Author Archives: Dongwook

What I learned from advising other people

Since the last year, I started to advise a postdoc and PhD student who work on topics of my expertise. In our group, it’s quite often that there are one or two more experienced researchers involved in advising together with direct supervisors who are usually professors. I find this system quite nice becaues one needs to learn not only key ideas from physics perspective but also numerical techniques to study these subjects. About the latter, sometimes it requires to sit down together for several hours in the office and struggle together, but a more important thing would be to bring “physics concepts” into specific problems that can be solved by the tools we have. I believe this is what I learned during the PhD and postdoc and what I am particularly good at.

Recently, I started to think different ways of supervising. These are what I’ve been trying to do so far:

  • Explaining key ideas in a “broad” perspective
  • Setting up a long-term vision of the research together.
  • Iteratively giving feedbacks so that the project does not lose momentum and get distracted.
  • Teaching technical aspects (theoretical background and methods)
  • Recommending good materials for study (textbooks, lecture notes, thesis of others, review papers, etc.)
  • Asking continuously rather than giving my own opinions so that they find their own motivation and paths of the research.

Also, finally I decided to go to actual office rather than working at home more often as the I got vaccinated from corona virus. So, I can chat with my colleagues . What I immediately realize is that everyday’s short discussion is much better than a lengthy discussion that occurs once a week. Maybe human brain is more adapted to continuous communication for processing information rather than being logical all the time. Well, I experienced many times during my studies that I just started to understand things that I could not understand even though I haven’t really studied nor thought about it.

Anyway, coming back to the original topic… I feel quite rewarding nowadays to see my colleagues finding their own paths and methods as the time goes on.

Spin-orbit torques in low-symmetry materials

Last Friday, I joined a seminar by Cheng Song on OSSS. While most of the time I just watch recorded talks from the OSSS because the usual schedule is from 3 pm ET, which is 9 pm CET and a bit late for me (especially for Friday evenings). But this time, since the speaker was from Beijing, the talk started at 4 pm CET, almost at the end of the week for me. The seminar was mainly about using AFMs to generate “non-trivial” components of the SOT, whose directions are different from the usual spin Hall or Rashba torque directions. For example, Song and his colleagues demonstrated this in Mn2Au/Py. Such non-equilibrium spin accumulation is called “antiferromagnetic spin Hall effect” in the paper. We discussed a lot on the microscopic mechanisms after the talk, such as whether how the surface termination of Mn2Au at the Py interface affects the SOT or whether the mechanism is indeed from the bulk Mn2Au or from the interface.

Regardless, I think it’s a recurring theme in spin-orbitronics community that people seek for low-symmetry materials to generate non-trivial components of the SOT. Not only from fundamental aspect, but also technologically this is important to achieve field-free switching of the magnetization. Usually, switching of PMA magnets by the SOT requires an external magnetic field that breaks the degeneracy between +z and -z magnetization configurations. The community has tried many different ways in order to achieve the field-free switching. For example, in 2014, a UCLA group achieved a symmetry breaking in a latral direction by inhomogeneous oxygen contents in a wedge structure. Also, a KAIST group demonstrated the field-free switching in IrMn/CoFeB/MgO structures by using the exchange bias between the antiferromagnetic IrMn and ferromagnetic CoFeB.

Since then, many different clever ways are being proposed. I think there are two differnet directions overall: (1) Using nonmagnetic materials which breaks a in-plane mirror symmety by the crystal structure, (2) Using magnetic materials where the mirror symmetry is broken by the magnetization direction.

Typical examples of (1) are 2D materials. Many 2D materials have C3v symmetry. It means that, if the crystal is symmetric with respect to mirror operation x -> -x, it does not have a mirror symmetry for y -> -y. Thus, at the magnetic interface, it can generate additional component of the SOT as well as the conventional damping-like and field-like torques. This was experimentally demonstrated in WTe2/Py structure. A microscopic theory can be found in this paper. From a similar motivation, the field-free switching was also achieved in L1_1 ordered CuPt/CoPt interface. Also, a recent theory paper proposed a SOT in a single-layer FGT, where the SOT also originates from the C3v symmetry.

