Symposium X—Frontiers of Materials Research

SymposiumX_800WidthErik Bakkers, Technische Universiteit Eindhoven

Bottom-Up Grown Nanowire Quantum Devices

Written by Don Monroe

Despite years of effort, scientists have yet to agree on a physical system for realizing the tremendous potential of quantum computing. Erik Bakkers described a relatively new strategy based on networks of epitaxial III-V nanowires that could exploit exotic collective electronic excitations to enable these applications.

In principle, quantum computing provides exponential increases in computational power from a relatively small number of quantum bits, or qubits. Such qubits can simultaneously represent two possible quantum states, but only until interaction with the environment destroys their coherent relationship. “Decoherence is the big problem of a quantum computer,” Bakkers said. “This is really the fundamental bottleneck.”

Bakkers and his collaborators have been exploring a strategy for overcoming this challenge using “Majorana fermions,” which are their own antiparticle and should be highly resistant to decoherence because they have “no charge, no spin, and no energy.” These entities were proposed decades ago as a model for neutrinos, but have recently been suggested to occur as quasiparticles in condensed-matter systems, in particular in one-dimensional superconductors. “If we can find this particle and control the quantum state, we could have very long decoherence time,” Bakkers said.

His group has been looking for this elusive particle in the proximity-induced superconductive state of InSb nanowires, whose electrons have very low effective mass, strong coupling to magnetic fields, and high spin-orbit coupling. Tunneling spectroscopy revealed the expected conductance peak in the center of the superconducting energy gap. “The data is all consistent with having Majoranas,” Bakkers concluded, although the first experiments were limited by a high density of states in the gap.

To make the wires, Bakkers’ team used an established method in which a ball of metal forms a eutectic with a semiconductor, and subsequent layer-by-layer growth on (111) facets produces long, highly uniform single-crystal nanowires. They developed ways to reduce As and P impurities, grow wires up to 60 µm long, and improve the interfaces, achieving low-temperature mobilities as high as 60,000 cm2/V·s. These nanowires produced extremely clean-induced superconducting gaps, and a magnetic field produces a clear mid-gap state with the predicted quantum conductance of 2e2/h. “We believe this is a very strong signature of having these Majorana states,” Bakkers said.

Exploiting these states for quantum computing will require fashioning these high-quality wires into loops and other circuits. Bakkers showed how to create such structures within the evacuated growth chamber. The gold balls that seed the nanowire growth were defined lithographically on (111) facets of a V-groove etched into a (100) substrate. Choosing the size and location of these seeds let the team create long wires that cross, merge, or shadow each other during later deposition.

In particular, Bakkers described “hashtag” structures (#) with pairs of InSb wires from opposing faces that merge to make “very high-quality wire-wire junctions,” although there can be twin boundaries at the interfaces. Painstakingly dislodging these structures in the electron microscope with a micromanipulator and transferring them to a separate substrate for electrical contacting allows their electrical characterization. The junctions showed the quantized conductance characteristic of ballistic transport, and the oscillating conductance as a function of magnetic field through the loop of the hashtag indicated a coherence length as large as 60 µm. Majorana-based qubits will require somewhat more complex structures, but these appear to be feasible.

Still, Bakkers noted that these techniques are “nice for academic studies, but not really scalable.” He and his collaborators are therefore exploring an alternative future path based on in-plane selective-area growth to create arbitrarily complex circuits in vacuum.

Symposium X—Frontiers of Materials Research features lectures aimed at a broad audience to provide meeting attendees with an overview of leading-edge topics.

Symposium X—Frontiers of Materials Research

SymposiumX_800Width 230x230Mariana Bertoni, Arizona State University

What Is Next for Solar PV Technology?

Written by Arthur L. Robinson

Despite all the excitement now directed at perovskites as a candidate material for solar cells, Mariana Bertoni made it clear at the outset of her Symposium X presentation Wednesday that her vision of the future of solar photovoltaic technology is firmly grounded in silicon. Beginning with a look at progress in lowering the cost of solar technology, Bertoni then provided a tour through the many challenges and obstacles to its reaching the goal of becoming a mainline energy technology. The breadth of these challenges from basic materials performance to integration of the technology into daily life is evident from one many of us might not have thought of—aesthetics. For example, rooftop solar installations are, to be blunt, ugly, which can dampen consumer acceptance. As a result, solar shingles that address this issue are now coming online.

