Prove that $2^{(log(n)^{1/2})}$ is $O(n^a)$

Hopefully this is the right section. I need to prove that $2^{(log(n)^{1/2})}$ is $O(n^a)$. From the basic principle of Big-O notation, I know I need to find two numbers $c$ and $N$ so that $f(n) \le c\cdot g(n)$ for all values of $n \ge N$. In this case, $f(n)$ is $2^{(log(n)^{1/2})}$, and $g(n)$ is $n^a$.

The first problem in the set had me prove that $2^{n+a}$ is $O(2^n)$. For this problem, I had separated the $f(n)$ to $2^n \cdot 2^a$. I was guided by another user to set $c = 2^a$. This way, with $g(n) = 2^n$, $cg(n) = 2^a2^n$, which is obviously $\ge f(n)$, $2^{n+a}$, for all values of $n \ge 0$. This problem, however, I don't find as easy.

Could someone point me in the right direction? I've been trying to rearrange the problem and solve it for a while now.

2 Answers

$n^a = \exp(a \ln n) = \exp(a \ln n \frac{\ln 2}{\ln 2}) = 2^{\frac{a \ln n}{\ln 2}}$. Can you prove that if $h(n) = O(k(n))$ then $2^{h(n)} = O(2^{k(n)})$ ?

It holds that $\log x^y=y\log x$. Hence we have $$2^{\log n^{1/2}}=2^{1/2 \log n}=(2^{\log n})^{1/2}=n^{1/2}=O(n^{1/2}).$$

So this is mostly about computing with logarithms, and not so much about the big-O notation.

• I am not quite sure, that by $log(n)^{1/2}$ he meant $log(n^{1/2})$. Maybe he actually meant $log^{1/2}(n)$. I think OP should clarify.
– Nejc
Nov 17, 2012 at 14:35