Mordellbles

Very direct inspiration (and the discussion with @Pi Han Goh and @David Vreken)

You find some pebbles and arrange them in groups of increasing diamond-shapes as follows:

You notice that number of pebbles in the first grouping is a perfect cube, ( 1 3 = 1 1^3=1 ), but run out of pebbles before you can find another one.

If the pattern continues, how many other groupings would also use a cube number of pebbles?

Finitely many others Infinitely many others None (it only works with one pebble)

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.

2 solutions

Chris Lewis
Mar 18, 2021

The n th n^\text{th} diamond pattern uses n 2 + ( n 1 ) 2 n^2+(n-1)^2 pebbles, as below:

So the equation we want to solve in positive integers is m 3 = 2 n 2 2 n + 1 m^3=2n^2-2n+1

Multiply by 8 8 : 8 m 3 = 16 n 2 16 n + 8 = ( 4 n 2 ) 2 + 4 8m^3=16n^2-16n+8=(4n-2)^2+4

Put x = 2 m x=2m and y = 4 n 2 y=4n-2 to get y 2 = x 3 4 y^2=x^3-4

which is an example of Mordell's equation . Better still, this precise example is solved on page 6 of this awesome paper by Keith Conrad. He shows the only solutions in integers to the above equation are ( x , y ) = ( 2 , ± 2 ) (x,y)=(2,\pm2) and ( x , y ) = ( 5 , ± 11 ) (x,y)=(5,\pm11) .

The first of these leads to solutions in the original equation ( m , n ) = ( 1 , 0 ) (m,n)=(1,0) and ( m , n ) = ( 1 , 1 ) (m,n)=(1,1) (but we can't have n = 0 n=0 ). The second gives non-integer values for ( m , n ) (m,n) so is discounted.

Hence there are no other solutions .

Ha, I cheated and asked MAGMA for the integral points on the elliptic curve y 2 = x 3 4. y^2 = x^3-4. That "elementary" solution in Conrad's paper looks a little tricky!

Patrick Corn - 2 months, 3 weeks ago

Log in to reply

It's not cheating that much - you converted it into canonical form, then looked it up; I can't claim to have done more than that! Do you know of any online searchable resources for this type of Diophantine equation?

By the way, does MAGMA cope with y 2 = x 3 24 y^2=x^3-24 ? (It's another example in that paper)

Chris Lewis - 2 months, 3 weeks ago

Log in to reply

Yes, it instantly says there are no solutions. There are general algorithms for finding all integral points on a Weierstrass model of an elliptic curve, implemented in MAGMA--I think it's the usual situation where they are potentially/theoretically extremely slow, but in practice, for any specific elliptic curve you care about, they're quite fast.

Patrick Corn - 2 months, 3 weeks ago

Log in to reply

@Patrick Corn Ah, sorry, I mistyped the equation, I meant y 2 = x 3 + 24 y^2=x^3+24 which does have solutions; it may be an interesting test because some are quite large. Wolfram|Alpha doesn't find all of them.

Chris Lewis - 2 months, 3 weeks ago

Log in to reply

@Chris Lewis It takes about half a second to find x = 2 , 1 , 10 , 8158. x= -2, 1, 10, 8158. You can try it yourself: go to http://magma.maths.usyd.edu.au/calc and put in E := EllipticCurve([0,24]); E; IntegralPoints(E);

Patrick Corn - 2 months, 3 weeks ago

Log in to reply

@Patrick Corn Great stuff, thank you for that!