The world’s installed solar energy capacity is rapidly growing, up from 8 GW in 2007 to 402 GW in 2017 and is on track to add 120 GW/year after 2019, which will result in 1 TW installed capacity (residential, commercial, and utility) by 2023. Looking at the levelized (lifetime) cost of solar photovoltaic energy from 2009 to 2017 reveals that this cost has dropped so significantly that the cost per megawatt-hour for utility-scale installations is already comparable to that of natural gas and less expensive than coal and nuclear energy, while rooftop-installation costs remain in the nuclear range. A long-time US Department of Energy cost goal of $1 per Watt, initially greeted with incredulous dismay, was met in 2017, three years earlier than projected, and new reports from China suggest $0.30 per Watt or less will soon be achievable.

Bertoni likened today’s crystalline silicon technology to that of a Ford Model T: it is an established technology that has demonstrated its affordability and its potential but we are just getting started. For example, efficiency is a major performance factor for solar cells. Silicon (single crystalline and polycrystalline) dominates the market today with 90% of market share. In comparison to the theoretical efficiency limit of 29.1%, polycrystalline silicon solar cells were at 21.5% in 2017 and single-crystalline cells up to 26.6%, within 10% of the theoretical limit. Still, increased efficiency remains as a major opportunity area, joining increased yield and lower manufacturing and processing costs, for progress toward $0.03/kW-hr. Long-term performance, degradation, system-level and cell and module integration, and materials design provide further opportunities for cost lowering.

The solar photovoltaic production process begins with ingots or boules and proceeds in stages through wafers, photovoltaic cells, modules, and the installed system. The major cost elements are the silicon wafer (40%), the metallization (10-20%), and the module (comparable to the wafer cost). Among the examples she cited, Bertoni reviewed the use of a spalling technique to avoid the loss of material inherent in using a wire saw to cut wafers from a boule of crystalline silicon. In this technique, pulling a stressor layer attached to the end of the boule cleaves a silicon layer with thickness determined by the stressor layer and there is no material lost, although surface quality remains an issue. Then turning to the metallization process for the silver front contacts in the form of metal fingers that are narrow to expose the silicon to as much light as possible. A particle-filled paste is less expensive than pure silver but has a 10 times higher resistivity. A reactive ink process yields a resistivity comparable to pure silver, and solar cells made in this way have high performance.

When it comes to long-term performance, it turns out that changes in degradation rates are important in determining the power output over time; two different paths to the same power output in the far future can yield different integrated lifetime power production. Bertoni discussed imaging methods to investigate what causes degradation in modules due to module cracking. For example, stress levels concentrated near metal ribbons may depend on the encapsulation method used. Mapping water distribution and its effects in encapsulants is another example. And modeling how sodium from the front glass above the encapsulant diffuses through a module is a third. Interfaces also contribute to long-term performance degradation, with one example being an increase of series resistance with time due to surface recombination resulting from an increasing density of states at the interface.

Bertoni concluded her talk with a discussion of tandem solar cells with a good base cell and a wide-bandgap material with a high efficiency in the upper cell, which provide a way to increase the cell efficiency above 29%. A GaAs-Si tandem is the best current configuration. Even with high efficiency, the system cost can be a problem. Bertoni showed an analysis showing that tandem cells are never the best choice for utility-scale facilities, whereas for residential systems they are viable provided that the GaAs cost is low enough (less than $260/m2). Perovskites can play a role in the analysis if the top and bottom cells are both inexpensive relative to the overall cost of the system. From this, it appears that tandem cells will enter the market and compete with their silicon predecessors. In her final remark, Bertoni put in a plug for machine-learning in the search for new materials.

Symposium X—Frontiers of Materials Research features lectures aimed at a broad audience to provide meeting attendees with an overview of leading-edge topics.

Symposium X—Frontiers of Materials Research

Symp X-Kim-blogDae-Hyeong Kim, Seoul National University

Nanomaterials-Based Flexible and Stretchable Bioelectronics

Written by Don Monroe

Dae-Hyeong Kim of Seoul National University used three projects from his group to illustrate the potential for flexible and stretchable electronics for medical applications.

Kim’s first example concerned non-invasive monitoring of glucose levels for people with diabetes. Either high and low glucose is dangerous, and levels are currently controlled with frequent, painful blood measurements and insulin shots.

