today I was investigating least common multiples up to some integer $n$. I made a simple program to create a graph of $\frac{LCM(1, 2,..., n)}{n}$ and noticed a resemblence to a $\frac{1}{n}$ graph, though it's quite chaotic.
I then changed the function to $\sum_{1}^n\left(\frac{LCM(1, 2,..., n)}{n}\right)$, which came out much smoother as it is cumulative, however the scale is way off (in the billions), and I couldn't find a base that produced a linear relationship, so I was unsure if it was logarithmic.
Something I found quite interesting was when plotting $\frac{LCM(1, 2,..., n)}{LCM(1, 2,..., (n-1))}$, all of the peaks (integer values in the line y = x) were either primes or very close to primes (usually ±1, the max I found was a difference of 3).
My question is why these relationships emerge - are they linked to the primes (like the PNT)?
Thanks!
EDIT: Graph of $\frac{log\left(LCM(1, 2,..., n)\right)}{n}$ by request:




There is something strange about your graphs. Namely, $LCM(1, \dotsc, n)$ is no smaller than the product of all primes between $2$ and $n,$ which, the prime number theorem tells us, behaves like $e^n.$ Now, because of higher powers of primes, the LCM behaves roughly like $e^n e^{n^{(1/2)}} e^{n^{(1/3)}} \dots \ll \exp((2+ \epsilon) n),$ for any $\epsilon.$ So, what ARE the graphs of? If, as suggested by Aaron you are dividing by $n!,$ that make slightly more sense, but then the decay should be much faster than $1/n$