Introduction

Ever since Democritus proposed the existence of atoms as the smallest indivisible unit of matter around 400 bc, this idea and its consequences have captured the attention of the scientific community. In the 19th century, broad adoption of atomistic theory led to remarkable progress in physics and chemistry, and eventually to the dawn of quantum physics in the early 20th century.

The earliest direct observation of atoms originally dates to the first field-ion microscope in the early 1950s. ^{Reference Muller1} The first lattice fringes were observed in the 1950s by Menter using an electron microscope, building on the developments of Ruska, ^{Reference Knoll and Ruska2} von Ardenne, ^{Reference von Ardenne3,Reference von Ardenne4} and others, ^{Reference Menter5} with gradually improving resolution over the following decades. Direct observation of individual atoms in scanning transmission electron microscopy (STEM) was achieved by Crewe and co-workers in the 1960s, ^{Reference Crewe, Wall and Langmore6} made possible by his development of the cold-field-emission gun. ^{Reference Crewe, Eggenberger, Wall and Welter7} Using an annular detector, direct imaging of atomic structures became possible, ^{Reference Wall, Langmore, Isaacson and Crewe8–Reference Pennycook and Jesson10} and in the last 30 years, the progress in STEM ^{Reference Pennycook and Varela11} has rendered such observations routine. In comparison, scanning tunneling microscopy (STM) ^{Reference Binnig, Quate and Gerber12–Reference Binnig and Rohrer14} began with direct demonstration of atomic resolution on Si surfaces, even though earlier examples of current- and force-based mesoscopic profilometry are available.

Since the observation of atoms and the emergence of modern physics, a loftier dream has developed, as summarized by R. Feynman in his seminal talk “There is plenty of room at the bottom.” ^{Reference Feynman15} The modern field of nanotechnology, however, owes much to the experimental work of D. Eigler at IBM, ^{Reference Eigler and Schweizer16} and the visionary work of E. Drexler. ^{Reference Drexler17} The letters “I,” “B,” and “M” written in Xe atoms on a copper surface ^{Reference Eigler and Schweizer16} and the book “Engines of Creation” ^{Reference Drexler17} have provided the impetus by becoming firmly imprinted in societal perception of nanoscience as a pathway to control matter on the atomic scale, creating the machines and devices for ultimate medical devices and nanoassemblers. The evolution of social acceptance of nanotechnology is illustrated in Figure 1 .

Figure 1. (a) Google Ngram search showing the growing frequency of the words “nanotechnology” and “nanoscience” in online books. (b) Number of publications per year on “electron microscopy” and (keyword), according to ISI.

The first paradigm: Molecular machines

The first paradigm of nanotechnology was popularized by Drexler, who proposed the controlled chemical synthesis of molecular entities capable of autonomous motion and performing work, the nanomachines and nanoassemblers. ^{Reference Drexler17} This concept immediately captured the minds and imagination of broad swaths of the scientific community. ^{Reference Coskun, Banaszak, Astumian, Stoddart and Grzybowski18,Reference Browne and Feringa19} Experimentally, efforts toward light, ^{Reference Ruangsupapichat, Pollard, Harutyunyan and Feringa20,Reference Saha and Stoddart21} chemically ^{Reference Fletcher, Dumur, Pollard and Feringa22} and electrochemically ^{Reference Balzani, Gomez-Lopez and Stoddart23,Reference Kudernac, Ruangsupapichat, Parschau, Maciá, Katsonis, Harutyunyan, Ernst and Feringa24} activated molecular machines were undertaken, as recognized by the award of the 2016 Nobel Prize in Chemistry to J.-P. Sauvage, J. Fraser Stoddart, and B.L. Feringa.

Realization of this approach to nanoscience requires the simultaneous solution of three problems, namely, the design of molecular blocks with the required functionalities, the development of the synthetic pathways, and assembly in operational supramolecular structures. The power of modern computational methods makes the first viable, if not yet fully accomplished. Synthesis represents a more complex problem, traditionally requiring outstanding organic synthesis intuition and broad knowledge base. However, the recent establishment of reaction databases combined with advances in graphical search algorithms has enabled automatic identification of synthetic pathways for all known and many unknown (e.g., via retrosynthesis) compounds. ^{Reference Szymkuć, Gajewska, Klucznik, Molga, Dittwald, Startek, Bajczyk and Grzybowski25,Reference Cadeddu, Wylie, Jurczak, Wampler-Doty and Grzybowski26} However, it is the probing and assembly of molecular machines into operational structures that remains a central issue in the field.

