What superpowers would you have?

Find the least positive integer k k for which there exist two distinct positive integers a a and b b such that k 3 = a 4 + b 4 + ( a + b ) 4 . k^3=a^4+b^4+(a+b)^4.


The answer is 392.

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

Mark Hennings
Oct 14, 2013

All variables are assumed to represent integers. We want to solve k 3 = 2 ( a 4 + 2 a 3 b + 3 a 2 b 2 + 2 a b 3 + b 4 ) = 2 ( a 2 + a b + b 2 ) 2 k^3 \; = \; 2(a^4 + 2a^3b + 3a^2b^2 + 2ab^3 + b^4) \; = \; 2(a^2 + ab + b^2)^2 Thus k k must be even, which implies that ( a 2 + a b + b 2 ) 2 (a^2+ab+b^2)^2 is a multiple of 4 4 , and hence a 2 + a b + b 2 a^2+ab+b^2 is even. This implies that both a a and b b are even, and hence k = 2 x k = 2x , a = 2 y a = 2y , b = 2 z b = 2z where x 3 = 4 ( y 2 + y z + z 2 ) 2 x^3 \; = \; 4(y^2 + yz + z^2)^2 This in turn implies that x x is even, and hence that y 2 + y z + z 2 y^2+yz+z^2 is even, and hence that y y and z z are both even. Thus x = 2 X x = 2X , y = 2 Y y = 2Y , z = 2 Z z = 2Z where X 3 = 8 ( Y 2 + Y Z + Z 2 ) 2 X^3 \; = \; 8(Y^2 + YZ + Z^2)^2 which implies that X X is even, so that X = 2 U X = 2U where U 3 = ( Y 2 + Y Z + Z 2 ) 2 U^3 \; = \; (Y^2 + YZ + Z^2)^2 From this it is clear that U = V 2 U = V^2 and Y 2 + Y Z + Z 2 = V 3 Y^2 + YZ + Z^2 = V^3 .

The smallest possible solution for V V of the equation V 3 = Y 2 + Y Z + Z 2 V^3 \; = \; Y^2 + YZ + Z^2 with Y , Z Y,Z distinct positive integers is V = 7 V=7 (with Y = 14 Y=14 and Z = 7 Z=7 ), and so the smallest possible value of k = 8 U = 8 V 2 k=8U=8V^2 is 8 × 49 = 392 8\times49=392 .

The fact that V = 7 V=7 can be established by showing that 4 × 1 3 , 4 × 2 3 , , 4 × 6 3 4\times1^3,4\times2^3,\ldots,4\times6^3 cannot be written as ( 2 Y + Z ) 2 + 3 Z 2 (2Y+Z)^2 + 3Z^2 for distinct positive integer Y , Z Y,Z . To do this, just run through the finite number of possible Z Z values, checking that no Y Y value is possible. For example, if ( 2 Y + Z ) 2 + 3 Z 2 = 4 × 3 3 = 108 (2Y+Z)^2 + 3Z^2 = 4\times3^3 = 108 , then Z = 1 ( 2 Y + 1 ) 2 = 105 Z = 2 ( 2 Y + 2 ) 2 = 96 Z = 3 ( 2 Y + 3 ) 2 = 81 Y = 3 = Z Z = 4 ( 2 Y + 4 ) 2 = 60 Z = 5 ( 2 Y + 5 ) 2 = 33 Z 6 ( 2 Y + Z ) 2 0 \begin{array}{rcl} Z = 1 & \Rightarrow & (2Y+1)^2 = 105 \\ Z=2 & \Rightarrow & (2Y+2)^2 = 96 \\ Z=3 & \Rightarrow & (2Y+3)^2 = 81 \; \Rightarrow \; Y = 3 = Z \\ Z=4 & \Rightarrow & (2Y+4)^2 = 60 \\ Z=5 & \Rightarrow & (2Y+5)^2 = 33 \\ Z\ge6 & \Rightarrow & (2Y+Z)^2 \le 0 \end{array}

Great solution that is well explained. How did you come up with it?

