My objective is linear

a a and b b are positive integers such that

{ 7 b 11 a 1000 a b + a 2 = 1 + b 2 . \begin{cases} 7b-11a \leq 1000 \\ ab + a^2= 1+b^2. \\ \end{cases}

What is the largest possible value of 7 b 11 a 7b-11a ?

Details and assumptions

There is no restriction that a a or b b must be less than 1000.


The answer is 521.

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

James Aaronson
Sep 9, 2013

First, drop the restriction that a a and b b are positive and the inequality condition.

Now, all solutions satisfy:

a |a| and b |b| are consecutive Fibonacci numbers, such that a a is of odd index (1,2,5,13,34...) and b b is of even index (0,1,3,8,21,...). If b > a b>a , then a a and b b are of the same sign, and if b < a b<a , then a a and b b are of opposite sign.

To prove this, suppose it's not true, and take the solution with smallest a + b |a| + |b| - clearly neither of them can be zero, as the only such solution is (1,0) which is of the above form. It is also easy to check that a |a| and b |b| must both be greater than 3, by solving the quadratic in the cases that a |a| or b |b| are at most 3 and noting that all solutions are as above. Then, we can observe that in any solution with a |a| and b |b| greater than 3, a < 2 b |a| < 2|b| and b < 2 a |b| < 2|a| by supposing otherwise and noting that one side is too large, and that there is clearly no solution with a = b a = b .

(e.g. to prove a < 2 b |a| < 2|b| , suppose otherwise. So a b < 0 ab < 0 , and then a 2 2 a b > 1 + b 2 + a b a^2 \geq 2|ab| > 1 + b^2 + |ab| .)

Suppose that a > b |a| > |b| , and WLOG b > 0 b > 0 since we can multiply both a a and b b by 1 -1 at our leisure. Clearly, a b ab must be less than 1 and so negative, so a a is negative. By Vieta's formula, there is another solution with the same value of b b , namely ( ( b + a ) , b ) (-(b+a),b) . Since a < 2 b |a| < 2|b| , this is a smaller solution, so it must be of the above form by the induction hypothesis. However, that is impossible because that means that our original solution was also of the above form, contradicting our assumption. If a < b |a| < |b| , we run a similar argument, except we consider the equation as a quadratic in a a .

Therefore, all solutions are of the above form. Since a a and b b are positive, we have only the solutions in which a = F 2 k 1 a = F_{2k-1} , b = F 2 k b = F_{2k} .

It is easy to prove by induction that 7 b 11 a 7b - 11a takes alternating values in the Lucas sequence, 1, 4, 11, 29, ..., and the largest value less than 1000 is 521.

Moderator note:

Great job!

I didn't feel like submitting a solution since I don't think my answer is very elegant, but what I did was write a a in terms of b b using the quadratic formula to get 7 b 11 a = 25 b 11 5 b 4 + 4 2 1000 7b - 11a = \frac {25b - 11\sqrt{5b^4 + 4}}{2} \leq 1000 and then after finding the first few b b that makes 5 b 2 + 4 5b^2 + 4 a perfect square I noticed that they were alternating fibonacci numbers. I'm glad that now I know why that is because of your solution!

Michael Tong - 7 years, 9 months ago

Log in to reply

My original solution went a lot like your solution. I said b b is an integer, so the discriminant, 5 a 2 4 5a^2 - 4 , is a square. If you're more diligent than I am, you might solve this Pell equation straight away, but I chose the route of finding the first four solutions for a a , which were 1 , 2 , 5 , 13 1, 2, 5, 13 . Since we know from the theory of Pell equations that the solution is a second order linear recurrence, we know we're taking alternating Fibonacci numbers.

James Aaronson - 7 years, 9 months ago

Nice done

Vali Dobre - 7 years, 9 months ago

this is what I did

Purvam Modi - 7 years, 9 months ago

Same problem with me. I observed that the possible values of b b were actually even positioned fibonacci numbers. I progressed with that conjecture, and it ultimately gave the correct answer. However I could not provide justification why it was true. Anyways, this solution is great!

