Euler, Grinberg,... who's next?
Solution 1:
Let's restrict it further to bipartite, cubic, hamiltonian Graphs $G$ only made up of $4$- or $6$-faces. These graphs can build out of $6$ faces of degree $4$ and the rest of degree $6$. Let $F$ be the number of faces, $V$ the number of Vertices, $E$ the number of edges and $a_k$ be the number of face of degree $k$ inside and $b_k$ outside the Hamilton cycle. Starting from Euler, we get: $$ F+V=E+2-2g\\ F=\sum_{k\in \{4,6\}} a_k+b_k = E-V+2 $$ Here $g=0$ for graph embedded in the plane. For $3$-regular graphs, we have $2E=3V$ and $a_4+b_4=6$ so $$ 6+a_6+b_6= V\left(\frac{3}2-1\right)+2 $$ $$ 2(a_6+b_6)= V-8\tag{1} $$ By combining $(1)$ and $$a_4+b_4=6\tag{2}$$ with Grinberg's formula (divided by $2$) in the given case $$ (a_4-b_4)+2(a_6-b_6)=0\tag{3} $$ we arrive at
- $$ a_4+2a_6=\frac12V-1 \tag{$\frac12\left[(1)+(2)+(3)\right]$} $$
I checked some examples and it seems to work. Tell me if there is anything wrong or unclear.
Depending on $\frac V2$ being even or odd there are $1,3,5$ or $(0,)2,4,6$ potential $4$-faces inside the HC, where I'm not sure whether $0$ or $6$ is possible at all...
UPDATE:
My interest changed to graphs on double tori $g=2$, build up from faces with $6$ and $8$ edges. There are (again!) six octagons necessary, to build these. I focus on hamiltonian graphs that cut the double torus into two tori first, as mentioned here.
For $3$-regular graphs, we have $2E=3V$ and $a_8+b_8=6$ so $$ 6+a_6+b_6= V\left(\frac{3}2-1\right)-2\\ 2(a_6+b_6)= V-16\tag{1'} $$ By combining $(1')$ and $$3(a_8+b_8)=18\tag{2'}$$ with Grinberg's formula (divided by $2$) in the given case $$ 2(a_6-b_6)+3(a_8-b_8)=0\tag{3} $$ we arrive at
- $$ 2a_6+3a_8=V+2 $$