An electron leaves a molecule like a tenant leaving the top floor: to get out, you have to climb the energy steps. But in the quantum staircase, the steps are not empty: other electrons crowd on them, and the crowding changes the height of the climb. The standard ADC(3) method does not notice this crowding; it assumes each electron climbs an empty staircase. As a result, the detachment energy it predicts is systematically too high.
In the new work, the staircase is fixed. Instead of bare steps, which are differences of orbital energies, they substitute 'dressed' Epstein–Nesbet denominators. Each step gets a correction for the fact that someone is already standing on it: two electrons in the same orbital repel each other, making the climb more expensive. This correction is not a one-time action but an infinite series of 'ladder' interactions that the method accounts for automatically, without new integrals.
Tests on 23 molecules showed that the average error drops from 0.326 eV to 0.09 eV, almost a factor of four. This is better than the more expensive EOM-CCSD method (0.174 eV) and close to the benchmark EOM-CCSDT (0.04 eV). Moreover, the computational cost does not increase: all corrections are taken from quantities already computed.
But the staircase is not perfect. For open shells, where electrons are not paired, the improvement is more modest: from 0.16 to 0.07 eV for weakly spin-contaminated systems, and for strongly contaminated ones only from 0.17 to 0.15 eV. And for long chains of acenes and oligothiophenes, the method changes the sign of the error: it improves small molecules but worsens large ones.
Let us return to the tenant on the top floor. Previously, he paid only for the climb itself, and the calculation overestimated the cost of leaving. Now the staircase accounts for the fact that someone is already standing on the steps, and the bill becomes more accurate. For closed shells, the method is ready for routine use: photoelectron spectra, redox potentials, charge transfer. For open shells and long chains, the staircase still needs to be extended.
The Epstein–Nesbet shift can be interpreted as adding the repulsion energy of two electrons that end up in the same orbital. For a pair of virtual orbitals, this shift is always positive, making the denominator more negative and preventing divergences when orbital energies are degenerate. In calculations for the BN molecule, the second ionization potential changed by 0.15 eV solely due to the inclusion or exclusion of one shift term, illustrating how sensitive the system is.