I tried to prove the following theorem :
If $n>1$ then $\displaystyle1+\frac{1}{2}+\dots+\frac{1}{n} = \frac{k}{m}$ where $k$ is odd and $m$ is even.
and I'd like to know if there is any flaw in my argument.
Proof :
For $n = 2$ and $n=3%$ we verify the theorem directly.
When $n \geq 4$ we can argue as follows :
First of all we sum up all the fractions obtaining
$$1+\frac{1}{2}+\dots+\frac{1}{n}= \frac{n!+1\cdot3\cdots n+\dots+(n-1)!}{n!} \:\:\:\:\:\:\:(1)$$
Now I define the multi-parity function $P$ as a function which counts the number of twos in the factorization of an integer, for example $44 = 2^2 \cdot 11$ so $P(44) = 2$.
Every product $c$ in the numerator of $(1)$ has $P(c) \geq P(n!) - \max[P(x)]$ where $1\leq x \leq n$ . This is true because every product in the numerator has just one factor less than $n!$ , for example we know that $P(1\cdot 3 \cdot ... \cdot n)=P(n!)-1$.
Moreover we know that $P(c) = P(n!) - \max[P(x)]$ in only one case that is when $c$ hasn't the greatest power of two less than $n$ in the product.
So using $g = \max[P(x)]$, which is also always equal or greater than $1$, and $q = P(n!) - g$ we can gather $2^q$ in the numerator and $2^{q+g}$ in the denominator : $$\frac{n!+1\cdot3 \cdots n+\dots+(n-1)!}{n!} = \frac{2^q(a_1+\dots+a_n)}{2^{q+g}m} = \frac{a_1+\dots+a_n}{2^{g}m}$$
but the denominator is even because $g \geq 1$ and the numerator is odd since all the $a_i$ are even except for one.
Is my argument sound ?
Anyway I found the problem in the great book Introduction to algorithms by Udi Manber and I found there are lots of threads about this problem on MSE and my proof is similar to the elementary proof of the statement without Bertrand Postulate but I enjoyed proving it by myself :-)