The symmetry can also be broken by magnetic layers. For instance, soon after the theoretical proposal of “interface-generated” spin currents an experiment was performed in magnetic trilayer structures. Meanwhile, the spin current induced by the magnetic order is often called “magnetic” spin Hall effect to distinghiush from the “usual” spin Hall effect that is even under the time-reversal. The magnetic spin Hall effect was also found in Mn3Sb, one of the non-collinear antiferromagnets in the Kagome lattice. In this sense, the antiferromagnetic spin Hall effect is also somewhat similar.

My impression on this direction of works is that there’s a consensus on the phenomenology and the symmetry requirements, far more investigations seem to be required for microscopic understanding of the mechanism, e.g. how is it correlated with the wave function in the bulk or interfacial states, etc. Anyway, I have no doubt that these are quite interesting pieces of physics and worth investigating further.

Differnet confusing names of spin-orbit torques

If you’re new to spin-orbitronics, you must be very confused for many different terminologies that seem to specify the same effect but from slightly different contexts. Honestely, this is really confusing and was terrible, especially when I had to study this field of research from the scratch in the beginning of my PhD. Just to name some of them, I easily find that these terms appear in many different papers: field-like torque, damping-like torque, antidamping-like torque, spin Hall torque, Rashba torque, Edelstein effect, even/odd torque, inverse spin Galvanic effect, intrinsic spin-orbit torque, interfacial torque, etc. This came out of discussion with my colleagues today and I came to think about it. In this article, I want to explain you a bit of motivations behind those terminologies and what they (usually) mean the same thing or not.

In thin film heterostructures such as Pt/Co, W/Py, etc. we usually find two independent components of the SOT. For instance, when an electric field is applied along the X direction, we have SOTs proportional to M x Y and (M x Y) x M. For instance, heterostructures with C4v symmetry ((001) stacks of bcc/fcc) and C6v symmetry (111 stacks of fcc(111) or hcp(0001)) have these two SOTs, as well as poly-crystalline samples (C-infinite-v symmetry). Although there exists higher-order terms that involve more M’s, M x Y and (M x Y) x M are the two leading terms. Phenomenologically, because M x Y torque is analogous to a situation with a magnetic field pointing along Y, M x Y torque is called field-like torque. By adding one more cross product with M, (M x Y) x M points toward the direction of the Gilbert damping, so it’s called damping-like torque. But since the direction of the torque can change if one flips the direction of the electric current, it can also cancels with the Gilbert daming. In this context, people call it “anti-damping”-like torque.

Around 2010, the whole community was debating over possible mechanisms for the SOT. While the SOT was “intuitively” understood as the spin Hall effect arising in the nonmagnetic substrate and the spin-transfer torque exerted by the spin current passing through the interface between the nonmagnet and ferromagnet, another mechanism based on the interfacial Rashba-type state was also proposed and many evidences that shows that the Rashba effect causes the SOT have been found. At the end, strictly speaking, both the SHE and Rashba mechanisms predict coexistence of the field-like and damping-like SOTs. In the SHE mechanism, the damping-like torque is direct consequence of the STT with the spin polarization direction Y, and the field-like torque is a secondary effect in the presence of the damping. On the other hand, the Rashba effect explains the field-like torque more naturally. There’s an effect called Rashba-Edelstein effect. The key idea for this effect is that shift of the Fermi surface by the current (Boltzmann-like) leads to finite spin density at the interface. On the other hand, although the damping-like torque emerges from the Rashba effect, it involves Berry phase physics and more complicated theoretical concepts. Though I’m not fully sure, but my suspicion is that many experimenatlists who were not familiar with the Berry phase physics didn’t buy the Rashba mechainsm for the damping-like torque because it doesn’t really come with most people’s “intuition”. Same thing might have happened for the field-like torque arising from the SHE mechanism. So, many people started to use spin Hall torque for the damping-like torque and Rashba torque for the field-like torque, if they want to emphasize microscopic mechanisms in their papers. Some people also called the Rashba torque by Edelstein effect. Because the Rashba effect is mostly present at the interface, it is also called interfacial torque.

