Le Mans Université

In this paper, we propose a novel neural network model called RNN Encoder-Decoder that consists of two recurrent neural networks (RNN). One RNN encodes a sequence of symbols into a fixed-length vector representation, and the other decodes the representation into another sequence of symbols. The encoder and decoder of the proposed model are jointly trained to maximize the conditional probability of a target sequence given a source sequence. The performance of a statistical machine translation system is empirically found to improve by using the conditional probabilities of phrase pairs computed by the RNN Encoder-Decoder as an additional feature in the existing log-linear model. Qualitatively, we show that the proposed model learns a semantically and syntactically meaningful representation of linguistic phrases.

The Crystallography Open Database (COD), which is a project that aims to gather all available inorganic, metal-organic and small organic molecule structural data in one database, is described. The database adopts an open-access model. The COD currently contains ∼80 000 entries in crystallographic information file format, with nearly full coverage of the International Union of Crystallography publications, and is growing in size and quality.

The large-pore iron(III) carboxylate MIL-100(Fe) with a zeotype architecture has been isolated under hydrothermal conditions, its structure solved from synchrotron X-ray powder diffraction data, while Friedel-Crafts benzylation catalytic tests indicate a high activity and selectivity for MIL-100(Fe).

Methods extracting fast all the peak intensities from a complete powder diffraction pattern are reviewed. The genesis of the modern whole powder pattern decomposition methods (the so-called Pawley and Le Bail methods) is detailed and their importance and domains of application are decoded from the most cited papers citing them. It is concluded that these methods represented a decisive step toward the possibility to solve more easily, if not routinely, a structure solely from a powder sample. The review enlightens the contributions from the Louër’s group during the rising years 1987–1993.

The removal of As(V), one of the most poisonous groundwater pollutants, by synthetic nanoscale zero-valent iron (NZVI) was studied. Batch experiments were performed to investigate the influence of pH, adsorption kinetics, sorption mechanism, and anionic effects. Field emission scanning electron microscopy (FE-SEM), high-resolution transmission electron microscopy (HR-TEM), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and Mossbauer spectroscopy were used to characterize the particle size, surface morphology, and corrosion layer formation on pristine NZVI and As(V)-treated NZVI. The HR-TEM study of pristine NZVI showed a core-shell-like structure, where more than 90% of the nanoparticles were under 30 nm in diameter. Mössbauer spectroscopy further confirmed its structure in which 19% were in zero-valent state with a coat of 81% iron oxides. The XRD results showed that As(V)-treated NZVI was gradually converted into magnetite/maghemite corrosion products over 90 days. The XPS study confirmed that 25% As(V) was reduced to As(III) by NZVI after 90 days. As(V) adsorption kinetics were rapid and occurred within minutes following a pseudo-first-order rate expression with observed reaction rate constants (Kobs) of 0.02-0.71 min(-1) at various NZVI concentrations. Laser light scattering analysis confirmed that NZVI-As(V) forms an inner-sphere surface complexation. The effects of competing anions revealed that HCO3-, H4SiO4(0), and H2PO4(2-) are potential interfering agents in the As(V) adsorption reaction. Our results suggest that NZVI is a suitable candidate for As(V) remediation.

There's more to LiFe: When used as an electrode in a lithium half cell, the metal–organic framework [LixFeIIxFeIII1−x(OH)0.8F0.2L] (L=O2CC6H4CO2) shows a reversible redox process around 3.0 V versus Li+/Li0 with interesting capacity retention and rate capability (see voltage profile for a [LixFe(OH)0.8F0.2L] half cell; inset: representation of lithium insertion upon oxidation to Li+). Supporting information for this article is available on the WWW under http://www.wiley-vch.de/contents/jc_2002/2007/z605163_s.pdf or from the author. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Magnetite particles with an average size of 39 nm and good monodispersity have been synthesized by coprecipitation at 70 °C from ferrous Fe2+ and ferric Fe3+ ions by a (N(CH3)4OH) solution, followed by hydrothermal treatment at 250 °C. The magnetite nanoparticles before the hydrothermal step display an average size of 12 nm and are highly oxidized when they are in contact with air. Complementary microstructural and magnetic characterizations of nanoparticles after hydrothermal treatment show unambiguously that they consist of magnetite with only a slight deviation from stoichiometry (δ ≈ 0.05), leading to Fe2.95O4.

