THF versus Furan

This is probably a stupid question, but I am wondering about it, nonetheless. Why is the oxygen on THF sp³ and the oxygen on furan sp²?

When looking at the aromaticity of furan, people tend to say, well, the oxygen on furan is sp² hybridized and therefore one of its lone pairs is in a p orbital. Thus we have a p orbital on every atom in the ring, and the ring is planar. There are 6 pi electrons in the ring, so by Hückel's rule it follows that furan is aromatic.

But, but, but...both oxygens have two lone pairs on them and two adjacent carbons, so naively one would predict that both oxygens are sp³. Both oxygens are "locally planar" in the sense that the C-O-C area is definitely flat if we ignore the hydrogens (3 points define a plane).

I ran some MO calculations [1] (of unknown accuracy) and the frontier orbitals on THF looks like this (oxygen at bottom, slightly left):
The HOMO is -12.7 eV, the LUMO is +21.5 eV.

The frontier orbitals on furan looks like this:
The HOMO is -13.4 eV, the LUMO is -0.100 eV.

Sorry about the random orientations. In both cases the frontier orbitals are more-or-less unique (no degenerate MOs), although the energy of the orbital directly under the HOMO of furan is pretty close: -13.9 eV.

I can also look at dihydrofuran (DHF); it's almost planar (oxygen on the left).
The HOMO is -11.6 eV, the LUMO is 1.99 eV.

Atomic orbital energies are as follows:
For oxygen: 2s is -32.4 eV, 2p is -15.9 eV
For carbon: 2s is -19.5 eV, 2p is -11.3 eV
For hydrogen: 1s is -13.6 eV. It shouldn't matter, though.

Ok, now I have some pictures and numbers. Do they answer my question?

In MO theory, "lone pairs" don't really exist; they basically correspond to electrons in nonbonding (or "pseudo-nonbonding") molecular orbitals and have their electron densities all over the molecule. Certainly, lone pairs can't really be measured by spectroscopic techniques as far as I know. For the three molecules, the HOMOs are bonding with respect to some AOs and antibonding with respect to others (no surprises there), so I suppose one can argue that they are "psuedo-nonbonding" (eh, kinda, sorta). Going by looks, it seems like there's significant electron density "isolated" on the oxygen in the HOMO of THF and DHF, but none at all for the HOMO of furan, which suggests that the "lone pair" on oxygen does delocalize into the pi system (the other lone pair is in another MO) of furan. However, there's no "halfway" behavior with DHF.

But why?

I guess a possible explanation would involve ignoring everything about hybridization and just looking at the energies of furan MOs. If we were to promote the electrons of the HOMO to the LUMO, we would see some (albeit small) "lone-pair-like" electron density on the oxygen. But that LUMO is -0.100 eV. That's a huge energy difference from the actual energy of the HOMO, -13.4 eV.

In valence bond theory terms: if we were to insist that the furan oxygen is sp³ hybridized, the molecule is much less (energetically speaking) stable than if it was sp². The ability for a lone pair on the oxygen of furan to delocalize into the ring is a strongly stabilizing effect [2], and anyway, there's no lone-pair/lone-pair repulsion to speak of when one of the lone pairs is in the pi system of the ring. On the other hand, there would be lone-pair/lone-pair repulsion if the furan oxygen is sp³ hybridized, and we lose the stabilizing effect of delocalization. Thus, the oxygen on furan is sp² hybridized and not sp³. In contrast, THF and DHF are not sp² hybridized because there is no such thing as "delocalization" for them - the lone pairs are stuck where they are and so the lowest energy hybridization on the oxygen is sp³. For DHF, there does exist a resonance form where the oxygen adopts sp² hybridization, but the result is a carbanion, and a not particularly stable one at that. So, the lone pair is still more-or-less stuck.

I guess the problem comes about when people say that furan is aromatic *because* the oxygen is sp² hybridized. Any insights on this would be welcome, though!

Meanwhile, I better actually study the "reactions and mechanisms" part of the course instead of moping about aromaticity, what with the final being tomorrow. Especially the "memorize this list of reaction conditions" part (whereas mechanisms are usually reasonable things). (Aw, but writing blog posts is more fun! *Sigh*.)

Oh well. *Cracks whip over head.*

[1] On a click-and-run program - it's such a pleasant surprise when a professor gives out the licence codes for such software for students to use. Admittedly, the licence is university-wide, but that doesn't mean that a student can easily access such things.
[2] Is it obvious that delocalization is stabilizing, in valence bond theory?