Quite interesting fact is that although physical principles do not exclude either of the damping-like torque and field-like torque for both Rashba and SHE mechanisms, several DFT calculations showed that the damping-like torque is indeed dominated by the SHE while the field-like torque is mainly due to the Rashba effect. In the meantime, several theory papers were also publisehd by that time, showing that the Berry phase effect explains the damping-like torque from the Rashba effect. But these papers used specifically invese spin Galvanic effect, thus still people use invese spin Galvanic effect in order to specify the damping-like torque arising from the Rashba effect from the Berry phase mechanism. But because of the intrinsic mechanism involved in the Berry phase mechanism, it is also called intrinsic spin-orbit torque. By the way, since M x Y torque invovles only “one” M, which is odd with respect to the time-reversal, so it is called odd torque. Likewise, (M x Y) x M torque is called even torque.

In summary, (anti)damping-like torque, spin Hall torque, inverse spin Galvanic effect, even torque mostly mean the same term ~ (M x y) x M, and people use field-like torque, Rashba torque, Edelstein effect, Rashba torque, interfacial torque, odd torque to specify the term ~ M x Y.

Final remark: These terminologies were coined with poly-crystalline heterostructures in mind, or at least C6v or C4v symmetries. So, they don’t really make sense to apply to other low-symmetry materials. Recently, 2D materials such as FGT or TMD were found and the SOT are being actively discussed. Many 2D materials have C3v symmetry and the properties are very different from other crystals. For example, there’s a new SOT term which acts like an electric-field-induced magnetic anisotropy in a single-layer FGT. So phenomenologically, it acts like an effective anisotropy field, but the torque is even in M. Another example is a SOT in TMD/FM. Although many people discuss interface mechanisms here, which is believed to be the main contribution, there are cases that an interfacial contribution does not exactly correspond to the SOT arising from the Rashba effect in metallic thin films. I hope that the community uses vocabularies more carefully rather than extrapolating what we knew in metallic thin films and try to set up right terms to call new components of the torques that are being discovered in new materials.

Writing a paper in the early morning

About 2 weeks ago, starting from the end of the vacation, I decided to wake up ealier than usual. I found that I do not have enough time when I can concentrate on one topic without any interruption. Since the lockdown with COVID, messengers like Slack and Skype/Zoom discussion became daily events. While these technologies definitely help us to communicate better, one shortcoming is that I get easily distracted. I even have tried to use an airplane mode at some point when I needed to concentrate, but it didn’t work quite well. At the end, I felt a bit isolated and could not keep up with all the recent updates within our group.

That’s why I started to work in the early morning. Since I work at home most of the time, waking up at 5:30 or 6:00 is already sufficient. I just need 10 mins to prepare and go to my office which is the livingroom. So, I am pretty much on my own from 6 am till 9 am. I have a breakfast meal in between, and start to do “regular works” from 9 am. From this time, I am quite availalbe for discussion with my colleagues. Therefore, it turned out to be a very nice solution. The only problem is I have to sleep early in the night, and sometimes I just cannot especially I have a large meal for dinner and drink a couple of beers and it takes a few more hours for the digestion.

Anyway, in the morning hours, I’m trying to do works that requires most of the brain activity and creativity, such as research planning/proposal, writing a paper, and theoretical formulations and derivations. By now, it’s been almost 2 weeks, and I find the result very satisfying. I mostly spent time writing a paper, and I found myself highly productive than usual. Also, I finish work quite early, near 16:00, and I have time to enjoy walking in the park and exercising. I hope I can keep this habit!

A little break

Finally I decided to take a vacation for 2 weeks, which started from the last Monday. I have been working quite hard nonstop for several months since the beginning of the year, and I noticed that the productivity was becoming low and I got easily tired, which is a sign that I need a break.

But with the COVID situation, however, I cannot do anything special in the last years. I used to travel around with my partner. Staying in a hotel, eating in different restaurants, visiting torusitic places, … all seem to have happened a long time ago. I don’t know if I ever had a whole vacation at home so far in my life. I only remember that when I’m back from a trip, I had calm days at home 1-2 days before going back to work, checking emails and psychologically ready for the work. But now that I have whole 2 weeks at home, to be honest, I really don’t find what to do. What I’m doing is intentially a bit away from science, and I check my email only once a day in case there’s an urgent administration.

