Egyptian Fractions & Square

1 a + 1 b + 1 c = x 2 n \dfrac{1}{a}+\dfrac{1}{b}+\dfrac{1}{c} = \dfrac{x^2}{n}

Let a , b , c , x a,b,c,x be pairwise coprime positive integers greater than 1, satisfying the equation above for some natural number n n coprime to x x .

What is the least possible value of a + b + c a+b+c ?


The answer is 20.

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.

1 solution

Leonel Castillo
May 26, 2018

First of all, the equation can be written as a b + b c + c a a b c = x 2 n \frac{ab + bc + ca}{abc} = \frac{x^2}{n} . I will now prove that gcd ( a b c , a b + b c + c a ) = 1 \gcd(abc,ab + bc + ca) = 1 :

Suppose that d a b c , a b + b c + c a d | abc , ab + bc + ca . Then d a b c + b c 2 + c 2 a d b c 2 + c 2 a d b 2 c 2 + b c 2 a d b 2 c 2 d | abc + bc^2 + c^2 a \implies d | bc^2 + c^2a \implies d | b^2c^2 + bc^2 a \implies d | b^2 c^2 . If d d does not divide b c bc then d d is divided by a higher power of a prime that divides b c bc but then d d would not divide a b c abc . This is a contradiction so d b c d | bc . By symmetry (or by repeating the procedure) we also get that d a b , b c , c a d | ab, bc , ca . So d gcd ( a b , b c ) = b d | \gcd(ab,bc) = b and d gcd ( b c , a c ) = c d | \gcd(bc, ac) = c so d gcd ( b , c ) = 1 d | \gcd(b,c) = 1 so d = 1 d=1 .

Because the expression of a positive rational number as a quotient of relatively prime positive integers is unique, a b + b c + c a a b c = x 2 n a b + b c + c a = x 2 , a b c = n \frac{ab + bc + ca}{abc} = \frac{x^2}{n} \implies ab + bc + ca = x^2, abc = n . Notice that this guarantees that gcd ( n , x 2 ) = 1 \gcd(n, x^2) = 1 . So any solution of the equation a b + b c + c a = x 2 ab + bc + ca = x^2 with relatively prime variables will yield a solution to the original problem by simply taking n = a b c n = abc . Solving this diophantine equation in general terms is relatively easy as a b + b c + c a = x 2 ( b + a ) ( c + a ) = x 2 + a 2 ab + bc + ca = x^2 \implies (b+a)(c+a) = x^2 + a^2 so by computing any particular value of x 2 + a 2 x^2 + a^2 and then factorizing it one would find a solution to this equation, but this will not guarantee that the solution will meet our conditions.

The rest of my solution is heuristic, there is probably a better solution :

As x 2 0 , 1 m o d 4 x^2 \equiv 0,1 \mod 4 we can break it down into cases. Suppose a b + b c + c a 0 m o d 4 ab + bc + ca \equiv 0 \mod 4 . Suppose that a a is even, then b , c b,c are odd. But we have a ( b + c ) + b c 0 m o d 4 b c 0 m o d 4 a(b + c) + bc \equiv 0 \mod 4 \implies bc \equiv 0 \mod 4 so at least another variable has to be even, this is a contradiction. Suppose no variable is even, then we would have ± 1 ± 1 ± 1 0 m o d 4 \pm 1 \pm 1 \pm 1 \equiv 0 \mod 4 . Regardless of how you choose your signs, this is impossible. This guarantees that x 2 1 m o d 4 x^2 \equiv 1 \mod 4 .

Suppose that none of the variables a , b , c a,b,c are even. If a , b , c 1 m o d 4 a,b,c \equiv -1 \mod 4 then a b + b c + c a 3 m o d 4 ab + bc + ca \equiv 3 \mod 4 which is impossible so suppose c = 1 c=1 , then a b + b + a 1 m o d 4 ab + b + a \equiv 1 \mod 4 . If a 1 a \equiv -1 then 1 1 m o d 4 -1 \equiv 1 \mod 4 . Impossible so suppose a 1 a \equiv 1 , then 2 b 0 m o d 4 2b \equiv 0 \mod 4 which implies b b is even, a contradiction. This means that at least one (and at most one) of the variables has to be even. Let's suppose that a a is even. Then a ( b + c ) + b c 1 m o d 4 b c 1 m o d 4 b c m o d 4 a(b+c) + bc \equiv 1 \mod 4 \implies bc \equiv 1 \mod 4 \implies b \equiv c \mod 4 . In short, all solutions must have one even variable and the two other variables must be both congruent to either 1 1 or 1 m o d 4 -1 \mod 4 .

Using the previous result to guide the search, if we restrict ourselves to small numbers (specifically, numbers no larger than 10) then by trial and error one finds the solution a = 10 , b = 3 , c = 7 a=10, b=3, c=7 with x = 11 x = 11 by making a list of relatively prime triples among the naturals 2 , 3 , 4 , . . . , 10 2,3,4,...,10 that also satisfy the previous congruences. These are:

( 2 , 3 , 7 ) , ( 2 , 5 , 9 ) , ( 4 , 3 , 7 ) , ( 4 , 5 , 9 ) , ( 8 , 3 , 7 ) , ( 8 , 5 , 9 ) , ( 10 , 3 , 7 ) (2,3,7), (2,5,9), (4,3,7), (4,5,9), (8,3,7), (8,5,9), (10,3,7) . You can easily find them by noticing that a = 2 , 4 , 6 , 8 , 10 a = 2,4,6,8,10 and b , c b,c are either 3 , 7 3,7 or 5 , 9 5,9 . With this computation we find that the only solution in small variables is the one mentioned with a + b + c = 10 + 3 + 7 = 20 a+b+c = 10 + 3 + 7 = 20 . We know that this is the smallest value of a + b + c a+b+c because any other solution must have at least two variables with two digits so a + b + c > 10 + 10 = 20 a + b + c > 10 + 10 = 20 or have a variable bigger than 10 10 so a + b + c > 10 + 3 + 7 = 20 a + b + c > 10 + 3 + 7 = 20 .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...