Calvin Lin Staff - 7 years, 7 months ago

Log in to reply

The main trick was to expand/collect terms in the expression a 4 + b 4 + ( a + b ) 4 a^4+b^4+(a+b)^4 , and spot the factorisation 2 ( a 2 + a b + b 2 ) 2 2(a^2+ab+b^2)^2 . The 12321 12321 coefficient pattern is symptomatic of a three-term square.

After that, identifying as many common factors of 2 2 as possible helped to cut the remaining problem down to size. There are two advantages in doing this:

(a) if you remove common factors (like 2 2 ) there may be coprime-like arguments that can be applied to the residual problem,

(b) the more common factors you identify, the smaller the range of possible answers becomes. The answer for k k had to be less than 1000 1000 , so since k = 8 V 2 k = 8V^2 we know that V V can be no bigger than 11 11 . This means that any remaining checking, if necessary, could be done by hand.

Argument (b) is, of course, only applicable to Brilliant problems, where the size of answer is known in advance!

Mark Hennings - 7 years, 7 months ago

Log in to reply

Good solution, but it was actually possible to find the solution directly, without checking cases. For that, it's enough to note that the quadratic form is actually a norm on Eisenstein integers. So by using cubic roots of unity you can factorize it complete the solution.

Marek Bernat - 7 years, 7 months ago

A small typo. Line 8 should read : "and hence that y y and z z are both even".

Mark Hennings - 7 years, 7 months ago

Fantastic solution and neatly written. Keep this up

indulal gopal - 7 years, 8 months ago

brilliant

basil babu - 7 years, 8 months ago

impressive solution

Lay Yash - 7 years, 7 months ago

I obtained the same solution by, after getting to the V^3= Y^2+YZ+Z^2 stage, letting GCF(Y,Z)=d, Y=dy, Z=dz, V^3=d^2(y^2+yz+z^2), then choose smallest distinct y,z: y=1, z=2, and d=1+1x2+2^2=7. Of course, we still need to check that no smaller values are possible.

Yezi Joy - 7 years, 7 months ago

Did it the same way! :)

A Brilliant Member - 7 years, 7 months ago

I did all the steps until it comes to the last part (checking each possible Z value)\ :[

Wei Jie Tan - 7 years, 7 months ago
Adrabi Abderrahim
Oct 19, 2013

let k 3 = a 4 + b 4 + ( a + b ) 4 k^3 = a^4 + b^4 + (a+b)^4 so:

k 3 = 2 ( a 2 + a b + b 2 ) 2 k^3 = 2(a^2 + ab + b^2)^2

then k k need to be even so let ( 1 ) , k = 2 n (1), k = 2n where n n is positive integer, by substituting we've:

k 3 = 2 ( a 2 + a b + b 2 ) 2 k^3 = 2(a^2 + ab + b^2)^2

become

( 2 n ) 3 = 2 3 n 3 = 2 ( a 2 + a b + b 2 ) 2 (2n)^3 = 2^3n^3 = 2(a^2 + ab + b^2)^2

divide both sides by 2 2

2 2 n 3 = ( a 2 + a b + b 2 ) 2 2^2n^3 = (a^2 + ab + b^2)^2

let ( 2 ) , n = m 2 (2), n = m^2 where m m is positive integer then:

2 2 n 3 = 2 2 ( m 2 ) 3 = 2 2 ( m 3 ) 2 = ( 2 m 3 ) 2 = ( a 2 + a b + b 2 ) 2 2^2n^3 = 2^2(m^2)^3 = 2^2(m^3)^2 = (2m^3)^2 = (a^2 + ab + b^2)^2

add root square to both sides (because m , a , b m, a, b is positive integers) so:

2 m 3 = a 2 + a b + b 2 2m^3 = a^2 + ab + b^2

for a 2 + a b + b 2 a^2 + ab + b^2 can be even, a , b a,b must be even so: let a = 2 x , b = 2 y a = 2x, b=2y and we've a , b a,b are distinct positive integers, without loss of generality let b = 2 y = a + 2 z = 2 x + 2 z = 2 ( x + z ) b = 2y = a + 2z = 2x + 2z = 2(x+z) then:

