Roadmap on emerging hardware and technology for machine learning.

Recent progress in artificial intelligence is largely attributed to the rapid development of machine learning, especially in the algorithm and neural network models. However, it is the performance of the hardware, in particular the energy efficiency of a computing system that sets the fundamental limit of the capability of machine learning. Data-centric computing requires a revolution in hardware systems, since traditional digital computers based on transistors and the von Neumann architecture were not purposely designed for neuromorphic computing. A hardware platform based on emerging devices and new architecture is the hope for future computing with dramatically improved throughput and energy efficiency. Building such a system, nevertheless, faces a number of challenges, ranging from materials selection, device optimization, circuit fabrication and system integration, to name a few. The aim of this Roadmap is to present a snapshot of emerging hardware technologies that are potentially beneficial for machine learning, providing the Nanotechnology readers with a perspective of challenges and opportunities in this burgeoning field.

Multimodal Characterization of the Morphology and Functional Interfaces in Composite Electrodes for Li-S Batteries by Li Ion and Electron Beams.

We report the characterization of multiscale 3D structural architectures of novel poly[sulfur-random-(1,3-diisopropenylbenzene)] copolymer-based cathodes for high-energy-density Li-S batteries capable of realizing discharge capacities >1000 mAh/g and long cycling lifetimes >500 cycles. Hierarchical morphologies and interfacial structures have been investigated by a combination of focused Li ion beam (LiFIB) and analytical electron microscopy in relation to the electrochemical performance and physicomechanical stability of the cathodes. Charge-free surface topography and composition-sensitive imaging of the electrodes was performed using recently introduced low-energy scanning LiFIB with Li+ probe sizes of a few tens of nanometers at 5 keV energy and 1 pA probe current. Furthermore, we demonstrate that LiFIB has the ability to inject a certain number of Li cations into the material with nanoscale precision, potentially enabling control of the state of discharge in the selected area. We show that chemical modification of the cathodes by replacing the elemental sulfur with organosulfur copolymers significantly improves its structural integrity and compositional homogeneity down to the sub-5-nm length scale, resulting in the creation of (a) robust functional interfaces and percolated conductive pathways involving graphitic-like outer shells of aggregated nanocarbons and (b) extended micro- and mesoscale porosities required for effective ion transport.

Comparison of Nanoscale Focused Ion Beam and Electrochemical Lithiation in ß-Sn Microspheres.

The development of Li focused ion beams (Li-FIB) enables controlled Li ion insertion into materials with nanoscale resolution. We take the first step toward establishing the relevance of the Li-FIB for studies of ion dynamics in electrochemically active materials by comparing FIB lithiation with conventional electrochemical lithiation of isolated ß-Sn microspheres. Samples are characterized by cross-sectioning with Ga FIB and imaging via electron microscopy. The Li-FIB and electrochemical lithiated Sn exhibit similarities that suggest that the Li-FIB can be a powerful tool for exploring dynamical Li ion-material interactions at the nanoscale in a range of battery materials.

Temperature Distribution in TaO x Resistive Switching Devices Assessed In Operando by Scanning Thermal Microscopy.

Understanding the physical changes during electroformation and switching processes in transition-metal-oxide-based non-volatile memory devices is important for advancing this technology. Relatively few characteristics of these devices have been assessed in operando. In this work, we present scanning thermal microscopy measurements in vacuum on TaO x -based memory devices electroformed in both positive and negative polarities and high- and low-resistance states. The observed surface temperature footprints of the filament showed higher peak temperatures and narrower temperature distributions when the top electrode served as the anode in the electroformation process. This is consistent with a model in which a hot spot is created by a gap in the conducting filament that forms closest to the anode. A similar behavior was seen on comparing the high-resistance state to the low-resistance state, with the low-resistance footprint showing a lower peak and a larger width, consistent with the gap disappearing when the device is switched from high resistance to low resistance.

