I did my PhD in a related area of magnetics to this. One of the problems is finding appropriate materials, and that applies for all magnetic devices. Crystal

I did my PhD in a related area of magnetics to this. One of the problems is finding appropriate materials, and that applies for all magnetic devices. Crystal structures are hugely complex, and the way that atoms interact manifests primarily through the 'exchange' interaction. This is caused by a combination of the Coulomb interaction (charges repelling/attracting), and through the quantum mechanical Pauli exclusion principle, which basically stops electrons with the same state from sitting in the same position. The problem is, while it's the dominant energy term, there are several forms of exchange interaction - RKKY, Dzyaloshinskii-Moriya (which affects the boundaries at interfaces and in certain crystals), etc. and so while the length scale over which the exchange interaction acts is normally on the order of a single atom, it can often actually act on next-nearest neighbour atoms and cause odd effects. This means that finding a 'pure' antiferromagnet is difficult; many things are actually ferrimagentic. Many studies which suggest wonderful new storage technologies are computational and have little bearing on the experimental reality. There are serious difficulties in predicting the magnetic properties of materials with density functional theory because the results are often oscillatory as you refine the model, meaning that the computational cost is enormous - the properties of even simple systems such as the two species alloy FePt can't be easily predicted. Real crystal structures are hugely more complex than this, and so it's questionable whether results given by people are any more than junk. The work in this area is hampered by the fact that DFT research is overwhelmingly focused on calculating bonding and chemical structures. DFT scales poorly, and while there are a few codes that scale better, they're licensed under weird license structures. There's also some odd fallings-out between various code authors with accusations of stealing... The 'next level' in the heirarchy of modelling neglects quantum effects completely and is known as atomistic modelling. These models are still enormously costly, but it's the only proper way to treat properly antiferromagnetism, as the next level in the heirarchy, micromagnetism, takes a continuum limit of this discrete model, and by doing so has assumptions that antiferromagnets break. Parameterising these models is enormously difficult, because it is pretty much impossible to measure experimentally the values that you need, and DFT gives answers which when used bear little resemblance to the real material. You have to sort of realise that many people in magnetics have lost funding because industry has moved on, and academic labs seem to be left behind. So everyone tries to work on the 'next big thing' - there are hundreds of various proposals for devices that haven't worked out, and it's worth being very, very sceptical of articles like this which promise that a new technology will revolutionise things. Just as an example, one of the biggest wastes of time over the last 10 years has been that of Skyrmions, which have been cited as the possible hallelujah of magnetic storage, and for which there is basically no evidence of utility after millions of dollars of research funding worldwide, and thousands of articles. Many scientists jumped onto this topic in my opinion because a Nobel prize winner wrote a brief, speculative computational paper suggesting that Skyrmions could be the panacea to the woes of current magnetic storage, despite there being even a lot of evidence at that point to the contrary - in particular, work by Stuart Parkin at IBM in trying to use magnetic domain walls for storage, and a whole host of work on magnetic bubble memory that even made it to market in the 70s and 80s but which was replaced because of poorer performance.