Showing papers by "Technical University of Berlin published in 2018"

Microscopic Traffic Simulation using SUMO

Abstract: Microscopic traffic simulation is an invaluable tool for traffic research. In recent years, both the scope of research and the capabilities of the tools have been extended considerably. This article presents the latest developments concerning intermodal traffic solutions, simulator coupling and model development and validation on the example of the open source traffic simulator SUMO.

SchNet - A deep learning architecture for molecules and materials.

Abstract: Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Deep One-Class Classification

Abstract: Despite the great advances made by deep learning in many machine learning problems, there is a relative dearth of deep learning approaches for anomaly detection. Those approaches which do exist involve networks trained to perform a task other than anomaly detection, namely generative models or compression, which are in turn adapted for use in anomaly detection; they are not trained on an anomaly detection based objective. In this paper we introduce a new anomaly detection method—Deep Support Vector Data Description—, which is trained on an anomaly detection based objective. The adaptation to the deep regime necessitates that our neural network and training procedure satisfy certain properties, which we demonstrate theoretically. We show the effectiveness of our method on MNIST and CIFAR-10 image benchmark datasets as well as on the detection of adversarial examples of GTSRB stop signs.

Negative emissions-Part 2 : Costs, potentials and side effects

Abstract: The most recent IPCC assessment has shown an important role for negative emissions technologies (NETs) in limiting global warming to 2 °C cost-effectively. However, a bottom-up, systematic, reproducible, and transparent literature assessment of the different options to remove CO2 from the atmosphere is currently missing. In part 1 of this three-part review on NETs, we assemble a comprehensive set of the relevant literature so far published, focusing on seven technologies: bioenergy with carbon capture and storage (BECCS), afforestation and reforestation, direct air carbon capture and storage (DACCS), enhanced weathering, ocean fertilisation, biochar, and soil carbon sequestration. In this part, part 2 of the review, we present estimates of costs, potentials, and side-effects for these technologies, and qualify them with the authors' assessment. Part 3 reviews the innovation and scaling challenges that must be addressed to realise NETs deployment as a viable climate mitigation strategy. Based on a systematic review of the literature, our best estimates for sustainable global NET potentials in 2050 are 0.5–3.6 GtCO₂ yr⁻¹ for afforestation and reforestation, 0.5–5 GtCO₂ yr⁻¹ for BECCS, 0.5–2 GtCO₂ yr⁻¹ for biochar, 2–4 GtCO₂ yr⁻¹ for enhanced weathering, 0.5–5 GtCO₂ yr⁻¹ for DACCS, and up to 5 GtCO2 yr⁻¹ for soil carbon sequestration. Costs vary widely across the technologies, as do their permanency and cumulative potentials beyond 2050. It is unlikely that a single NET will be able to sustainably meet the rates of carbon uptake described in integrated assessment pathways consistent with 1.5 °C of global warming.

A Survey of Physical Layer Security Techniques for 5G Wireless Networks and Challenges Ahead

Abstract: Physical layer security which safeguards data confidentiality based on the information-theoretic approaches has received significant research interest recently. The key idea behind physical layer security is to utilize the intrinsic randomness of the transmission channel to guarantee the security in physical layer. The evolution toward 5G wireless communications poses new challenges for physical layer security research. This paper provides a latest survey of the physical layer security research on various promising 5G technologies, including physical layer security coding, massive multiple-input multiple-output, millimeter wave communications, heterogeneous networks, non-orthogonal multiple access, full duplex technology, and so on. Technical challenges which remain unresolved at the time of writing are summarized and the future trends of physical layer security in 5G and beyond are discussed.

