Every textbook draws water with two identical "rabbit ear" lone pairs. Shine light on water and knock those electrons out, and they leave at two different energies: 12.6 eV and 14.8 eV. Here is the real computed electron cloud, the textbook picture, and how one turns into the other. Drag to rotate.
Two identical lobes, like rabbit ears. They are real: rotate the four valence orbitals until each is as compact as possible (Boys localisation) and two of them come out as mirror-image twins, pointing up and out of the plane.
Mix the two ears 50:50 and you get two new orbitals: one is exactly the out-of-plane p orbital 1b1 (overlap 0.9997 on our grid), the other an in-plane lone pair. The total cloud never changes while you rotate. Both pictures describe the same molecule.
Photoelectron spectroscopy measures how much energy it takes to remove each electron. Water's two highest peaks sit at 12.6 eV (1b1) and 14.8 eV (3a1). Our EOM-IP-CCSD calculation gives 12.59 and 14.78 eV. Two different energies: the lone pairs are not twins.
The rabbit ears are not "wrong". Any rotation of the filled orbitals gives the same total cloud and the same energy; ears suit bonding pictures, the split pair matches spectra. 3a1 also mixes a little with the O–H bonds (overlap 0.81 with the pure in-plane pair). One molecule in vacuum, no liquid water.
Our geometry (B3LYP/aug-cc-pVTZ): H–O–H 105.1°, O–H 0.962 Å, dipole 1.85 D. Measured in the gas phase: 104.52°, 0.9572 Å, 1.855 D.
PySCF · aug-cc-pVTZ basis
geometry: B3LYP, Berny optimiser
orbitals + density: Hartree–Fock
ionisation: EOM-IP-CCSD
ears: Boys localisation
grid 72³ · ray-marched in the browser