The second paradigm: Scanning probe manipulation

The second paradigm of nanotechnology is based on direct atomic manipulation by scanning probes. From the early days of scanning tunneling microscopy, it was realized that the probe tip can induce the motion of loosely bound atoms on material surfaces. In some cases, the surface can be imaged prior to and after the manipulation, providing atomically resolved views of surface changes. This provided the ultimate opportunity to study cause and effect on the atomic scale.

The real breakthrough came with the experiments of Eigler at IBM, who developed an approach to position deposited atoms in predefined configurations. ^{Reference Eigler and Schweizer16} In the 25 years since this work, the field has seen multiple advances in fabrication, basic physics, and societal impact. Some of the breakthrough concepts introduced by the Eigler group include quantum corrals ^{Reference Heller, Crommie, Lutz and Eigler27} and molecular cascades, ^{Reference Heinrich, Lutz, Gupta and Eigler28} opening pathways for probing the fundamental physics of quantum states in real space and molecular motion-based computational devices. Other advances, including ultrahigh density storage ^{Reference Donati, Rusponi, Stepanow, Wäckerlin, Singha, Persichetti, Baltic, Diller, Patthey, Fernandes and Dreiser29} and holographic memories have been reported. The movie The Boy and His Atom ³⁰ captured public interest in this, collecting more than 6 M views on YouTube. Many of these advances are summarized in a number of recent reviews, and some of the highlights are shown in Figure 2 . ^{Reference Morgenstern, Lorente and Rieder31–Reference Benjamin and Kelly35}

Figure 2. Examples of single-atom manipulation with scanning probes. (a) Xenon atoms on nickel(110) surface forming “IBM.” Image originally created by IBM Corporation. (b) Carbon monoxide man on platinum(111). Image originally created by IBM Corporation. (c) Scanning tunneling micrograph of a four-wheeled “nanocar” on a Au(111) surface. Adapted with permission from Reference Shirai, Osgood, Zhao, Kelly and TourReference 33. © 2005 American Chemical Society. (d) Quantum corral, iron on copper(111). Image originally created by IBM Corporation. (e) Single-atom transistor with a single Si atom. Note: S, source; D, drain. Reprinted with permission from Reference Fuechsle, Miwa, Mahapatra, Ryu, Lee, Warschkow, Hollenberg, Klimeck and SimmonsReference 34. © 2012 Nature Publishing Group. (f) Nb and Al circuits on a Si substrate, forming a multiple-qubit device. Reprinted with permission from Reference Benjamin and KellyReference 35. © 2015 Nature Publishing Group.

The fundamental limitation of STM-based fabrication is that in many cases, it is limited to low temperatures, with liquid helium temperature (4 K) being the norm, and requiring atomically flat clean surfaces. These approaches are also fairly slow and have limited throughput. At the same time, mainstream nanotechnology necessitates room-temperature stability (either at operational temperatures or intermediate fabrication steps), and reasonably fast fabrication. Simmons et al. ^{Reference Schofield, Curson, Simmons, Rueß, Hallam, Oberbeck and Clark36,Reference Fuechsle, Miwa, Mahapatra, Ryu, Lee, Warschkow, Hollenberg, Klimeck and Simmons37} and several other groups demonstrated the achievement of this goal by a combination of STM manipulation with classical surface science techniques for fabrication of atomically defined surfaces.

