Some quick updates on this plan, mostly in picture form.
First of all, some quick concept diagrams of the overall mission architecture:
Nothing here is necessarily to scale, the use of two stages, one for TLI and one for LOI/TEI is not necessarily an inherent part of the "new two module idea", and the colours are only to make it easy to tell things apart. The small blue headlamp shaped module is the descent module, the large green cylinder is the orbital module. I made the orbital module cylindrical as per Soyuz rather than roughly spherical like Soyuz purely because I think it looks a little nicer: if there ends up being a good reason to prefer one approach over the other I won't hesitate to make the best choice.
Now, onto some more realistic things:
As part of the big mass analysis I've been working on (which has been temporarily shelved as I have a tonne of uni work to do), one propulsion architecture I am considering involves a LH2/LOX stage purely for TLI and an RP1/LOX stage that handles LOI and TEI, i.e. this is a two-stage architecture like that shown above. Using RP1 (kerosene) instead of LH2 isn't something we've considered much before, but I'm starting to prefer it because I think it's not anywhere near as worse than LH2/LOX as one would think. I think we have understimated how inconvenient using LH2, which is deeply cryogenic, for burns which occur days after the initial launch will be (we have never included in our mass analyses an allowance for LH2 boiloff, which is pretty much inevitable, and when considering the impact of RP1's lower Isp we have neglected the counterbalancing fact that RP1 tanks are much, much lighter than LH2 tanks, due to RP1's much higher density and the need for less insulation. A "bad" LH2 tank can have a mass of around 15-20% of the propellant it holds, whereas really good RP1 tanks can have a mass of just 1% of their capacity. That's a big difference.
Long story short: I currently believe that if we use LH2/LOX for TLI and RP1/LOX for everything else (including the lander), then with a lander structural mass of 200 kg (this is not including engine, propellant or propellant tanks – it is frame, seat, and avionics) we can have a total DM-OM mass of 1132 kg.
Knowing all the masses of propellant and oxidiser required for this arrangement, I have spent some time thinking about how to physically arrange stuff, and making sure everything fits in the Falcon 9 payload fairing. To this end, I have prepared some scale diagrams:
The first diagram shows the situation where each lot of fuel and oxidiser gets its own spherical tank. This leaves very little room for the lunar lander, so I don't think it's workable.
The second diagram shows the situation where the TLI propellants have been collapsed into a single spherical tank with a dividing bulkhead. This frees up what looks like enough room, but there's little room for error, so I'm not too comfortable with it.
The third diagram shows the situation with all TLI propellants in a single spherical tank and the LOI/TEI propellants put in a cylindrical tank. The cylindrical tank is much more efficient from a space point of view, but also much less efficient from a mass point of view – however, as mentioned, RP1 tanks can be so light that making that sacrifice on this particular tank is not that big a deal. This arrangement has room for the LL and plenty of room for error. I also feel like this arrangement has the benefit of minimising the total number of spherical tanks: while these are the most mass efficient, they're also the hardest to manufacture.
A quick note on engine dimensions: originally, when I drew these diagrams, for the engines I used the measurements of the RL-10 engine (as used on Saturn and Centaur) for the LH2/LOX stage, and SpaceX's Kestrel engine for the RP1/LOX stage. However, they just looked comically huge. The truth of the matter is that the propellant and payload masses we are considering here are so far below what is normal in astronautics that standard engines are generally overkill. So for now in these diagrams I have just drawn engines which "look about right". This is obviously very imprecise which is why I am concerned about leaving room for error.
All feedback super welcome.