Theoretical model of errors in micromirror-based three-dimensional particle tracking.

Several recently developed particle-tracking and imaging methods have achieved three-dimensional sensitivity through the introduction of angled micromirrors into the observation volume of an optical microscope. We model the imaging response of such devices and show how the direct and reflected images of a fluorescent particle are affected. In particle-tracking applications, asymmetric image degradation manifests itself as systematic tracking errors. Based on our results, we identify strategies for reducing systematic errors to the 10nm level in practical applications.

Direct-Write Lithiation of Silicon Using a Focused Ion Beam of Li.

Electrochemical processes that govern the performance of lithium ion batteries involve numerous parallel reactions and interfacial phenomena that complicate the microscopic understanding of these systems. To study the behavior of ion transport and reaction in these applications, we report the use of a focused ion beam of Li+ to locally insert controlled quantities of lithium with high spatial resolution into electrochemically relevant materials in vacuo. To benchmark the technique, we present results on direct-write lithiation of 35 nm thick crystalline silicon membranes using a 2 keV beam of Li+ at doses up to 1018 cm-2 (104 nm-2). We confirm quantitative sub-µm control of lithium insertion and characterize the concomitant morphological, structural, and functional changes of the system using a combination of electron and scanning probe microscopy. We observe saturation of interstitial lithium in the silicon membrane at ≈10% dopant number density and spillover of excess lithium onto the membrane's surface. The implanted Li+ is demonstrated to remain electrochemically active. This technique will enable controlled studies and improve understanding of Li+ ion interaction with local defect structures and interfaces in electrode and solid-electrolyte materials.

Streaming Batch Eigenupdates for Hardware Neural Networks.

Neural networks based on nanodevices, such as metal oxide memristors, phase change memories, and flash memory cells, have generated considerable interest for their increased energy efficiency and density in comparison to graphics processing units (GPUs) and central processing units (CPUs). Though immense acceleration of the training process can be achieved by leveraging the fact that the time complexity of training does not scale with the network size, it is limited by the space complexity of stochastic gradient descent, which grows quadratically. The main objective of this work is to reduce this space complexity by using low-rank approximations of stochastic gradient descent. This low spatial complexity combined with streaming methods allows for significant reductions in memory and compute overhead, opening the door for improvements in area, time and energy efficiency of training. We refer to this algorithm and architecture to implement it as the streaming batch eigenupdate (SBE) approach.

Spontaneous current constriction in threshold switching devices.

Threshold switching devices are of increasing importance for a number of applications including solid-state memories and neuromorphic circuits. Their non-linear characteristics are thought to be associated with a spontaneous (occurring without an apparent external stimulus) current flow constriction but the extent and the underlying mechanism are a subject of debate. Here we use Scanning Joule Expansion Microscopy to demonstrate that, in functional layers with thermally activated electrical conductivity, the current spontaneously and gradually constricts when a device is biased into the negative differential resistance region. We also show that the S-type negative differential resistance I-V characteristics are only a subset of possible solutions and it is possible to have multiple current density distributions corresponding to the same value of the device voltage. In materials with steep dependence of current on temperature the current constriction can occur in nanoscale devices, making this effect relevant for computing applications.

Fast, bias-free algorithm for tracking single particles with variable size and shape.

We introduce a fast and robust technique for single-particle tracking with nanometer accuracy. We extract the center-of-mass of the image of a single particle with a simple, iterative algorithm that efficiently suppresses background-induced bias in a simplistic centroid estimator. Unlike many commonly used algorithms, our position estimator requires no prior information about the shape or size of the tracked particle image and uses only simple arithmetic operations, making it appropriate for future hardware implementation and real-time feedback applications. We demonstrate it both numerically and experimentally, using an inexpensive CCD camera to localize 190 nm fluorescent microspheres to better than 5 nm.

Publisher Correction: Stateful characterization of resistive switching TiO2 with electron beam induced currents.