The third paradigm: Electron beams

The introduction of high-resolution aberration-corrected electron microscopy and related progress in electron energy-loss spectroscopy (EELS) in the early 2000s have revolutionized the field of condensed-matter physics and materials science. Following the initial demonstration of single-atom sensitivity in EELS ^{Reference Varela, Findlay, Lupini, Christen, Borisevich, Dellby, Krivanek, Nellist, Oxley, Allen and Pennycook38} and three-dimensional (3D) imaging capability via focal series, ^{Reference Borisevich, Lupini and Pennycook39} the increased spatial resolution and sensitivity of STEM have enabled advances such as direct mapping of polarization, ^{Reference Chang, Kalinin, Morozovska, Huijben, Chu, Yu, Ramesh, Eliseev, Svechnikov, Pennycook and Borisevich40,Reference Chun-Lin, Nagarajan, Jia-Qing, Houben, Zhao, Ramesh, Urban and Waser41} in-plane octahedral tilts, ^{Reference Borisevich, Chang, Huijben, Oxley, Okamoto, Niranjan, Burton, Tsymbal, Chu, Yu and Ramesh42} and chemical strains, ^{Reference Kim, He, Biegalski, Ambaye, Lauter, Christen, Pantelides, Pennycook, Kalinin and Borisevich43,Reference Yankovich, Berkels, Dahmen, Binev, Sanchez, Bradley, Li, Szlufarska and Voyles44} and was recently extended to probe tilt systems in the z-direction. ^{Reference He, Ishikawa, Lupini, Qiao, Moon, Ovchinnikov, May, Biegalski and Borisevich45} Advances in the quantification of STEM have determined the location of single vacancy centers, ^{Reference Kim, Lin, Weickert, Kenzelmann, Bauer, Ronning, Thompson and Movshovich46} and are likely to lead to further breakthroughs. In parallel with the mainstream development of STEM as a purely imaging/spectroscopy tool, a number of groups have explored the potential of atomically focused beams for resist-based lithography, ^{Reference Manfrinato, Wen, Zhang, Yang, Hobbs, Baker, Su, Zakharov, Zaluzec, Miller and Stach47,Reference Manfrinato, Zhang, Su, Duan, Hobbs, Stach and Berggren48} now approaching 1-nm resolution, and e-beam deposition. However, in the vast majority of work to date, STEM has been perceived only as an imaging tool, and any beam-induced modifications were viewed as undesirable beam damage.

A brief historical overview of e-beam, ion beam, and particle beam literature suggests that during the last three decades, these beams were found to induce significant modification in the structure of solids. One example is beam-induced crystallization and amorphization. This area was actively explored in the 1980s and 1990s, and e-beam crystallization of a number of important semiconductors such as Si ^{Reference Dai, Zhao, Xie, Cai, Wang and Zhu49–Reference Jencic, Bench, Robertson and Kirk52} and GaAs ^{Reference Jencic, Bench, Robertson and Kirk52–Reference Li54} has been reported. Similarly, the beam can result in selective removal of material, and when integrated with beam-induced reactions, it can enable fabrication of nanoscale structures, as summarized in recent reviews by Krasheninnikov, ^{Reference Krasheninnikov and Nordlund55} Gonzales-Martinez, ^{Reference Gonzalez-Martinez, Bachmatiuk, Bezugly, Kunstmann, Gemming, Liu, Cuniberti and Rümmeli56} and Jesse. ^{Reference Jesse, Borisevich, Fowlkes, Lupini, Rack, Unocic, Sumpter, Kalinin, Belianinov and Ovchinnikova57} The associated mechanisms are discussed by Jiang. ^{Reference Jiang58} However, these beam-fabrication processes primarily explored mesoscopic-level changes in materials structure, as limited by the electron-microscopy platforms of the time.

In the last five years, the proliferation of high-resolution STEMs and their intrinsic propensity for beam-induced modifications in solids have led several groups to explore and report atomic-level beam-induced modifications, including phase transitions, vacancy creation, and atomic motion, as summarized in Table I . ^{Reference Dai, Zhao, Xie, Cai, Wang and Zhu49,Reference Krasheninnikov and Nordlund55,Reference Jesse, He, Lupinin, Leonard, Oxley, Ovchinnikov, Unocic, Tselev, Fuentes-Cabrera, Sumpter, Pennycook, Kalinin and Borisevich59–Reference Unocic, Lupini, Borisevich, Cullen, Kalinin and Jesse69} Some of these are further illustrated in Figure 3 . ^{Reference Hart, Liu, Lang, Hubert, Zukauskas, Canalias, Beanland, Rappe, Arredondo and Taheri62,Reference Kim, Jang, Kim, Cho, Kang and Lee70–Reference Ishikawa, Lupini, Findlay, Taniguchi and Pennycook76} What is remarkable is that these changes often involve one atom or small groups of atoms, and can be monitored in real time with atomic resolution. Further examples of controllable electron-beam-induced phenomena are discussed throughout this issue.

Table I. Atomistic and nanoscale beam-induced phenomena.