Diacetylene Functionalized Covalent Organic Framework (COF) for Photocatalytic Hydrogen Generation

Abstract: Covalent organic frameworks (COFs) are crystalline, highly porous, two- or three-dimensional polymers with tunable topology and functionalities. Because of their higher chemical stabilities in comparison to their boron-linked counterparts, imine or β-ketoenamine linked COFs have been utilized for a broad range of applications, including gas storage, heterogeneous catalysis, energy storage devices, or proton-conductive membranes. Herein, we report the synthesis of highly porous and chemically stable acetylene (−C≡C−) and diacetylene (−C≡C–C≡C−) functionalized β-ketoenamine COFs, which have been applied as photocatalyst for hydrogen generation from water. It is shown that the diacetylene moieties have a profound effect as the diacetylene-based COF largely outperforms the acetylene-based COF in terms of photocatalytic activity. As a combined effect of high porosity, easily accessible diacetylene (−C≡C–C≡C−) functionalities and considerable chemical stability, an efficient and recyclable heterogeneous photoca...

Efficient Neural Audio Synthesis

Abstract: Sequential models achieve state-of-the-art results in audio, visual and textual domains with respect to both estimating the data distribution and generating high-quality samples. Efficient sampling for this class of models has however remained an elusive problem. With a focus on text-to-speech synthesis, we describe a set of general techniques for reducing sampling time while maintaining high output quality. We first describe a single-layer recurrent neural network, the WaveRNN, with a dual softmax layer that matches the quality of the state-of-the-art WaveNet model. The compact form of the network makes it possible to generate 24kHz 16-bit audio 4x faster than real time on a GPU. Second, we apply a weight pruning technique to reduce the number of weights in the WaveRNN. We find that, for a constant number of parameters, large sparse networks perform better than small dense networks and this relationship holds for sparsity levels beyond 96%. The small number of weights in a Sparse WaveRNN makes it possible to sample high-fidelity audio on a mobile CPU in real time. Finally, we propose a new generation scheme based on subscaling that folds a long sequence into a batch of shorter sequences and allows one to generate multiple samples at once. The Subscale WaveRNN produces 16 samples per step without loss of quality and offers an orthogonal method for increasing sampling efficiency.

Negative emissions—Part 1: Research landscape and synthesis

Abstract: With the Paris Agreement's ambition of limiting climate change to well below 2 °C, negative emission technologies (NETs) have moved into the limelight of discussions in climate science and policy. Despite several assessments, the current knowledge on NETs is still diffuse and incomplete, but also growing fast. Here, we synthesize a comprehensive body of NETs literature, using scientometric tools and performing an in-depth assessment of the quantitative and qualitative evidence therein. We clarify the role of NETs in climate change mitigation scenarios, their ethical implications, as well as the challenges involved in bringing the various NETs to the market and scaling them up in time. There are six major findings arising from our assessment: first, keeping warming below 1.5 °C requires the large-scale deployment of NETs, but this dependency can still be kept to a minimum for the 2 °C warming limit. Second, accounting for economic and biophysical limits, we identify relevant potentials for all NETs except ocean fertilization. Third, any single NET is unlikely to sustainably achieve the large NETs deployment observed in many 1.5 °C and 2 °C mitigation scenarios. Yet, portfolios of multiple NETs, each deployed at modest scales, could be invaluable for reaching the climate goals. Fourth, a substantial gap exists between the upscaling and rapid diffusion of NETs implied in scenarios and progress in actual innovation and deployment. If NETs are required at the scales currently discussed, the resulting urgency of implementation is currently neither reflected in science nor policy. Fifth, NETs face severe barriers to implementation and are only weakly incentivized so far. Finally, we identify distinct ethical discourses relevant for NETs, but highlight the need to root them firmly in the available evidence in order to render such discussions relevant in practice.

