I would like to perform a DFT+U calculation on transition metal oxides using CASTEP. To learn about the method, I tried some simple calculations that are published in the literature about doped TiO2. I have set up the model, giving U a value for the transition metals in the cell, and calculated CATEP with LDA+U checked. The result I got was a closed shell result for a couple of U value choices, while I was expecting some unpaired spin appear (as predicted by the papers I try to reproduce). Is that a problem with the choice of U? Does CASTEP automatically convert the result closed shell (showing only one DOS during analysis) when up and down spin space has the same DOS?