Moving silicon atoms in graphene with atomic precision

aka The Story of an(other) Article

NOTE: This post is a greatly expanded adaptation of a press release I wrote to publicize our recent work appearing in Physical Review Letters (doi:10.1103/PhysRevLett.113.115501).

A video abstract where I briefly describe the research that is explained in more detail in this blog post (by Viktor Zdrachal from the Österreichische Zentralbibliothek für Physik).

This post explains the science behind the study for the general audience in as simple terms as I can manage.

The main part is followed by a description of getting the study published, intended more for working scientists.

Two years ago I wrote a blog post titled "The Story of an Article" about the process of publishing a research article in ACS Nano. Now, I am again the lead author of a work appearing in Physical Review Letters, likewise the result of an international collaboration. Thanks to the open access policy of my funder, the Austrian Science Fund (FWF), we were able to pay a fee* to make the article free to read for everyone. My sincere thanks to the FWF!

Is it possible to see and control the movement of individual atoms in materials?

Legendary physicist Richard Feynman famously posed a question in his seminal 1959 talk "There is Plenty of Room at the Bottom", which some consider to be a major inspiration for the field of nanoscience – research into the structure and properties of matter in the nanometer scale (10-9 m, or one billionth of a meter). His question was: is it possible to see and control the movement of individual atoms in materials? For a time this vision seemed more science fiction than science, but starting with groundbreaking scanning tunneling microscopy experiments in the late 1980s it has become scientific reality.

In recent years, it has become possible to directly see individual atoms also using electron microscopy. This is naturally extremely valuable for us trying to study matter at these scales. Electron microscopes work in similar way as normal optical microscopes: light that scatters from an object makes it visible to the eye or a camera. However, the wavelength of visible light is so large compared to atoms, molecules or nanostructures that it is not possible to see these using light. Electrons, on the other hand, interact readily with nano-objects, and thus using them instead of visible light allows scientists to image and visualize things like carbon nanomaterials – nanosized materials composed of carbon – or even single atoms. Lenses to focus the electron beams need to use electric and magnetic fields instead of being made of glass, and a fluorescent screen is needed to make the electrons visible. These days these are invariably combined with so-called charge-coupled device (CCD) cameras to record the images. Nonetheless, the basic principle is quite similar, as shown below.

One further detail is worth mentioning: instead of standard transmission electron microscopy (TEM) where a parallel beam of electrons pass through the sample, an alternative is to focus the beam into a very small spot (these days around 1 Ångström, or 1/10 of a nanometer, in diameter), scan the beam over the sample, and measure the electrons that are scattered from each position of the probe. This is called scanning transmission electron microscopy (STEM) and it has certain advantages over TEM, most of which are too technical to get into here. But one thing is important for our purposes: since STEM images are usually formed by collecting electrons that have scattered to certain angles by the sample, heavier atoms that have more positively charged protons in their nucleus cause greater scattering of the electrons, and thus show up bright in the images (so called Z-contrast, Z being the atomic number).

A visualisation of single-layer graphene created by Jani Kotakoski.
A visualisation of single-layer graphene created by Jani Kotakoski.

Unlike light, electrons have mass and carry enough energy that they can cause damage by hitting the nuclei of the material under study.

Electron microscopy is especially useful for studying graphene, which is a one-atom-thick sheet of pure carbon that was discovered in 2004 and received a Nobel Prize in Physics a few years later. This is because electrons passing through such a thin material only interact with a single plane of atoms, making interpretation of the images very straightforward. However, there is a crucial difference in using electrons instead of light to image materials: electrons have mass, and they need to be accelerated to rather high energies for good imaging. This means that the electrons carry enough energy that they can cause damage by hitting the nuclei of the material under study, knocking out atoms in a billiard-ball fashion. For this and other reasons, electron microscopy experiments in graphene have typically involved imperfect structures, or resulted in damage during observation.

With this preliminary information in mind, let's turn to our latest study. For the 6 months or so before the recent publication, I had been leading an international collaboration that grew between two groups here at the University of Vienna — the Tailored Hybrid Structures group of Paola Ayala where I work, and the electron microscopy team of Jannik Meyer — along with two research teams in the UK: a microscopy team at the SuperSTEM laboratory in Daresbury led by Quentin Ramasse, and materials scientists at the University of Manchester led by Ursel Bangert (now at the University of Limerick in Ireland).

