Solution:
It suffices to find a non-empty, finite set σ∈N such that
n∣s∈σ∑ds
for all d≥2. Consider the set σ={φ(n),2φ(n),…,nφ(n)}. For d≥2, if (d,n)=1:
s∈σ∑ds=k=1∑n(dφ(n))k=n
Else, if g=(d,n)=1, let n=xy with (x,y)=(y,d)=1 and y maximal:
s∈σ∑ds=k=1∑n(dφ(n))k≡dφ(n)−1dφ(n)(dnφ(n)−1)≡0modx
which makes sense, as dφ(n)−1 is coprime to x, and true since vp(x)≤vp(n)≤φ(n) (ask raphi for an elaborate explanation). And y divides this sum by the first argument as above. Now n=xy divides the sum since (x,y)=1.