The original version of this Article contained an error in Eq. 1. The arrows between the symbols "T" and "B", and "B" and "T", were written "↔" but should have been "→", and incorrectly read: IEBIC=IEBAC+ISEE+I(e↔h)+IEBICT↔B+IESEEB↔T The correct from of the Eq. 1 is as follows:IEBIC=IEBAC+ISEE+I(e↔h)+IEBICT→B+IESEEB→T This has now been corrected in both the PDF and HTML versions of the article.

Two-dimensional imaging and modification of nanophotonic resonator modes using a focused ion beam.

High-resolution imaging of optical resonator modes is a key step in the development and characterization of nanophotonic devices. Many sub-wavelength mode-imaging techniques have been developed using optical and electron beam excitation-each with its own limitations in spectral and spatial resolution. Here, we report a 2D imaging technique using a pulsed, low-energy focused ion beam of Li+ to probe the near-surface fields inside photonic resonators. The ion beam locally modifies the resonator structure, causing temporally varying spectroscopic shifts of the resonator. We demonstrate this imaging technique on several optical modes of silicon microdisk resonators by rastering the ion beam across the disk surface and extracting the maximum mode shift at the location of each ion pulse. A small shift caused by ion beam heating is also observed and is independently extracted to directly measure the thermal response of the device. This technique enables visualization of the splitting of degenerate modes into spatially-resolved standing waves and permits persistent optical mode editing. Ion beam probing enables minimally perturbative, in operando imaging of nanophotonic devices with high resolution and speed.

Stateful characterization of resistive switching TiO2 with electron beam induced currents.

Metal oxide resistive switches are increasingly important as possible artificial synapses in next-generation neuromorphic networks. Nevertheless, there is still no codified set of tools for studying properties of the devices. To this end, we demonstrate electron beam-induced current measurements as a powerful method to monitor the development of local resistive switching in TiO2-based devices. By comparing beam energy-dependent electron beam-induced currents with Monte Carlo simulations of the energy absorption in different device layers, it is possible to deconstruct the origins of filament image formation and relate this to both morphological changes and the state of the switch. By clarifying the contrast mechanisms in electron beam-induced current microscopy, it is possible to gain new insights into the scaling of the resistive switching phenomenon and observe the formation of a current leakage region around the switching filament. Additionally, analysis of symmetric device structures reveals propagating polarization domains.

Imaging Nanophotonic Modes of Microresonators using a Focused Ion Beam.

Optical microresonators have proven powerful in a wide range of applications, including cavity quantum electrodynamics1-3, biosensing4, microfludics5, and cavity optomechanics6-8. Their performance depends critically on the exact distribution of optical energy, confined and shaped by the nanoscale device geometry. Near-field optical probes9 can image this distribution, but the physical probe necessarily perturbs the near field, which is particularly problematic for sensitive high quality factor resonances10,11. We present a new approach to mapping nanophotonic modes that uses a controllably small and local optomechanical perturbation introduced by a focused lithium ion beam12. An ion beam (radius ≈50 nm) induces a picometer-scale dynamic deformation of the resonator surface, which we detect through a shift in the optical resonance wavelength. We map five modes of a silicon microdisk resonator (Q≥20,000) with both high spatial and spectral resolution. Our technique also enables in-situ observation of ion implantation damage and relaxation dynamics in a silicon lattice13,14.

Accuracy of Nanoscale Pitch Standards Fabricated by Laser-Focused Atomic Deposition.

The pitch accuracy of a grating formed by laser-focused atomic deposition is evaluated from the point of view of fabricating nanoscale pitch standard artifacts. The average pitch obtained by the process, nominally half the laser wavelength, is simply traceable with small uncertainty to an atomic frequency and hence can be known with very high accuracy. An error budget is presented for a Cr on sapphire sample, showing that a combined standard uncertainty of 0.0049 nm, or a relative uncertainty of 2.3 × 10(-5), is readily obtained, provided the substrate temperature does not change. Precision measurements of the diffraction of the 351.1 nm argon ion laser line from such an artifact are also presented. These yield an average pitch of (212.7777 ± 0.0069) nm, which agrees well with the expected value, as corrected for thermal contraction, of (212.7705 ± 0.0049) nm.