The WULCA consensus characterization model for water scarcity footprints: assessing impacts of water consumption based on available water remaining (AWARE)

Abstract: Life cycle assessment (LCA) has been used to assess freshwater-related impacts according to a new water footprint framework formalized in the ISO 14046 standard. To date, no consensus-based approach exists for applying this standard and results are not always comparable when different scarcity or stress indicators are used for characterization of impacts. This paper presents the outcome of a 2-year consensus building process by the Water Use in Life Cycle Assessment (WULCA), a working group of the UNEP-SETAC Life Cycle Initiative, on a water scarcity midpoint method for use in LCA and for water scarcity footprint assessments. In the previous work, the question to be answered was identified and different expert workshops around the world led to three different proposals. After eliminating one proposal showing low relevance for the question to be answered, the remaining two were evaluated against four criteria: stakeholder acceptance, robustness with closed basins, main normative choice, and physical meaning. The recommended method, AWARE, is based on the quantification of the relative available water remaining per area once the demand of humans and aquatic ecosystems has been met, answering the question “What is the potential to deprive another user (human or ecosystem) when consuming water in this area?” The resulting characterization factor (CF) ranges between 0.1 and 100 and can be used to calculate water scarcity footprints as defined in the ISO standard. After 8 years of development on water use impact assessment methods, and 2 years of consensus building, this method represents the state of the art of the current knowledge on how to assess potential impacts from water use in LCA, assessing both human and ecosystem users’ potential deprivation, at the midpoint level, and provides a consensus-based methodology for the calculation of a water scarcity footprint as per ISO 14046.

Fast current-driven domain walls and small skyrmions in a compensated ferrimagnet.

Abstract: Spintronics is a research field that aims to understand and control spins on the nanoscale and should enable next-generation data storage and manipulation. One technological and scientific key challenge is to stabilize small spin textures and to move them efficiently with high velocities. For a long time, research focused on ferromagnetic materials, but ferromagnets show fundamental limits for speed and size. Here, we circumvent these limits using compensated ferrimagnets. Using ferrimagnetic Pt/Gd44Co56/TaOx films with a sizeable Dzyaloshinskii–Moriya interaction, we realize a current-driven domain wall motion with a speed of 1.3 km s–1 near the angular momentum compensation temperature (TA) and room-temperature-stable skyrmions with minimum diameters close to 10 nm near the magnetic compensation temperature (TM). Both the size and dynamics of the ferrimagnet are in excellent agreement with a simplified effective ferromagnet theory. Our work shows that high-speed, high-density spintronics devices based on current-driven spin textures can be realized using materials in which TA and TM are close together. Ferrimagnetic Gd44Co56 near the compensation temperature enables domain wall motion with a speed of 1.3 km s–1 and room temperature skyrmions with diameters close to 10 nm.

Towards exact molecular dynamics simulations with machine-learned force fields.

Abstract: Molecular dynamics (MD) simulations employing classical force fields constitute the cornerstone of contemporary atomistic modeling in chemistry, biology, and materials science. However, the predictive power of these simulations is only as good as the underlying interatomic potential. Classical potentials often fail to faithfully capture key quantum effects in molecules and materials. Here we enable the direct construction of flexible molecular force fields from high-level ab initio calculations by incorporating spatial and temporal physical symmetries into a gradient-domain machine learning (sGDML) model in an automatic data-driven way. The developed sGDML approach faithfully reproduces global force fields at quantum-chemical CCSD(T) level of accuracy and allows converged molecular dynamics simulations with fully quantized electrons and nuclei. We present MD simulations, for flexible molecules with up to a few dozen atoms and provide insights into the dynamical behavior of these molecules. Our approach provides the key missing ingredient for achieving spectroscopic accuracy in molecular simulations.

Towards demand-side solutions for mitigating climate change

Abstract: Research on climate change mitigation tends to focus on supply-side technology solutions A better understanding of demand-side solutions is missing We propose a transdisciplinary approach to identify demand-side climate solutions, investigate their mitigation potential, detail policy measures and assess their implications for well-being

The limits and potentials of deep learning for robotics

Abstract: The application of deep learning in robotics leads to very specific problems and research questions that are typically not addressed by the computer vision and machine learning communities. In this paper we discuss a number of robotics-specific learning, reasoning, and embodiment challenges for deep learning. We explain the need for better evaluation metrics, highlight the importance and unique challenges for deep robotic learning in simulation, and explore the spectrum between purely data-driven and model-driven approaches. We hope this paper provides a motivating overview of important research directions to overcome the current limitations, and helps to fulfill the promising potentials of deep learning in robotics.

