This is Chapter $3$, Problem $46$(b) from Stanley's Enumerative Combinatorics.
Let $f(n)$ be the number of sublattices of rank $n$ of the Boolean algebra $B_n$... Let $g(n)$ be the number of sublattices of $B_n$ that contain $\emptyset$ and $[n]$. Write \begin{align} F(x) &= \sum_{n \ge 0} f(n)\frac{x^n}{n!}\\ G(x) &= \sum_{n \ge 0} g(n) \frac{x^n}{n!}. \end{align} Show that $G(x)=F(e^x-1)$.
By the Compositional Formula, $F(e^x-1)$ counts the number of ways to take an unordered partition of $[n]$, then put a "$e^x-1$"-structure on each block, and then put a $F(x)$ structure (sublattice with rank $n$) on the set of blocks.
The exponential generating function $e^x-1$ has coefficients $1$ everywhere except for $[x^0]$. Hence it just eliminates situations where some blocks of our unordered partition were empty.
So we can interpret $[x^n]F(e^x-1)$ as counting the number of ways to take an unordered partition of $[n]$ into $k$ nonempty blocks, and then (corresponding the $k$ blocks with [k] as needed) find a rank-$k$ sublattice of $B_k$.
I searched this site for a similar question, and found Trevor Gunn's answer here. I can understand how Trevor Gunn takes some lattice $L\subseteq B_n$ and then determines the associated $k$. However, after trying some examples, I'm still struggling to see the correspondence that his construction creates between the set of sublattices of $B_n$ containing $\emptyset$ and $[n]$, and selections of a partition $S_1, ..., S_k$ of $[n]$ combined with a rank $k$ graded sublattice of $B_k$.
I was wondering if someone could help clarify this correspondence, or point me in the direction of a different proof?
Given a lattice $L\subseteq B_n$, define an equivalence relation $\sim$ on $[n]$ by $x\sim y$ iff for all $a\in L$, $x\in a$ iff $y\in a$. Say that $\sim$ has $k$ equivalence classes. We can then naturally identify $L$ with a lattice $L'\subseteq B_k$ if we identify $[k]$ with $[n]/{\sim}$ (replace each element of $L$ with the set of equivalence classes that it contains). I claim that if $L$ contains $\emptyset$ and $[n]$, then $L'$ has rank $k$.
Indeed, suppose $\emptyset=a_0\subseteq a_1\subseteq \dots\subseteq a_m=[k]$ is a maximal chain in $L'$ but $m<k$. Then for some $i$, $a_{i+1}\setminus a_i$ contains two distinct elements $x$ and $y$. Now there exists some element $b\in L'$ which contains exactly one of $x$ or $y$ (otherwise $\sim$ would have identified $x$ and $y$); say $x\in b$ and $y\not\in b$. Observe now that $a_i\cup (b\cap a_{i+1})\in L'$ contains $x$ but not $y$ and so is strictly between $a_i$ and $a_{i+1}$. This contradicts the assumption that our chain was maximal.
Conversely, if we start with a lattice $L'\subseteq B_k$ and an identification of $[k]$ with a partition of $[n]$, we obtain a lattice $L$ on $[n]$ by just replacing each element of $L'$ with the union of the corresponding subsets of $[n]$. If $L'$ had rank $k$, then $L$ must contain $\emptyset$ and $[n]$ and the equivalence relation $\sim$ induced by $L$ as above is exactly the equivalence relation of the partition of $[n]$ we used (since $L'$ distinguishes all elements of $[k]$). It is then easy to see that this construction is inverse to the one described in the first paragraph.