The Sona method, described in 1968 by Peter Sona, has been used in polarized sources of the Lamb-shift type and is still important at optically pumped ion sources, e.g. at BNL. The trick of this method is that an electron polarization of a hydrogen beam, e.g. produced by charge exchange of a proton beam with optically pumped rubidium atoms, can be transferred into nuclear polarization. For this purpose, the electron-polarized hydrogen atoms have to pass a zero-crossing of a longitudinal magnetic field that acts as quantization axis. This non-adiabatic passage exchanges the occupation numbers of the “pure” hyperfine substates $|1>$ and $|3>$, but keeps the “mixed” states $|2>$ and $|4>$. Thus, the atoms in a hydrogen beam in the states $|1>$ and $|2>$, both characterized by $m_J = +1/2$, will end up in the states $|2>$ and $|3>$ that have both $m_I = -1/2$.

Like other groups operating such a Sona unit for metastable hydrogen atoms, we observed strong oscillations of the occupation numbers of the involved hyperfine substates. These depend on several parameters like the magnetic field shape and amplitude of the Sona unit or the velocity of the hydrogen beam. In this proceeding we discussed the theoretical explanation of this effect and possible application for future polarized sources.