Analysis and Observations From the First Amazon Picking Challenge

Abstract: This paper presents an overview of the inaugural Amazon Picking Challenge along with a summary of a survey conducted among the 26 participating teams. The challenge goal was to design an autonomous robot to pick items from a warehouse shelf. This task is currently performed by human workers, and there is hope that robots can someday help increase efficiency and throughput while lowering cost. We report on a 28-question survey posed to the teams to learn about each team’s background, mechanism design, perception apparatus, planning, and control approach. We identify trends in this data, correlate it with each team’s success in the competition, and discuss observations and lessons learned based on survey results and the authors’ personal experiences during the challenge. Note to Practitioners —Perception, motion planning, grasping, and robotic system engineering have reached a level of maturity that makes it possible to explore automating simple warehouse tasks in semistructured environments that involve high-mix, low-volume picking applications. This survey summarizes lessons learned from the first Amazon Picking Challenge, highlighting mechanism design, perception, and motion planning algorithms, as well as software engineering practices that were most successful in solving a simplified order fulfillment task. While the choice of mechanism mostly affects execution speed, the competition demonstrated the systems challenges of robotics and illustrated the importance of combining reactive control with deliberative planning.

Efficient Neural Audio Synthesis

Abstract: Sequential models achieve state-of-the-art results in audio, visual and textual domains with respect to both estimating the data distribution and generating high-quality samples. Efficient sampling for this class of models has however remained an elusive problem. With a focus on text-to-speech synthesis, we describe a set of general techniques for reducing sampling time while maintaining high output quality. We first describe a single-layer recurrent neural network, the WaveRNN, with a dual softmax layer that matches the quality of the state-of-the-art WaveNet model. The compact form of the network makes it possible to generate 24kHz 16-bit audio 4x faster than real time on a GPU. Second, we apply a weight pruning technique to reduce the number of weights in the WaveRNN. We find that, for a constant number of parameters, large sparse networks perform better than small dense networks and this relationship holds for sparsity levels beyond 96%. The small number of weights in a Sparse WaveRNN makes it possible to sample high-fidelity audio on a mobile CPU in real time. Finally, we propose a new generation scheme based on subscaling that folds a long sequence into a batch of shorter sequences and allows one to generate multiple samples at once. The Subscale WaveRNN produces 16 samples per step without loss of quality and offers an orthogonal method for increasing sampling efficiency.

A unique oxygen ligand environment facilitates water oxidation in hole-doped IrNiOx core–shell electrocatalysts

Abstract: The electro-oxidation of water to oxygen is expected to play a major role in the development of future electrochemical energy conversion and storage technologies. However, the slow rate of the oxygen evolution reaction remains a key challenge that requires fundamental understanding to facilitate the design of more active and stable electrocatalysts. Here, we probe the local geometric ligand environment and electronic metal states of oxygen-coordinated iridium centres in nickel-leached IrNi@IrOx metal oxide core–shell nanoparticles under catalytic oxygen evolution conditions using operando X-ray absorption spectroscopy, resonant high-energy X-ray diffraction and differential atomic pair correlation analysis. Nickel leaching during catalyst activation generates lattice vacancies, which in turn produce uniquely shortened Ir–O metal ligand bonds and an unusually large number of d-band holes in the iridium oxide shell. Density functional theory calculations show that this increase in the formal iridium oxidation state drives the formation of holes on the oxygen ligands in direct proximity to lattice vacancies. We argue that their electrophilic character renders these oxygen ligands susceptible to nucleophilic acid–base-type O–O bond formation at reduced kinetic barriers, resulting in strongly enhanced reactivities. The precise understanding of the active phase under reaction conditions at the molecular level is crucial for the design of improved catalysts. Now, Strasser, Jones and colleagues correlate the high activity of IrNi@IrOx core–shell nanoparticles with the amount of lattice vacancies produced by the nickel leaching process that takes place before and during water oxidation, and elucidate the underlying structural-electronic effects.

