess is that the answer is 49. Take all points where the x or y coordinate is 1. That gives you a column with 30 points, plus a row with 20, less the one point where they intersect. 30+20-1=49.. Add any other point and you get a rectangle.

If you were approaching this problem for a rectangle, as opposed to a parallelogram, the answer is not 49.

For example, in a 3 × 3 3 \times 3 grid, you would have said 3 + 3 1 = 5 3 + 3 - 1 = 5 , but we can have 6, like so:

× × × × × × \begin{array} { | l | l | l | } \hline \times & \times & \\ \hline \times & & \times \\ \hline & \times & \times \\ \hline \end{array}

It is worthwhile to think about the rectangle case too.

Calvin Lin Staff - 7 years, 7 months ago

How do you ensure this is the minimum possible? There can be numerous other cases, which might allow for a parallelogram (maybe not a rectangle).

However, that's the idea I first had in mind when I solved this problem. Your thinking is on the right track. :)

Sreejato Bhattacharya - 7 years, 7 months ago

Log in to reply

Indeed. A specific equality case is easy to guess, but hard to justify.

Calvin Lin Staff - 7 years, 7 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...