Here we consider a bent triatomic molecule in \(C_{s}\) symmetry. We have only one fermionic IRREP and complex MOs. This is indicated in the ADC output reporting complex ADC matrices. We want to do single ionizations (DOSIPS=T) at the ADC(3) level (ADCLEVEL=3) including constant diagrams (DOCONST=default) and compute all final states in symmetry 1, the only available one (XREPS=1). If the basis set is large, we restrict the iterations for the constant diagrams by releasing convergence a bit (VCONV=1.0E-05) because this step takes longest. We request 1000 Lanczos iterations and consider ionization energies up to 100 eV on screen. The complete spectrum is written to SSPEC.01. You get as many entries as you have Lanczos iterations. A characteristic of the iterative diagonalizer is that the eigenvalues at the edge of the spectrum converge very fast and are reproduced for high iteration numbers. In order to get higher eigenvalues equally tightly converged the number of iterations has to be increased accordingly. In this case the spurious (already converged) eigenvalues are projected out. For the \(C_{s}\) molecule the ADC input would then look like this:

**RELADC
.DOSIPS
.SIPREPS
 1 # How many requiered IRREPS are going to follow in the next line
 1
.VCONV
 1.0E-05
**LANCZOS
.SIPITER
 1000
.SIPPRNT
 100.0

This part has to be included into an input file, which contains the information about the Hamiltonian, the appropriate number of electrons, … .

Since we do not activate DIPs nothing else needs to be specified in the Lanczos section.