It is argued that spin alignment can and will occur for ion-core spins in superconductors, but that the alignment is in the form of extremely small domains. Central to the argument is the concept of the nonlocal susceptibility $\chi (r-r^{\prime })$, which leads to a positive short-range Kittel-Ruderman-Yosida interaction of ion-core spins, but a negative long-range interaction of range ${\xi }_0$.

Very general arguments suggest that purely ferromagnetic alignment should not be observed in preference to this domain-like "cryptoferromagnetic" alignment.

• Received 25 June 1959

DOI:https://doi.org/10.1103/PhysRev.116.898