Denoting y1=x1, yk+1=xk+1−xk, for k∈{1,2,…,n−1}, we get xk=y1+y2+⋯+yk, cu y1>0,yk≥0,∀k=2,n. With these notations, the inequality becomes (a1+a2+⋯+an−c(b1+b2+⋯+bn))y1+(a2+a3+⋯+an−c(b2+b3+⋯+bn))y2+⋯+(an−cbn)yn≥0 (∗).
Plugging y1=1,y2=⋯=yn=0, shows that necessarily c≤b1+b2+⋯+bna1+a2+⋯+an.
We show that, conversely, for every c≤b1+b2+⋯+bna1+a2+⋯+an the inequality under consideration is true, so b1+b2+⋯+bna1+a2+⋯+an is the required maximum.
We notice that c≤b1+b2+⋯+bna1+a2+⋯+an≤b2+b3+⋯+bna2+a3+⋯+an≤⋯≤bnan. Indeed, the inequality bk+bk+1+⋯+bnak+ak+1+⋯+an≤bk+1+⋯+bnak+1+⋯+an can be written ak(bk+1+⋯+bn)≤bk(ak+1+⋯+an) and is justified by the inequalities bkak≤bk+1ak+1,bkak≤bk+2ak+2,…,bkak≤bnan.
This shows that every term of the left member of (∗) is at least 0, so (∗) is true.