2 m 3 = a 2 + a b + b 2 2m^3 = a^2 + ab + b^2

become

2 m 3 = ( 2 x ) 2 + ( 2 x ) ( 2 ( x + z ) ) + ( 2 ( x + z ) ) 2 2m^3 = (2x)^2 + (2x)(2(x+z)) + (2(x+z))^2

divide the both sides by 2 2 and factorize:

m 3 = 2 ( 3 x 2 + 3 x z + z 2 ) m^3 = 2(3x^2 + 3xz + z^2) and let again ( 3 ) , m = 2 l (3), m = 2l where l l is positive integer so:

m 3 = 2 ( 3 x 2 + 3 x z + z 2 ) m^3 = 2(3x^2 + 3xz + z^2)

( 2 l ) 3 = 2 ( 3 x 2 + 3 x z + z 2 ) (2l)^3 = 2(3x^2 + 3xz + z^2)

l 3 = ( 3 x 2 + 3 x z + z 2 ) / 4 l^3 = (3x^2 + 3xz + z^2)/4

for 3 x 2 + 3 x z + z 2 3x^2 + 3xz + z^2 be multiple of 4 4 must the both x , z x,z be even also unnecessary for x , z x,z be distinct positive integers, so when:

x = z = 2 , l 3 = ( 3 x 2 + 3 x z + z 2 ) / 4 = 7 = 1 × 1 × 7 x = z = 2, l^3 = (3x^2 + 3xz + z^2)/4 = 7 = 1\times1\times7

x = z = 4 , l 3 = ( 3 x 2 + 3 x z + z 2 ) / 4 = 28 = 2 × 2 × 7 x = z = 4, l^3 = (3x^2 + 3xz + z^2)/4 = 28 = 2\times2\times7

...

x = z = 14 , l 3 = ( 3 x 2 + 3 x z + z 2 ) / 4 = 343 = 7 × 7 × 7 x = z = 14, l^3 = (3x^2 + 3xz + z^2)/4 = 343 = 7\times7\times7

we've k = 2 ( ( 2 × l ) 2 ) k = 2((2\times l)^2) from ( 1 ) , ( 2 ) , ( 3 ) (1), (2), (3) where l = 7 l = 7 then k = 392 k = 392

Angel Leon
Oct 14, 2013

for those of us who think in code

public class LeastPositiveK {
public static void main(String[] args) {
    for (int k = 1; k < 1000; k++) {
        for (int a = 1; a <= 1000; a++) {
            for (int b = 1; b <= 1000; b++) {
                int kCube = k*k*k;
                int a4 = a*a*a*a;
                int b4 = b*b*b*b;
                int ab = a+b;
                int ab4 = ab*ab*ab*ab;
                int sumRight = a4+b4+ab4;
                if (sumRight == kCube && a != b) {
                    System.out.println("k="+k+", a="+a+", b="+b);
                }
            }
        }
    }
}
}

this is a kind of cheating...

Cuong Doan - 7 years, 8 months ago

This can be posted as a solution as long as you prove that your code isn't faulty and you didn't make any assumptions. As they said in the Brilliant FAQ, you probably aren't seeing the elegant way to solve things.

Anton Than Trong - 7 years, 8 months ago

What program did you use?

Edward Jiang - 7 years, 8 months ago

Log in to reply

that's just a java program. coded it first in python but it was too slow.

Angel Leon - 7 years, 8 months ago

Log in to reply

You could have speeded your algorithm up quite a bit, since k 3 > 2 a 4 , 2 b 4 k^3 > 2a^4,2b^4 . Thus you only needed to check for 2 a 149 2 \le a \le 149 and 1 b a 1 1 \le b \le a-1 .

Mark Hennings - 7 years, 7 months ago

Log in to reply

@Mark Hennings this is what I'm looking for when I post these algorithmic solutions. Thank you!

Angel Leon - 7 years, 7 months ago

Log in to reply

@Angel Leon Can you please translate the code so it can be implemented into Pari/GP?

Edward Jiang - 7 years, 7 months ago

Log in to reply

@Edward Jiang I'm sorry I'm afraid I don't know the syntax of Pari/GP, just checked it out, very interesting project for mathematicians.

Angel Leon - 7 years, 7 months ago

I'll try to post next one in C

Angel Leon - 7 years, 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...