Tartar 5
Tartar 5
Tartar 5
Your question about Fibonacci numbers (with F0 = 0 and F1 = 1, giving the sequence 1, 1, 2, 3, 5, 8,
13, 21, 34, etc) is a consequence of the formula
2
since one deduces that F2n+1 = (Fn2 + Fn+1
2
)2 = (Fn+1
2
− Fn2 )2 + (2Fn Fn+1 )2 , and one may replace Fn+1
2
− Fn2
by (Fn+1 − Fn ) (Fn+1 + Fn ) = Fn−1 Fn+2 . One proves (1) by induction, since both formulas are true for
n = 1, and assuming the formulas true for n, one has
and
2
F2n+4 = F2n+2 + F2n+3 = Fn+2 − Fn2 + Fn+1
2 2
+ Fn+2 2
= 2Fn+2 2
+ Fn+1 − (Fn+2 − Fn+1 )2
2
(3)
= Fn+2 + 2Fn+2 Fn+1 = (Fn+2 + Fn+1 )2 − Fn+1
2 2
= Fn+3 2
− Fn+1 .
1
2 2 2 2 2
a = m n, and c = n +m 2 . Then the condition is that A B = 4(A + B + C), i.e. g m n(n2 −m ) = 4g (n2 + m n),
n−m
or g m 2 = 4. The case g = 4 gives m = 1 and n = 3, i.e. (a, b, c) = 4(3, 4, 5) = (12, 16, 20); the case
g = 2 gives m = 1 and n = 5, i.e. (a, b, c) = 2(5, 12, 13) = (10, 24, 26); the case g = 1 gives m = 1 and n = 9,
i.e. (a, b, c) = (9, 40, 41).
Your problem 5 (Putnam1966-B1): Let a convex polygon P be contained in a square of side one. Show that
the sum of the squares of the sides of P is less than or equal to 4.
I had not written a solution before: One considers the smallest rectangle with sides parallel to the axes x and
y defined by the square; if the rectangle has size a × b, one shows that the sum of the squares of the sides of P
is less than or equal to 2(a2 + b2 ), which is ≤ 4, since 0 < a, b ≤ 1. By Pythagoras, the square of a side is the
sum of the square of its projection on the x axis and the square of its projection on the y axis; the projections
in x cover twice the interval of length a, and if `1 + . . . + `k = a then `21 + . . . + `2k ≤ (`1 + . . . + `k )2 = a2 ,
so that the sum of the squares of the projections on the x axis is ≤ 2a2 ; similarly, the sum of the squares of
the projections on the y axis is ≤ 2a2 .
Your problem 6 (Putnam1966-B4): Let 0 < a1 < a2 < · · · < am n+1 be m n + 1 integers. Prove that you can
select either m + 1 of them no one of which divides any other, or n + 1 of them dividing the following one.
I had not written a solution before: For each i, let f (i) be the length of the longest chain starting at ai and
such that each term divides the following one. If for some i one has f (i) ≥ n + 1 the second part of the
assertion is true. If for all i one has f (i) ≤ n, then by the pigeon-hole principle there is a value k ∈ {1, . . . , n}
such that {i | f (i) = k} has at least m + 1 elements; if i 6= j and f (i) = f (j) = k, then ai does not divide
aj , so that the first part of the assertion is true: indeed, if ai divided aj , the chain starting with ai and
continuing with a longest chain starting at aj would imply f (i) ≥ k + 1.
If p1 , . . . , pm are distinct odd primes, and one considers the integers 2i pj for i = 1, . . . , n and j = 1, . . . , m,
then one obtains m n distinct integers; one integer in this list divides another integer in this list if and only
if they have the same value of j, so that the length of a chain such that each term divides the following one
is ≤ n; in any subset of m + 1 elements in this list, two of them must have the same value of j, so that one
divides the other.