Kim’s research team developed devices to monitor glucose painlessly using sweat, rather than blood. The 100-fold lower glucose concentration is harder to measure, but Kim noted that the sweat and blood glucose concentration are highly correlated. Electronics incorporated in a patch or other device on sweating skin need to be stretchable, however.

The team improved the sensitivity by using nanostructures to increase the surface area. They exploited the same glucose oxidase enzyme used in current glucose strips, which generates electrically conductive species from glucose. This reaction is sensitive to temperature, humidity, and pH, however, so the team integrated sensors for these environmental variables to provide accurate measurements. Kim also noted that accuracy improved when participants followed a well-defined protocol, such as exercise or a foot bath to encourage sweating.

A later-generation device replaced the re-usable patch with a disposable strip, avoiding degradation of measurement accuracy over time. This device also included sensors for heart rate, blood oxygenation, and motion, and can be wirelessly connected to a cell phone or electronics. Kim illustrated monitoring of volunteers over 30 days, showing clear glucose concentration increases after meals, matching well with blood tests.

Ideally the glucose monitor will be coupled to integrated drug delivery. Kim described a transdermal delivery system using microneedles made of a biodegradable polymer containing a drug such as insulin. To control administration, his team has used a protective layer that can be melted away by applying heat to expose the polymer.

In his second example, Kim described an implantable retinal prosthesis. Previous external imagers are bulky and have low resolution, he noted, while fully implanted chips based on rigid silicon technology cover only a small fraction of the visual field and degrade over time due to their poor mechanical match with tissue.

Kim’s group developed a soft large-area image-sensor array based on MoS¬2 and graphene, extending the area using an icosahedral arrangement of patches. The overall thickness of the array was less than 1 µm, and it was connected to a flexible printed-circuit board using flexible cable. Animal experiments showed an evoked potential in the visual cortex reflecting transmission of the detected image. This “proves the system was working quite well,” Kim said.

Kim’s third example addressed heart failure, which reflects inefficient pumping due to erratic and poorly synchronized heart-tissue contraction. Current implanted-device strategies include electrical pacing by localized electrodes, which generally increases pumping by about 10%, or an external mesh that externally compresses the ventricles.

“Our approach combines these two approaches,” Kim said, by creating a flexible, conductive mesh for extended stimulation. Commercially available conductive rubbers are too resistive, so the team developed a material based on silver nanowires, which increased pumping by 50%.

“The effect is quite meaningful, but there are many issues for clinical translation,” Kim noted, mostly related to materials. In particular, the silver wires readily oxidize, increasing resistance, and dissolved silver is toxic to cells and tissues. The researchers solved these issues by passivating the wires with a gold sheath.

They also improved the tradeoff between flexibility and electrical conductivity by modifying the curing process to create phase segregation between nanowire-rich and polymer-rich regions. The resulting material showed up to 840% stretch with a conductivity up to 72,000 S/cm.

Symposium X—Frontiers of Materials Research features talks aimed at a broad materials audience to provide meeting attendees with an overview of leading-edge topics.

Symposium X—Frontiers of Materials Research


SymposiumX200x200Sergei V. Kalinin, Oak Ridge National Laboratory

The Lab on a Beam—Big Data and Artificial Intelligence in Scanning Transmission Electron Microscopy

Written by Arthur L. Robinson

In his Symposium X presentation Monday, Sergei Kalinin of the Oak Ridge National Laboratory presented his vision for a future in which there is a marriage joining materials data at the atomic level and big data, including machine learning and artificial intelligence. In his vision, this marriage can position scanning transmission electron microscopy (STEM) to transition from a purely imaging tool to an atomic-scale laboratory of electronic, phonon, and quantum phenomena in atomically engineered structures.

In what would become a theme running through his talk, Kalinin contrasted the traditional question of “how can we make our imaging tools better” to “what can we learn from these images?” For example, aberration-corrected STEM became commercially available about 10 years ago and opened the way for wide-spread atomic-resolution imaging in which atomic positions can be determined down to the picometer level in a variety of environments beyond the ultra-high vacuum previously required for this kind of resolution. A floodgate has opened for a huge flow of information, all available from STEM experiments. But what can we do with that information; how can we interpret it?” Kalinin asked.

As an example, Kalinin showed a video of an electron beam knocking out sulfur atoms one at a time while imaging MoS2, in effect watching a chemical reaction as it occurs. But what is really wanted is a way to convert the images to information about atomic coordinates and trajectories. It turns out that a neural network can be a useful way to address this task, once you are able to teach the network the physics it needs to identify features of interest, he said. For example, a downloadable deep learning network very accurately identified the mix of breeds in a dog from a photograph but also kept finding animals in an electron micrograph.