So far I have done various tiny things, mostly cooking nice foods, exercising a bit more than usual, watching series and documentaries, reading books, playing computer games alone or pplaying board games with my partner — leisure things that I also do at the end of working days. Though I was feeling a bit guilty for spending time like that, I intentially let myself to chill a bit, hoping that an inspiration comes naturally, something that is satisfying and broadening my perspective or developing my skills.

I’m slowly searching for simple physics problems that I can solve/demonstrate with computer codes. I think it doesn’t matter it’s studied or not. Even for well-known old problems, I always find them fascinating. A problem that is easy but not too easy but possibly giving cool plot/animation is what I’m looking for. Also, I’d like to explore Python a bit more. Although Python may not ben ideal for large-scale supercomputing, I believe that it can deal with most of the problems in physics. But so far, my Python skillset is quite limited, stricktly speaking.

For example, something like this would be cool.

Also, it’s quite nice to get back to basics – solving time-dependent Schrödinger equation. I expect that I can get this result roughly within 5-10 hours or 2-3 days.

What I learned from the online hands-on DFT workshop

During the last week, our group organized a hands-on workshop on the DFT code FLEUR. I participated the workshop as a tutor. I gave two talks and led a hands-on session on the Wannier function.

About the talk, one was about the basics of the Wannier function and its interface with the FLEUR code and the other was about the usage of the code for spin-orbitronics research. The first talk was provided as a recorded format since it is quite general and may be useful for users even after the end of the workshop, which can be found in this link. But regarding my own research I decided to give a live talk. To be honest, I was not quite comfortable with recording the talk about my own research because the recording makes it harder for me to be open to the audience. For example, I wanted to give my own perspective on the field, which some people might have different opinions. I think this kind of “raw” ideas are what inspires others the most, at least according to my own experience, even though the argument is not fully supported by solid evidences or proofs. Sure, I also wanted to present unpublished results from on-going projects.

Thus, I think for established techniques and knowledge, recording makes sense and can be helpful to many people. But about on-going research topics, it is likely that there are unsettled controveries in the field. Even I often my point of view changes over a few months. This means that something I say right now might not make much sense a few months later. I mean, I would diagree with my own opinion which I thought is correct in a few months before. Also, I think talking about only established or well-supported results makes all science activities quite boring. Imagine a conference without any controversy and everyone agrees. Honestly I wouldn’t enjoy this kind of science at all. I like to open myself a bit more when I give a talk. People might disagree, but I can learn from them if they give me reasons why my argument isn’t right. This gives room for me to learn and possibly to solve the problem by community-driven efforts rather than working alone.

Talking about the hands-on session, I found that it was more interactive than I have expected. I noticed that the participants can be roughly classified into two. The first group was quite active, and they ask so many questions, both regarding the workshop and their own research. They already knew what they want to learn, especially for experienced researchers. But PhD students who just started were asking questions in various aspects, which was very refreshing to be honest (and I liked them very much). I found that it’s quite nice to have a chance to hear different research topics and challenges to achive goals. The second group consists of kind of self-learners, who seem to be a bit shy. They didn’t ask questions and were very silient most of the time. During the session, I tried to talk to individual participants as much as possible but I found that several people were just silent. Some didn’t even answer at all. Probably they’re too shy to talk with camear and microphone. By the way, there were also a couple of funny characters. Some avatars were following me a lot whenever I was helping other participants or chatting with other tutors. It made us a bit uncomfortable because these people were listening to what we were speaking about but never showed themselves.

Let alone the workshop, I found also nice to dedicate a full week to think about the code itself. We discussed a lot about the future direction for developing the code. Especially I had two other colleagues who were helping me during the hands-on session. When there was no question, we chatted over various topics regarding the code. Sure we also discussed the best pizzaria in the city, which was very helpful for me since I haven’t tried many pizzarias yet since I moved.