Figure 3. Examples of the electron-beam- and ion-beam-induced changes in materials structure resolved at atomic or near-atomic level. (a) Sculpting: scanning electron microscope images before and after straightening carbon nanotubes by Ar ion irradiation. Reprinted with permission from Reference Kim, Jang, Kim, Cho, Kang and LeeReference 70. © 2003 Elsevier. (b) Crystallization/amorphization: structural changes of Sr₂Nd₈(SiO₄)₆O₂ under electron irradiation. Reprinted with permission from Reference Bae, Zhang, Weber, Higuchi and GiannuzziReference 60. © 2007 AIP Publishing. (c) Phase transitions: phase transition from γ-CaSo₄ to β-CaSo₄. Adapted with permission from Reference Cao, Zheng, Jia, Bai, Li, Sheng, Wu, Han, Li, Wen and YuReference 71. © 2015 American Chemical Society. (d) Atomic column rearrangement: annular dark-field images showing phase front advancement in transition from Mn₃O₄ to MnO. ^{Reference Pennycook, Jones, Pettersson, Coelho, Canavan, Mendoza-Sanchez, Nicolosi and Nellist72} (e) Domain switching: electron-beam-induced domain switching in Rb-doped KTiOPO₄. Reprinted with permission from Reference Hart, Liu, Lang, Hubert, Zukauskas, Canalias, Beanland, Rappe, Arredondo and TaheriReference 62. © 2016 American Physical Society. (f) Bond formation: experimental (right) and simulated (left) images of bond formation in perchlorocoronene under electron-beam irradiation. Adapted with permission from Reference Chamberlain, Biskupek, Skowron, Markevich, Kurasch, Reimer, Walker, Rance, Feng, Müllen, Turchanin, Lebedeva, Majouga, Nenajdenko, Kaiser, Besley and KhlobystovReference 73. © 2017 American Chemical Society. (g) Vacancy formation in graphene sheet after irradiation by a focused electron beam. Adapted with permission from Reference Robertson, Allen, Wu, He, Olivier, Neethling, Kirkland and WarnerReference 74. © 2012 Macmillan Publishers Ltd. (h) Vacancy ordering in SrCoO_2.7. Reprinted with permission from Reference Mefford, Rong, Abakumov, Hardin, Dai, Kolpak, Johnston and StevensonReference 75. (i) Atomic motion: movement of Ce dopant laterally within wurtzite-type aluminum nitride. ^{Reference Ishikawa, Lupini, Findlay, Taniguchi and Pennycook76}

This broad range of well-defined beam-induced processes suggests tremendous potential for materials science, chemistry, and nanofabrication. Some examples of direct e-beam fabrication are illustrated in Figure 4 . ^{Reference Manfrinato, Stein, Zhang, Nam, Yager, Stach and Black77–Reference Kalinin79} Imagine being able to implement the same spectrum of phenomena—or more—with an electron beam in the bulk as with scanning probe microscopy on a material’s surface—from fabrication of single-atom devices, atom-by-atom assembly, editing defect configurations in solids, to the realization of electron-beam-driven molecular motors, and more.

Figure 4. Examples of electron-beam-based fabrication. (a) Scanning transmission electron microscope image of a nanoscale topographic map of the world drawn using electron-beam-induced deposition and a tungsten precursor. Adapted with permission from Reference Manfrinato, Stein, Zhang, Nam, Yager, Stach and BlackReference 77. © 2008 World Scientific. (b) Aberration-corrected electron-beam lithography in poly(methyl methacrylate). ^{Reference Kalinin, Borisevich and Jesse78} (c) IFIM. (d) Text “ORNL” patterned on the interface of amorphous and crystalline SrTiO₃. ^{Reference Kalinin79}

At this point, we are only at the beginning of this pathway. The observed phenomena at this point are poorly controlled (e.g., both crystallization and amorphization can be induced by electron beams for slightly different parameters, but we don’t know why or how to predict it), and the associated mechanisms acting at the atomic level are poorly understood.

Achieving further progress and harnessing beam-induced atomic motion and fabrication for nanofabrication requires solving a set of interlinked problems. The first is the generation of forward-predictive models for electron-beam–matter interactions that take into account both elastic and inelastic scattering events and allow calculation of the sequence of changes in electronic and atomic structure of materials during electron irradiation. Such models are intrinsically complex since they aim to link the stochastic aspects of energy transfer from keV electrons with the electronic and ionic subsystem over a time scale ranging from femtoseconds to seconds. Development of two temperature or equivalent models coupled with molecular dynamics simulations is a step in the right direction. The key step forward from the present state of the art will be to transition from single-particle simulations to the effects of multiple particles in a larger volume that will allow researchers to elucidate the statistical parameters of the process, classify the probabilities of atomic knock-on events and resulting displacements, and also quantify low-probability processes. Similarly of interest are the models that explore the evolution of charges and electron and ion currents in the irradiated area.