Kalinin’s Oak Ridge group has developed the tools to convert the image data into high-quality representations and they are publicly available. With these tools, the Oak Ridge group is constructing defect libraries cataloging defects that actually do occur in a given material. Kalinin pointed out the similarity that now arises to astrophysics and high-energy physics where the individual researcher model has largely disappeared and work is done via communities sharing resources, and asked whether materials researchers should adopt a similar approach.

Returning to his theme of what can materials researchers learn from the information, Kalinin illustrated the case of surfaces and interfaces, defects, and spatially inhomogeneous ferroelectric materials for which theories for bulk behavior are well in hand. Specifically, he showed that the value of the flexoelectric coupling of polarization and strain gradient in heterostructures of lead and strontium titanate can be determined from the shape of vortices in STEM images. Moreover, one can choose microscopic models from atomic-resolution image data, as Kalinin illustrated with the example of a superconductor comprising a tellurium-selenium solid solution in one plane and iron in the other. The segregation of the tellurium and selenium is controlled by a single parameter, the specific enthalpy for segregation. With neural networks, one can reliably determine the value of this parameter from the atomic positions and from this, determine the entire phase diagram.

These examples have been based on static images, but dynamic imaging is possible and brings us into the realm of chemistry. In one example, tungsten sulfide with molybdenum impurities and sulfur vacancies created by the electron beam, Kalinin described how to use neural networks to identify all the defects and then plot their trajectories in time. From this information, dominant point defects can be identified, diffusion parameters analyzed, and transformation pathways of composite defects studied.

For the next part of his presentation, Kalinin turned to the possibility of manipulation of matter and fabrication of structures atom by atom with the use of an electron beam. Parts of the puzzle to be solved include finding phenomena that are amenable to control by an electron beam, learning how to control the beam to follow different trajectories, expanding the dynamic range, incorporating feedback from other measurement tools, and ideally having an appropriate theory at hand. For example, the group incorporated feedback from the beam detector to tell the control system that crystallization had occurred and to move the beam.

An early example of automation was combining STEM with the control system that allows scanning probes to write pattern. In this case, the electron beam causes the crystallization of amorphous strontium titanate an atom at a time to write the ORNL pattern in the bulk surrounded by amorphous material. The process also works with silicon, where it was also demonstrated that controlled dopant profiles could be created by moving the impurities with the electron beam.

Looking further in the future, Kalinin turned to learning from Nature. Insects, for example, are in one sense ultimate machines but they are not controllable and cannot be miniaturized to the nanolevel. If one tries to build a nanomachine with similar capabilities a host of issues such as thinking, locomotion, and energy sources arise from trying to integrate too many functions in the nanomachine. Kalinin asked, “What if we use a single atom or small atom assembly as a functional element of a moving nanomachine and defer control and power functions to external entities?” As a first example, he discussed the use of an electron beam to move a cluster of silicon atoms on graphene.

In his summary, Kalinin pointed out the evolution of imaging from description to control. We are now at the stage where dynamic microscopies can tell us what atoms do. We still need to find out: what are local atomic functionalities; why do atoms do what they do; how can we direct them to do what we want.

Symposium X—Frontiers of Materials Research features talks aimed at a broad materials audience to provide meeting attendees with an overview of leading-edge topics.

Symposium X - David J. Mooney introduced fascinating founding in biomaterials

David J. Mooney,  a Professor of Bioengineering at the Harvard School of Engineering and Applied Sciences as well as a core faculty member at the Wyss Institute for Biologically Inspired Engineering at Harvard University, was invited as an honored speaker in the MRS Symposium X meeting. Mooney is a member of both the National Academy of Engineering and the National Academy of Medicine. The Mooney lab focuses on designing biomaterials that affect specific cells functions and making therapies more effective and practical through study of the mechanism of the chemical and mechanical signal that were sensed by cells. His research now focuses on therapeutic angiogenesis, regeneration of musculoskeletal tissues and cancer therapies. In the meeting, he introduced the influence of stress relaxation of hydrogel on cells and demonstrated that faster relaxation of gels promote cell spreading and enhance osteogenesis and new bone formation. Moreover, he put forward that ferrogel with magnetic stimulation can promote new tissue regeneration because active mechanical stimulation share a similar mechanism.