The collaboration

In modern nanoscience, computer simulations based on quantum mechanics are a crucial tool for understanding and interpreting experiments.

The study started when myself and my good friend and colleague Jani Kotakoski were looking at the electron beam stability of a related but slightly different system using computer simulations. We were surprised by our predictions, but had no direct means to test whether the findings would bear out in reality. Luckily, I spotted a fresh-off-the-presses article whose authors included Ursel, Quentin, Demie Kapaptsoglou and Recep Zan, again on a similar but related system. We got with touch with them via email, and it soon turned out that Quentin and Demie had been working on a material that was a perfect test case for our findings.

Finally, researchers from the Nion Company headed by Ondrej Krivanek from the US joined the collaboration to help in the analysis and interpretation of the data. Nion is the developer and manufacturer of the cutting-edge electron microscopes used by both our teams. Due to their advanced electron optics (correcting aberrations, or imperfections, unavoidably present in electromagnetic lenses), the Nion microscopes only need to accelerate electrons to energies of 60 kiloelectronvolts (keV, thousands of electron volts) and still be able to see individual atoms. This is lower than most microscopes, which require energies of 80-200 keV for a similar resolution.

A silicon dopant in graphene

The study now published in Physical Review Letters focused on single-layer graphene with silicon atoms embedded into the lattice (such atoms are often called dopants), previously synthesized and characterized by our UK collaborators. Due to the larger size of silicon compared to carbon, these dopant atoms protrude – stick out – from the plane (as shown above), which makes for interesting dynamics under electron irradiation. The detailed computer simulations that I performed in Vienna showed that 60 keV electrons are not energetic enough to likely cause the outright ejection of atoms, in line with what had been observed.

Experimental electron microscopy observation of the (bright) silicon atom jumping from one lattice site to the next.
Experimental electron microscopy observation of the (bright) silicon atom jumping from one lattice site to the next.

Due to the interplay of the moving atoms, the silicon–carbon bond is inverted.

Crucially, however, carbon atoms next to a silicon dopant are slightly less strongly bound than other atoms in the structure. Because of this, they can receive just enough of a kick even from 60 keV electrons so that they almost escape from the lattice, but are recaptured due to an attractive interaction with the silicon atom (basically, the carbon atom would rather be bonded with the silicon and other carbon atoms than by itself). Meanwhile, the silicon relaxes (moves into an energetically more preferred place) into to the lattice position left empty by the impacted carbon atom, which thus lands back into the lattice on the opposite side from where it started. In effect, the silicon–carbon bond is inverted, which was directly seen by the microscopy teams, an example of which is shown below. The video shown at the top of this post also illustrates this process based on the computer simulations.

Visualisation of the silicon-carbon bond inversion by Toma Susi (CC-BY, doi:10.1103/PhysRevLett.113.115501).

We found very good correspondence between the simulations and the experiments.

In total, the SuperSTEM microscopy team observed 38 such jumps, 19 of them in a single unbroken time series (basically, a movie of the silicon atom jumping around), schematically shown in the figure below. We were able to analyze the jumps as a stochastic process (something that happens due to a random influence, which in this case is the impact of a electron to the nucleus of the carbon atom). This gave us an estimate for the probability of the bond inversion, which in turn could be directly compared to the simulations, and a remarkable agreement was obtained. The close correspondence between the simulations (which describe the motion of atoms) and the imaging (which is too slow to see the motion, but gives a solid reason to believe the process really happens) was one of the main reasons our study was accepted for publication in Physical Review Letters.

An illustration of a time series of 19 consecutive "jumps" of the silicon atom in the graphene lattice. By Toma Susi, from the PRL article.
An illustration of a time series of 19 consecutive "jumps" of the silicon atom in the graphene lattice. By Toma Susi, from the PRL article.

Besides being beautiful physics, the findings open promising possibilities for atomic-scale engineering. What makes our results truly intriguing is that the bond flip is directional – the silicon moves to take the place of the carbon atom that was hit by a probe electron. This means that it should be possible to control the movement of one or more silicon atoms in the lattice with atomic precision. So perhaps we’ll see a new kind of quantum corral or a university logo made of atoms of silicon in graphene in the near future :)

That concludes my explanation! I hope you got something out of it, though I'm painfully aware of how difficult it is to talk about research in simple terms.

