Solution 1
Let one optimal set of integers be {a1,…,a2k+1} with a1>a2>⋯>a2k+1>0 .
The two conditions can now be rewritten as a1+⋯+ak≤N/2 and a1+⋯+a2k+1>N .
Subtracting, we get that ak+1+⋯+a2k+1>N/2 , and hence ak+1+⋯+a2k+1>a1+⋯+ak .
In words, the sum of the k+1 smallest numbers must exceed the sum of the k largest ones.
Let ak+1=C . As all the numbers are distinct integers, we must have ∀i∈{1,…,k}: ak+1−i≥C+i , and also ∀i∈{1,…,k}: ak+1+i≤C−i .
Thus we get that a1+⋯+ak≥kC+2k(k+1) , and ak+1+⋯+a2k+1≤(k+1)C−2k(k+1) .
As we want the second sum to be larger, clearly we must have (k+1)C−2k(k+1)>kC+2k(k+1) .
This simplifies to C>k(k+1) .
Hence we get that:
N≥2(a1+⋯+ak)≥2(kC+2k(k+1))=2kC+k(k+1)≥2k(k2+k+1)+k(k+1)=2k3+3k2+3k
On the other hand, for the set {k2+k+1+i ∣ i∈{−k,…,k}} the sum of the largest k elements is exactly k3+k2+k+2k(k+1) , and the sum of the entire set is (k2+k+1)(2k+1)=2k3+3k2+3k+1 , which is more than twice the sum of the largest set.
Hence the smallest possible N is N=2k3+3k2+3k .
Alternate solutions are always welcome. If you have a different, elegant solution to this problem, please add it to this page.