The manuscript “Non-trivial charge-to-spin conversion in ferromagnetic metal/Cu/Al2O3 by orbital transport” was finally accepted from Physical Review B Rapid Communications (now it’s called Letter). It was not really easy to get this manuscript published, which took more than 2 years.

It was beginning of 2018 when I heard that SOT in AlOx/Cu/FM is surprisingly large. My first impression was about half-and-half. Around 2018, I didn’t know whether the orbital current may generate torque on the magnetization, but I knew that it is possible “in principle” although the magnitude may be tiny (My orbital torque paper‘s preprint came out about a year after). In this regard, I thought that the experiment might have measured something related to orbital current. In fact, it is the reason why I started to investigate “orbital torque” more seriously. On the other hand, I knew that orbital current in Cu is strongly suppressed simply because valence electronic state in Cu consists of s-orbital and d-orbitals are all occupied, which means there’s no angular momentum in the valence state. Because of this, to be frank, I was 50% skeptical on the experimental data: Maybe it’s related to current shunting effect in Cu.

But as time goes on, it became clear that the measurement was correct, and it was not likely that the signal is a measurement artifact. Dr. Junyeon Kim, who is the leading author of the paper in Otani-group in RIKEN-CEMS, checked in various different methods: (1) ST-FMR, (2) spin pumping, (3) Edelstein MR, (4) Current-induced magnetization switching. More importantly, he found a very unusual dependence of the SOT efficiency on the choice of the ferromagnets: SOT efficiency was more than 100 times larger in AlOx/Cu/CoFe samples than the efficiency measured in AlOx/Cu/Ni samples. We already knew that a concept of “spin transparency” can explain variation of the SOT efficneicy for different interfaces, but for heavy metal/3d ferromagnet interfaces deviation is by maximum 3-5 times. We noticed that the factor 100 difference that we observed implis that the mechanism is something beyond a conventional spin injection/torque mechanism.

At the end, this “interface dependence” became a crucial evidnece from which we concluded that it is very likely to be orbital-injection torque. Indeed, TEM images revealed that Cu/CoFe interface is very clean while Cu/Ni interface is “intermixed”. The idea is the following. For the spin injection, spin interacts with the interface disorder “indirectly” via spin-orbit interaction. Thus, if it is the spin that is carrying angular momentum in AlOx/Cu/FM structure, we shouldn’t observe huge variation of the torque efficiency. On the other hand, the orbital degree interacts with the lattice directly, regardless of spin-orbit interaction. It means, injection efficiency of the orbital current will depend significantly on the interface crystallinity (This is discussed in my orbital toqrue paper, but to be frank, there’s not yet a direct numerical result that supports this hypothesis). For this reason, we concluded that the torque that we observed in AlOx/Cu/FM is medaited by “orbital current”, instead of spin current.

Then why did it take so long before the publication? Well, we got REJECTED SO MANY TIMES from different journals. While reviewers pointed out different aspects overall, the common reason for rejections was “there’s no DIRECT evidence that angular momentum carrier is orbital degree of freedom”. I understand that one has to be skeptical, especially someone claims a whole new concept in the field. But even for spin current, which is regarded as an established concept, is there any direct evidence of spin current even in STT or SOT?

But overall, our experimental evidence was far from complete to support the orbital torque mechanism. There are many speculations and open questions. In my opinion, it is inevitable for any kind of measurement of a new effect. Usually, people go though this step by resporting to theoretical predictions. But for us, it was not easy because investigaton of interface crystallinity and disorder is still very difficult to estimate from theoretical/numerical calculations (I mean, not only to me, but to everyone in this field). Well, I am planning to attack this problem later, anyway.

An interesthing thing is that different groups also observed that oxidized Cu is somewhat strange (in the sense of usual SOT mechanisms) but easily explained by the orbital torque mechanism. This includes a preprint from Kazuya Ando’s group and a recent PRL from Kläui Lab (in which I also participated). If we combine all these results, then it’s hard to deny that orbital transport is involved. But in the individual papers alone, they are not quite complete. Meanwhile, I also recently published a preprint reporting that a surface oxidized Cu film exhibits exceptionally large orbital Rashba effect.

Anyway, I am quite glad that finally our paper is accepted despite all the harsh reviews. I consider it as the first step for establishing a bigger picture of the orbitronics. Obviously, we will need to figure out more details from both experiement and theory.