Sreejato Bhattacharya - 7 years, 9 months ago

Nicely done! This is a great solution of a hard problem. I am sure some people will want to know more details, hopefully you can find the time to answer!

Alexander Borisov - 7 years, 9 months ago

Log in to reply

"hard" varies a lot from person to person, this is one of the great things about maths :) Actually this was the single fastest problem out of this week's for me. On the other hand I would be embarrassed to admit how long I spent on the tripping triples...

Matt McNabb - 7 years, 9 months ago

Nice.

One thought. You say: "Since a < 2 b |a|<2|b| , this is a smaller solution" but do you even need a < 2 b |a|<2|b| ? Specifically, could you conclude that it's a smaller solution just because 0 < b < a 0 < b<-a means that a < b + a < 0 a < b+a<0 ?

Peter Byers - 7 years, 9 months ago
Matt McNabb
Sep 15, 2013

Rearranging the second condition: b 2 a 2 = a b 1 b^2 - a^2 = ab - 1

We see that we must have b > a b>a , except for the trivial solution 1 , 0 1,0 and 1 , 1 1,1 .

Let b = a + c b = a+c . Substituting this in the above equation ( a + c ) 2 a 2 = a ( a + c ) 1 2 a c + c 2 = a 2 + a c 1 a 2 c 2 = a c + 1 \begin{aligned} (a+c)^2 - a^2 &= a(a+c)-1 \\ 2ac + c^2 &= a^2 + ac - 1 \\ a^2 - c^2 &= ac + 1 \end{aligned}

Here we see, again that a > c a>c since a c + 1 ac+1 must be positive. So, subsituting a = c + d a = c+d in the last equation:

( c + d ) 2 c 2 = c ( c + d ) + 1 2 c d + d 2 = c 2 + c d + 1 c 2 d 2 = c d 1 \begin{aligned} (c+d)^2 - c^2 &= c(c+d) + 1 \\ 2cd + d^2 &= c^2 + cd + 1 \\ c^2 - d^2 &= cd - 1 \end{aligned}

But this is actually the same equation that we started with. So we have shown that if a solution ( a , b ) (a,b) exists, then there is a smaller solution ( c , d ) (c,d) . In fact, if ( c , d ) (c,d) is a solution then the next-biggest solution is ( c + d , 2 c + d ) (c+d, 2c+d) .

This means that we can work back from the smallest solution c = 1 , d = 1 c=1, d=1 in order to generate all possible solutions. (The special case ( 1 , 0 ) (1,0) gives ( 1 , 1 ) (1,1) anyway so we don't have to worry about that).

This gives us ( 1 , 1 ) , ( 2 , 3 ) , ( 5 , 8 ) , ( 13 , 21 ) , . . (1,1), (2,3), (5,8), (13, 21), ..

i.e. F 2 n 1 , F 2 n F_{2n-1}, F_{2n} .

To quickly find the solution from here we note that as n n gets large, the ratio between successive Fibonacci numbers approaches the golden ratio ϕ \phi . So we would want 7 ϕ a 11 a < 1000 7 \phi a - 11a \lt 1000 , i.e. a ( 7 ϕ 11 ) < 1000 a(7 \phi - 11) < 1000 , i.e. a < 3066 a \lt 3066 .

Checking a list of Fibonacci numbers shows that a = 1597 , b = 2584 a=1597, b=2584 is the largest pair meeting this condition, and 7 b 11 a 7b - 11a here is 521 \boxed{521} .

I could have been more formal about when the descent ends: we keep doing it until we end up with c c not being greater than d d , which can only be ( 1 , 1 ) o r ( 0 , 1 ) (1,1) or (0,1) .

Matt McNabb - 7 years, 9 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...