Complementary to this will be the creation of a library of structures and beam-induced transformations based on experimental observations, akin to the reaction pathway analysis in biochemical and catalysis communities. This will, in turn, necessitate fully harnessing data flow from detectors, the creation of rapid image analytic tools to identify observed atomic structures from local imaging, also ptychography (imaging based on analyzing the diffraction signal from STEM), ^{Reference Rodenburg, McCallum and Nellist80–Reference Hüe, Rodenburg, Maiden, Sweeney and Midgley82} and establishing common knowledge spaces to integrate information from multiple microscopic platforms (similar to research models in astronomy or genomics). This requires deconvolution of the microscope transfer function (a measure of resolution and performance) and calibration, so that materials-specific phenomena are separated from instrumental factors. It also requires establishing common file formats and analysis tools. When available, these experimental cause and effect libraries can be compared and used to improve theoretical ones. With the information on cause and effect at hand, advances in machine learning can be used to produce required beam trajectories and parameters to enable atomic-level fabrication.

Perspective

Electron-beam control and direction of matter will yield multiple research opportunities in areas spanning basic and applied science and nanofabrication. In basic science research, the dual potential of STEMs to image atomic structure and probe associated electronic properties combined with beam atom-by-atom fabrication opens pathways to create new atomic configurations and probe their electronic, plasmonic, and phononic properties. Similarly, while the early stages of e-beam-induced phenomena are intrinsic to this field, observations of later stages of molecular, vacancy, and single-atom dynamics can provide insights into atomically resolved mechanisms of phase transitions, chemical, and electrochemical reactions.

Similarly, the ability to create predefined atomic configurations in the bulk, or edit STM- or lithographically fabricated devices offers tremendous potential for atomically precise device fabrication, ranging from single-spin magnetoelectronics to quantum and exotic phonon devices. While the throughput of the e-beam (as any sequential) technique will be limited, it is likely to surpass STM-based fabrication in terms of speed. Further introduction of multibeam systems can accelerate the enabling process (e.g., 100-qubit device fabrication). Initial progress is likely to be assisted by a large fleet of extant STEM platforms that can be repurposed for these applications.

In this Issue

This issue of MRS Bulletin assembles contributions that detail recent progress in electron- and ion-beam fabrication on the atomic level. In their article in this issue, Mishra et al. ^{Reference Mishra, Ishikawa, Lupini and Pennycook83} provide an overview of the energy and momentum transfer processes in STEM, and further summarize recent observations of electron-beam-induced single-atom dynamics in STEM. In their article, Zhao et al. ^{Reference Zhao, Kotakoski, Meyer, Sutter, Sutter, Krasheninnikov, Kaiser and Zhou84} report on e-beam-induced modifications in the atomic structure of two-dimensional materials such as graphene and layered chalcogenides, including creation of single vacancies and defect clusters, beam-induced phase transitions, and single-atom motion. In their article, Belianinov et al. ^{Reference Belianinov, Burch, Kim, Tan, Hlawacek and Ovchinnikova85} report on ion-beam-based matter fabrication, advances enabled by rapidly emerging He and other ion microscopies. Finally, Jiang et al. ^{Reference Jiang, Zarkadoula, Narang, Maksov, Kravchenko, Borisevich, Jesse and Kalinin86} report on the applications of electron beams in inducing local structural changes in 3D solids, including amorphization-crystallization and phase transitions. Combined with beam control, this enables fabrication of 3D atomic structures with single-atomic plane precision.

Summary

Feynman famously said, “There’s plenty of room at the bottom.” ^{Reference Feynman15} Modern microelectronics development has allowed scientists to “think” at the “bottom,” or the nanoscale, as exemplified by sub-10-nm semiconductor and emerging quantum-computing technologies. ^{Reference Fuechsle, Miwa, Mahapatra, Ryu, Lee, Warschkow, Hollenberg, Klimeck and Simmons37} However, the realization of the vision of nanoscience—from nanorobotics to destroy cancer cells to reconfigurable electronics—also requires the capability to “act” and “build” on the nanoscale. The two primary paradigms of nanotechnology—scanning probe-based fabrication and chemical synthesis and self-assembly—have been guiding the development of this field for more than two decades. Now, a third paradigm, the “atom forge” ^{Reference Nellist, McCallum and Rodenburg81,Reference Hüe, Rodenburg, Maiden, Sweeney and Midgley82} —a toolbelt of electron-beam-based methods for direct atomic manipulation and atom-by-atom assembly—joins this field.