Chris Lewis - 2 months, 3 weeks ago
Paul Romero
Mar 21, 2021

@Chris Lewis , @David Vreken, @Pi Han Goh: Please, correct me if I am wrong:

Let's a k + 1 a_{k + 1} to be the general term of this series, then

(1) a k + 1 = 1 + k × ( 4 k + 4 ) 2 = 1 + k × ( 2 k + 2 ) = 1 + 2 k 2 + 2 k a_{k + 1} = 1 + \frac{k\times (4k + 4)}{2} = 1 + k\times(2k +2) = 1 + 2k^{2} + 2k

Now suppose that a k + 1 = m 3 a_{k + 1} = m^{3} , then

(2) m 3 = 1 + 2 k + 2 k 2 m^{3} = 1 + 2k + 2k^{2}

Now I will add on both sides of the equation k + k 2 + k 3 k + k^{2} + k^{3} , then

(3) m 3 + k + k 2 + k 3 = 1 + 3 k + 3 k 2 + k 3 = ( k + 1 ) 3 m^{3} + k + k^{2} + k^{3} = 1 + 3k + 3k^{2} + k^{3} = (k + 1)^{3} ,

which means that

(4) k + 1 > k m k + 1 > k \geq m

Now, we know that

(5) a + b + c 3 × ( a b c ) 1 3 a + b + c \geq 3\times (abc)^{\frac{1}{3}} so from (1) and (5)

(6) m 3 = 1 + 2 k + 2 k 2 3 × 4 1 3 × k m^{3} = 1 + 2k + 2k^{2} \geq 3\times 4^{\frac{1}{3}} \times k

which is a contradiction with (4). Therefore, we can get only a perfect cube with 1 pebble

Wow, this looks great. I tried but didn't get anywhere with adding k 3 + k 2 + k k^3+k^2+k , looks like you've got a lot further.

One question (and I may have just missed something obvious here, I haven't had a chance to write this down and go through it properly yet), but why does k m + 1 ( m 1 ) 2 + ( m 1 ) + 1 k-m+1|(m-1)^2+(m-1)+1 imply k m + 1 = 1 k-m+1=1 ? You show k m k \ge m at the beginning, so k m + 1 k-m+1 is positive, but why couldn't we have (say) m = 2 m=2 and k = 4 k=4 ? Then k m + 1 = 3 k-m+1=3 divides ( m 1 ) 2 + ( m 1 ) + 1 = 3 (m-1)^2+(m-1)+1=3 .

The same works in general if m = 3 a 1 m=3a-1 and k = 3 a + 1 k=3a+1 . Of course, these examples don't give solutions to the full problem (we know there aren't any others) but I'm not sure how that one deduction step works.

Again, sorry if I've missed something!

Chris Lewis - 2 months, 3 weeks ago

Log in to reply

I think what you said is wise. I will check the solution. Remember that I asked you, guys, for advice.

Paul Romero - 2 months, 3 weeks ago

k - m + 1 has to divide, at least, one of the term at the left

Paul Romero - 2 months, 3 weeks ago

@Chris Lewis : I am not done yet, but check my first edition, please, and I apologize if I am annoying you.

Paul Romero - 2 months, 3 weeks ago

Log in to reply

Hi, definitely not annoying! As I mentioned in the discussion on Pellbles, I'm really interested in this question (and the related ones - make sure you have a look at Legellbles).

In the latest edit here you have m 3 3 4 3 k m^3\ge 3 \sqrt[3]{4} k and claim this contradicts k m k\ge m but again I don't see why it should - there's no problem with m k 4 3 k m 3 m\le k \le \sqrt[3]{4} k \le m^3 .

One thing to think about here is that the equation does have solutions - in terms of your ( m , k ) (m,k) , the pairs ( 1 , 0 ) (1,0) and ( 1 , 1 ) (1,-1) both work.

This rules out a lot of methods; for example, modulo arithmetic can be used to either show there are no solutions, or infinitely many; so more is needed. The Conrad paper I linked to in my solution shows that quite often factoring and using coprimality is useful, which makes me think your original solution is a good way to go.

As I say, happy to keep discussing!

Chris Lewis - 2 months, 3 weeks ago

Log in to reply

That's why I erased :)

Paul Romero - 2 months, 3 weeks ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...