(I'll now turn to the details of getting a study like this published in a high impact journal, which is probably only interesting if you are a working scientist or looking for a "backstage" view. I do try to keep the discussion below as general as possible, though.)

The road from finding to publication is rarely entirely straightforward.

As often is the case – but rather too rarely talked about in public – the road from a finding to a publication is rarely entirely straightforward. This was the case for us as well. From the beginning, we felt we had very beautiful results, and furthermore, the agreement that emerged between our detailed simulations and the electron microscopy imaging gave us confidence that we could accurately explain what is happening at the atomic scale.

Being career-conscious scientists participating in the academic rat-race, we therefore chose to submit our work to as high an impact factor journal as possible suitable for the topic: in our case, the august Physical Review Letters. (Despite the well-known problems with using impact factors — arguably science's most misused metric — in evaluation, they still are the de facto career currency.) The higher the impact factor of a journal, however, the more submissions they receive, and thus the more stringent their selection criteria. Furthermore, the highest impact journals often have the broadest readership, which means articles accepted there need to be interesting to many of their readers, and written with a general audience in mind.

As is the standard procedure in publishing academic research, the editors of the journal first preliminarily evaluate a submission, and decide whether it seems to be in line with their editorial policy, and the work appropriate and suitably interesting for their journal. In our case we passed this hurdle (after having to condense the manuscript to fit PRL's stringent length limit), giving us hope that we had not seriously overestimated the appeal of our results.

Herein lies the well-known stumbling block of the whole system.

The editors then sent our manuscript out for peer review; that is, they requested other researchers familiar with the topic to read, evaluate and offer criticism of the work. Herein lies the well-known stumbling block of the whole system: if a referee misunderstands or unfairly criticizes the work, it can derail the publication of a deserving work when editors make their decisions on such negative reports — as they must, since they cannot be expected to understand each and every study themselves. For researchers trying to get their work published, these situations can however be frustrating. (However, it must be admitted that these situations can often be somewhat subjective, and thus not taking things personally and making a well-argued case for your own position is really the only way to go.)

After about a month (fairly swift as these things go), we received reports from two referees who had replied to the editors promptly. The first referee was rather dismissive of our work, but without very good justifications or much specific criticism. The second referee seemed to understand the topic better, and supported publication on condition of a few changes, the main of which was that the introduction was not written for a broad enough audience. Fortunately, the editor requested a revision instead of declining the manuscript outright, indicating that they would also send the revision out to new referees. After brief correspondence amongst the coauthors, we revised the manuscript as requested, and wrote a reply arguing against the first referee and for the work to be published in PRL.

Next we had to wait close to two months for word on the revised submission. The revision had gone out to three referees, apparently all of them new. One of the referees explicitly concluded that our reply had been satisfactory, and no communication with the previous negative referee was warranted. They were also the most positive about the revised manuscript, essentially recommending publication with a few minor clarifications. A second referee mentioned that while they considered the research to be of high quality, they had struggled to reach a decision on whether the work was suitable for PRL; in the end, they narrowly recommended publication.

The third referee became the sticking point.

However, the third referee became the sticking point. They were critical of the work, and recommended sending it to Physical Review B instead of PRL. Probably swayed by this opinion, the editor had decided to decline the manuscript, stating that it did not meet "the Physical Review Letters criteria of impact, innovation, and interest". After carefully reading all the comments and discussing with my coauthors, however, many of us became convinced that the third referee had misunderstood crucial aspects of the work, and furthermore, that we could justifiably argue that our work indeed met all the criteria. Others of us advised taking the easy road out and resubmitting instead to PRB or another lower impact journal, but weren't against an appeal if the lead authors wished to do so.

Physical Review Letters has an explicit resubmittal policy for situations such as these, which surely arise quite often. If the authors are not happy with a decision, they can submit a formal appeal, possibly along with a revised version of the manuscript. The editor will then consult a member of the PRL Editorial Board, a so-called Divisional Associate Editor (DAE), who will either decide the case outright, or send the manuscript out for additional peer review. We made a few changes, wrote a carefully argued appeal letter, and resubmitted with fingers crossed.

Believing in ourselves and being persistent paid off this time!

Another three weeks passed by until we got the judgment from the associate editor — and to our great joy, their comments were very positive! Thus no new review was needed, and our article was immediately accepted for publication.