A clear trend is emerging in theoretical reports of dilute magnetic semiconductors: blind enumeration.
do i = all_periodic_table
do j = all_semiconductors
dope $i in $j
run dft_code
if (magnetic_moment > 0)
echo eureka! time to write a paper
else if (magnetic_moment = 0)
echo maybe it's time to co-dope.
enddo
enddo
Standard practice for the brute force approach appears to be: (i) skim previous literature, (ii) forget about intrinsic defects, (iii) ignore any methods post-LDA/GGA, (iv) don't think too hard about the results. This week’s examples are ZnS, and just one more on top of a bottomless series of nitrogen in oxide investigations, this time on TiO2.
Unsurprisingly, aliovalent substitutions where the difference in formal oxidation states is +/- 1 will produce an unpaired electron, e.g. N on an O site or Li on a Zn site. In some cases this can be compensated by intrinsic defects, e.g. in ZrO2, Y(III) doping on the Zr(IV) site is compensated by the formation of oxygen vacancies to be overall charge neutral and closed shell: 2Y_Zr + V_O (YSZ). In a material like MgO, you can form localized oxygen hole states, which act as catalytic centres, but the room temperature stabilization of long range ferromagnetism is highly unlikely. It is possible that (magnetic) aether is staging a comeback, or the experiment on d0 ferromagnetism has gone as bananas as ferroelectrics.
(Update August 2010) An interesting review on the caveats of the theory of dilute magnetic semiconductors appeared in the new APS magazine Physics by Zunger et al. - not entirely objective, but some good points are made.