Also, with the recorded talks and tutorial materials, I believe that we can make our webpage more helpful. On my side, I will give these materials whenever a new student or postdoc joins our group. So far, we transferred the knowledge in an old-fashioned way: One just sits down next to an experienced one in the same office and go through examples one-by-one.

Personally, I also studied quite a lot on the code and methods we’re using. To be 100% frank, although people from the outside may think that we’re gurus or Yoda on the method we’re using and developing, we don’t study basics that much. In my case, sure I have read key papers many times but it was long time ago when I first learned the technique. It’s when I had to teach, I studied the material again and realized that how ignorant I have been. I think this experience makes my knoweledge more mature, and importantly this knowledge becomes integrated with practices and experiences.

My lecture on the Wannier function is now available on YouTube

Our group, Peter Grünberg Institut in Forschungszentrum Jülich, is organizing a DFT hands-on workshop. We will cover both theoretical and practical aspects of using the FLEUR code developed by our group. Unfortunately, the registration is already closed, but all the lecture videos will be available on YouTube. I myself also plan to watch all the videos during the workshop week to learn and remind basic knowledge.

In this workshop, I am going to be a tutor on the Wannier function. The Wannier function is very useful for calculating electron transport properties in general. It’s because it often requires dense sampling of k-points or the system size is quite big for directly dealing with standard DFT methods. I am using this method for calculating various spintronic and orbitronic effects, which are my main research interests.

If you’re interested in this technique, you can watch my video. I hope this is going to be helpful. Enjoy!

A new paper is out: Orbital Rashba effect in a surface-oxidized Cu

Finally the paper on the orbital Rashba effect in a surface-oxidized Cu has been accepted in the last week. Now you can see the publisehd manuscript in the following link.

The major change during the reviews was calculation of various stacking geometries and comparisions (fcc, hcp, bridge, etc.). At the end, they are not very different: All the structures lead to pronounced orbital Rashba effect. But thanks to this calculation, we are now sure that the orbital Rashba effect would be robust as a Cu film is oxidized from the surface. Additionally, we added Supplementary Material, and all the calculation steps are made transparent.

Anyway, I really hope that the theoretical prediction of the orbital Rashba effect is going to be confirmed soon by angle-resolved photoemission spectroscopy or momentum microscopy.

First-principles approach

While a first-principles calculation is often regarded a synoym for the density functional theory in condensed matter physics, the idea of “first-principles” is more deeper than that. It means that we assume nothing but the most fundamental truth (or something we believe), on which any falsification seems very unlikely at least in the near future, and arguments start from this point. For example, the existence of atoms, quantum mechanics, statistical mechanics are examples of the first-principles. On the other hand, DFT calculations are based on quite many other assumptions as a matter of facts.

Anyway, I believe that the first-principles approach is the very core philosophy in physics. There are numorous natural phenomena that hardly have an analogy or intuitive picture, such as the Bell’s inequality, superconductivity, the big bang, etc. If physicists had used only analogy or intuition, discovery of such phenomena would have been impossible. It required to write down equations, beginning from fundamental assuptions. And not to jump any steps but to expand equations line by line.

However, at the same time, so called a hand-waving argument or a comparision with a toy model can be very helpful in science too. When I give a talk, many people ask this kind of questions a lot:

You have explained this effect by blah-blah theory and blah-blah simulation. 
But is there an intuitive way to understand this (as an experimentalist)?

In fact, I also think hard to provide an intuitive physical picture whenever I prepare for a paper. But I want to say that making an analogy can be dangerous and one has to watch out not to overextend. I think it is important to know where is the starting point of this analogy (most physics analogy is based on a simple toy model, whose result is well-known — such as a harmonic oscillator, a two-level system) and where it fails as the system has more factors that are not taken into account in the model. In my opinion, using an analogy can achieve similar kind of works rather quickly, but it cannot really make a breakthrough which nobody has done.

Well, this thought came up to me while I was thinking about the orbital dynamics. Ananlogy of “arrow dynamics” is helpful, but I concluded that it cannot explain several important features of the oribtal dynamics. At the end, I decided to take a quantum mechanical description, for a reasonably simple but general model.