Making carbon pricing work for citizens

Abstract: The gap between actual carbon prices and those required to achieve ambitious climate change mitigation could be closed by enhancing the public acceptability of carbon pricing through appropriate use of the revenues raised. In this Perspective, we synthesize findings regarding the optimal use of carbon revenues from both traditional economic analyses and studies in behavioural and political science that are focused on public acceptability. We then compare real-world carbon pricing regimes with theoretical insights on distributional fairness, revenue salience, political trust and policy stability. We argue that traditional economic lessons on efficiency and equity are subsidiary to the primary challenge of garnering greater political acceptability and make recommendations for enhancing political support through appropriate revenue uses in different economic and political circumstances. Ambitious carbon pricing reform is needed to meet climate targets. This Perspective argues that effective revenue recycling schemes should prioritize behavioural considerations that are aimed at achieving greater political acceptance.

Unified structural motifs of the catalytically active state of Co(oxyhydr)oxides during the electrochemical oxygen evolution reaction

Abstract: Efficient catalysts for the anodic oxygen evolution reaction (OER) are critical for electrochemical H2 production. Their design requires structural knowledge of their catalytically active sites and state. Here, we track the atomic-scale structural evolution of well-defined CoOx(OH)y compounds into their catalytically active state during electrocatalytic operation through operando and surface-sensitive X-ray spectroscopy and surface voltammetry, supported by theoretical calculations. We find clear voltammetric evidence that electrochemically reducible near-surface Co3+–O sites play an organizing role for high OER activity. These sites invariably emerge independent of initial metal valency and coordination under catalytic OER conditions. Combining experiments and theory reveals the unified chemical structure motif as µ2-OH-bridged Co2+/3+ ion clusters formed on all three-dimensional cross-linked and layered CoOx(OH)y precursors and present in an oxidized form during the OER, as shown by operando X-ray spectroscopy. Together, the spectroscopic and electrochemical fingerprints offer a unified picture of our molecular understanding of the structure of catalytically active metal oxide OER sites. Knowledge of the active sites in catalysts—including the sites that form under working conditions—is vital for future design and development. Here, the authors track the atomic-scale changes in a series of well-defined cobalt-based oxide electrocatalysts, showing that the structurally distinct catalysts develop a similar structural motif as they transform into the catalytically active state.

Dynamic strain determination using fibre-optic cables allows imaging of seismological and structural features

Abstract: Natural hazard prediction and efficient crust exploration require dense seismic observations both in time and space. Seismological techniques provide ground-motion data, whose accuracy depends on sensor characteristics and spatial distribution. Here we demonstrate that dynamic strain determination is possible with conventional fibre-optic cables deployed for telecommunication. Extending recently distributed acoustic sensing (DAS) studies, we present high resolution spatially un-aliased broadband strain data. We recorded seismic signals from natural and man-made sources with 4-m spacing along a 15-km-long fibre-optic cable layout on Reykjanes Peninsula, SW-Iceland. We identify with unprecedented resolution structural features such as normal faults and volcanic dykes in the Reykjanes Oblique Rift, allowing us to infer new dynamic fault processes. Conventional seismometer recordings, acquired simultaneously, validate the spectral amplitude DAS response between 0.1 and 100 Hz bandwidth. We suggest that the networks of fibre-optic telecommunication lines worldwide could be used as seismometers opening a new window for Earth hazard assessment and exploration.

Efficient Electrochemical Hydrogen Peroxide Production from Molecular Oxygen on Nitrogen-Doped Mesoporous Carbon Catalysts

Abstract: Electrochemical hydrogen peroxide (H2O2) production by two-electron oxygen reduction is a promising alternative process to the established industrial anthraquinone process. Current challenges relate to finding cost-effective electrocatalysts with high electrocatalytic activity, stability, and product selectivity. Here, we explore the electrocatalytic activity and selectivity toward H2O2 production of a number of distinct nitrogen-doped mesoporous carbon catalysts and report a previously unachieved H2O2 selectivity of ∼95–98% in acidic solution. To explain our observations, we correlate their structural, compositional, and other physicochemical properties with their electrocatalytic performance and uncover a close correlation between the H2O2 product yield and the surface area and interfacial zeta potential. Nitrogen doping was found to sharply boost H2O2 activity and selectivity. Chronoamperometric H2O2 electrolysis confirms the exceptionally high H2O2 production rate and large H2O2 faradaic selectivity f...

