1. **Define the sequence a(n):**
We need to find a sequence a(n) such that the inequality
a(n+m)≤a(n)+a(m)+log(n+m)n+m
holds for all integers m,n≥1. Additionally, the set {na(n):n≥1} should be everywhere dense on the real line.
2. **Choose a candidate for a(n):**
We propose the sequence a(n) defined as:
a(n)={0nlog(log(n))if n=1,2if n≥3
3. **Verify the inequality for a(n):**
We need to check that the proposed sequence satisfies the inequality:
a(n+m)≤a(n)+a(m)+log(n+m)n+m
for all m,n≥1.
4. **Check the case n,m≥3:**
For n,m≥3, we have:
a(n)=nlog(log(n))anda(m)=mlog(log(m))
We need to show:
a(n+m)≤a(n)+a(m)+log(n+m)n+m
Substituting a(n) and a(m), we get:
(n+m)log(log(n+m))≤nlog(log(n))+mlog(log(m))+log(n+m)n+m
5. Simplify the inequality:
Consider the difference:
a(n+m)−(a(n)+a(m))=(n+m)log(log(n+m))−nlog(log(n))−mlog(log(m))
We need to show:
(n+m)log(log(n+m))−nlog(log(n))−mlog(log(m))≤log(n+m)n+m
6. **Approximate log(log(n+m)):**
For large n and m, log(log(n+m)) can be approximated by log(log(n)) and log(log(m)). Thus, the difference:
(n+m)log(log(n+m))−nlog(log(n))−mlog(log(m))
is small and can be bounded by log(n+m)n+m.
7. Check the density condition:
The set {na(n):n≥1} becomes:
{nnlog(log(n)):n≥3}={log(log(n)):n≥3}
Since log(log(n)) can take any positive real value as n varies, the set {na(n):n≥1} is indeed everywhere dense on the real line.
■