Scanning ion microscopy with low energy lithium ions.

Using an ion source based on photoionization of laser-cooled lithium atoms, we have developed a scanning ion microscope with probe sizes of a few tens of nanometers and beam energies from 500eV to 5keV. These beam energies are much lower than the typical operating energies of the helium ion microscope or gallium focused ion beam systems. We demonstrate how low energy can be advantageous in ion microscopy when detecting backscattered ions, due to a decreased interaction volume and the potential for surface sensitive composition analysis. As an example application that demonstrates these advantages, we non-destructively image the removal of a thin residual resist layer during plasma etching in a nano-imprint lithography process.

Electron vortex beams with high quanta of orbital angular momentum.

Electron beams with helical wavefronts carrying orbital angular momentum are expected to provide new capabilities for electron microscopy and other applications. We used nanofabricated diffraction holograms in an electron microscope to produce multiple electron vortex beams with well-defined topological charge. Beams carrying quantized amounts of orbital angular momentum (up to 100h) per electron were observed. We describe how the electrons can exhibit such orbital motion in free space in the absence of any confining potential or external field, and discuss how these beams can be applied to improved electron microscopy of magnetic and biological specimens.

3D particle trajectories observed by orthogonal tracking microscopy.

We demonstrate high-resolution, high-speed 3D nanoparticle tracking using angled micromirrors. When angled micromirrors are introduced into the field of view of an optical microscope, reflected side-on views of a diffusing nanoparticle are projected alongside the usual direct image. The experimental design allows us to find the 3D particle trajectory using fast, centroid-based image processing, with no nonlinear computing operations. We have tracked polystyrene particles of 190 nm diameter with position measurement precision <20 nm in 3D with 3 ms frame duration (i.e., at an imaging rate >330 frames per second). Because the image processing requires only approximately 1 ms per frame, this technique could enable real-time feedback-controlled nanoparticle assembly applications with nanometer precision.

Magneto-optical-trap-based, high brightness ion source for use as a nanoscale probe.

We report on the demonstration of a low emittance, high brightness ion source based on magneto-optically trapped neutral atoms. Our source has ion optical properties comparable to or better than those of the commonly used liquid metal ion source. In addition, it has several advantages that offer new possibilities, including high resolution ion microscopy with ion species tailored for specific applications, contamination-free ion milling, and nanoscale implantation of a variety of elements, either in large quantities, or one at a time, deterministically. Using laser-cooled Cr atoms, we create an ion beam with a normalized rms (root-mean-square) emittance of 6.0 x 10 (-7) mm mrad M e V and approximately 0.25 pA of current, yielding a brightness as high as 2.25 A cm (-2) sr (-1) eV (-1). These values of emittance and brightness show that, with suitable ion optics, an ion beam with a useful amount of current can be produced and focused to spot sizes of less than 1 nm.

Narrow-line magneto-optical cooling and trapping of strongly magnetic atoms.

Laser cooling on weak transitions is a useful technique for reaching ultracold temperatures in atoms with multiple valence electrons. However, for strongly magnetic atoms a conventional narrow-line magneto-optical trap (MOT) is destabilized by competition between optical and magnetic forces. We overcome this difficulty in Er by developing an unusual narrow-line MOT that balances optical and magnetic forces using laser light tuned to the blue side of a narrow (8 kHz) transition. The trap population is spin polarized with temperatures reaching below 2 muK. Our results constitute an alternative method for laser cooling on weak transitions, applicable to rare-earth-metal and metastable alkaline earth elements.