A chemical survey of exoplanets with ARIEL

Abstract: Thousands of exoplanets have now been discovered with a huge range of masses, sizes and orbits: from rocky Earth-like planets to large gas giants grazing the surface of their host star. However, the essential nature of these exoplanets remains largely mysterious: there is no known, discernible pattern linking the presence, size, or orbital parameters of a planet to the nature of its parent star. We have little idea whether the chemistry of a planet is linked to its formation environment, or whether the type of host star drives the physics and chemistry of the planet’s birth, and evolution. ARIEL was conceived to observe a large number (~1000) of transiting planets for statistical understanding, including gas giants, Neptunes, super-Earths and Earth-size planets around a range of host star types using transit spectroscopy in the 1.25–7.8 μm spectral range and multiple narrow-band photometry in the optical. ARIEL will focus on warm and hot planets to take advantage of their well-mixed atmospheres which should show minimal condensation and sequestration of high-Z materials compared to their colder Solar System siblings. Said warm and hot atmospheres are expected to be more representative of the planetary bulk composition. Observations of these warm/hot exoplanets, and in particular of their elemental composition (especially C, O, N, S, Si), will allow the understanding of the early stages of planetary and atmospheric formation during the nebular phase and the following few million years. ARIEL will thus provide a representative picture of the chemical nature of the exoplanets and relate this directly to the type and chemical environment of the host star. ARIEL is designed as a dedicated survey mission for combined-light spectroscopy, capable of observing a large and well-defined planet sample within its 4-year mission lifetime. Transit, eclipse and phase-curve spectroscopy methods, whereby the signal from the star and planet are differentiated using knowledge of the planetary ephemerides, allow us to measure atmospheric signals from the planet at levels of 10–100 part per million (ppm) relative to the star and, given the bright nature of targets, also allows more sophisticated techniques, such as eclipse mapping, to give a deeper insight into the nature of the atmosphere. These types of observations require a stable payload and satellite platform with broad, instantaneous wavelength coverage to detect many molecular species, probe the thermal structure, identify clouds and monitor the stellar activity. The wavelength range proposed covers all the expected major atmospheric gases from e.g. H2O, CO2, CH4 NH3, HCN, H2S through to the more exotic metallic compounds, such as TiO, VO, and condensed species. Simulations of ARIEL performance in conducting exoplanet surveys have been performed – using conservative estimates of mission performance and a full model of all significant noise sources in the measurement – using a list of potential ARIEL targets that incorporates the latest available exoplanet statistics. The conclusion at the end of the Phase A study, is that ARIEL – in line with the stated mission objectives – will be able to observe about 1000 exoplanets depending on the details of the adopted survey strategy, thus confirming the feasibility of the main science objectives.

Active Salt/Silica-Templated 2D Mesoporous FeCo-Nx -Carbon as Bifunctional Oxygen Electrodes for Zinc-Air Batteries.

Abstract: Two types of templates, an active metal salt and silica nanoparticles, are used concurrently to achieve the facile synthesis of hierarchical meso/microporous FeCo-Nx -carbon nanosheets (meso/micro-FeCo-Nx -CN) with highly dispersed metal sites. The resulting meso/micro-FeCo-Nx -CN shows high and reversible oxygen electrocatalytic performances for both ORR and OER, thus having potential for applications in rechargeable Zn-air battery. Our approach creates a new pathway to fabricate 2D meso/microporous structured carbon architectures for bifunctional oxygen electrodes in rechargeable Zn-air battery as well as opens avenues to the scale-up production of rationally designed heteroatom-doped catalytic materials for a broad range of applications.