Gas grabber: Reduction of the porous metal–organic framework MIL-100(Fe) (see picture; X−: OH− or F−; red: O of trimesate ligands) to form coordinatively unsaturated sites at iron(II) (depicted as a violet circle) is essential for selective gas purification or the selective removal of unsaturated gas impurities. Detailed facts of importance to specialist readers are published as ”Supporting Information”. Such documents are peer-reviewed, but not copy-edited or typeset. They are made available as submitted by the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Although much effort has recently been devoted to training high-quality sentence embeddings, we still have a poor understanding of what they are capturing. "Downstream" tasks, often based on sentence classification, are commonly used to evaluate the quality of sentence representations. The complexity of the tasks makes it however difficult to infer what kind of information is present in the representations. We introduce here 10 probing tasks designed to capture simple linguistic features of sentences, and we use them to study embeddings generated by three different encoders trained in eight distinct ways, uncovering intriguing properties of both encoders and training methods.

New expressions are given that can be used instead of the phenomenological equations of Delany and Bazley. They provide similar predictions in the range of validity of these equations, and in addition are valid at low frequencies where the equations of Delany and Bazley provide unphysical predictions. These new expressions have been worked out by using the general frequency dependence of the viscous forces in porous materials proposed by Johnson et al. [J. Fluid Mech. 176, 379 (1987)], with a transposition carried out to predict the dynamic bulk modulus of air. The model used suggests how sound propagation in fibrous materials can depend both on the diameter of the fibers and on the density of the material.

Experimental results obtained in different laboratories world-wide by researchers using surface-enhanced Raman scattering (SERS) can differ significantly. We, an international team of scientists with long-standing expertise in SERS, address this issue from our perspective by presenting considerations on reliable and quantitative SERS. The central idea of this joint effort is to highlight key parameters and pitfalls that are often encountered in the literature. To that end, we provide here a series of recommendations on: a) the characterization of solid and colloidal SERS substrates by correlative electron and optical microscopy and spectroscopy, b) on the determination of the SERS enhancement factor (EF), including suitable Raman reporter/probe molecules, and finally on c) good analytical practice. We hope that both newcomers and specialists will benefit from these recommendations to increase the inter-laboratory comparability of experimental SERS results and further establish SERS as an analytical tool.

The results of both a line-broadening study on a ceria sample and a size–strain round robin on diffraction line-broadening methods, which was sponsored by the Commission on Powder Diffraction of the International Union of Crystallography, are presented. The sample was prepared by heating hydrated ceria at 923 K for 45 h. Another ceria sample was prepared to correct for the effects of instrumental broadening by annealing commercially obtained ceria at 1573 K for 3 h and slowly cooling it in the furnace. The diffraction measurements were carried out with two laboratory and two synchrotron X-ray sources, two constant-wavelength neutron and a time-of-flight (TOF) neutron source. Diffraction measurements were analyzed by three methods: the model assuming a lognormal size distribution of spherical crystallites, Warren–Averbach analysis and Rietveld refinement. The last two methods detected a relatively small strain in the sample, as opposed to the first method. Assuming a strain-free sample, the results from all three methods agree well. The average real crystallite size, on the assumption of a spherical crystallite shape, is 191 (5) Å. The scatter of results given by different instruments is relatively small, although significantly larger than the estimated standard uncertainties. The Rietveld refinement results for this ceria sample indicate that the diffraction peaks can be successfully approximated with a pseudo-Voigt function. In a common approximation used in Rietveld refinement programs, this implies that the size-broadened profile cannot be approximated by a Lorentzian but by a full Voigt or pseudo-Voigt function. In the second part of this paper, the results of the round robin on the size–strain line-broadening analysis methods are presented, which was conducted through the participation of 18 groups from 12 countries. Participants have reported results obtained by analyzing data that were collected on the two ceria samples at seven instruments. The analysis of results received in terms of coherently diffracting, both volume-weighted and area-weighted apparent domain size are reported. Although there is a reasonable agreement, the reported results on the volume-weighted domain size show significantly higher scatter than those on the area-weighted domain size. This is most likely due to a significant number of results reporting a high value of strain. Most of those results were obtained by Rietveld refinement in which the Gaussian size parameter was not refined, thus erroneously assigning size-related broadening to other effects. A comparison of results with the average of the three-way comparative analysis from the first part shows a good agreement.

Recent work on end-to-end neural network-based architectures for machine translation has shown promising results for En-Fr and En-De translation. Arguably, one of the major factors behind this success has been the availability of high quality parallel corpora. In this work, we investigate how to leverage abundant monolingual corpora for neural machine translation. Compared to a phrase-based and hierarchical baseline, we obtain up to $1.96$ BLEU improvement on the low-resource language pair Turkish-English, and $1.59$ BLEU on the focused domain task of Chinese-English chat messages. While our method was initially targeted toward such tasks with less parallel data, we show that it also extends to high resource languages such as Cs-En and De-En where we obtain an improvement of $0.39$ and $0.47$ BLEU scores over the neural machine translation baselines, respectively.

Superparamagnetic ferrite nanoparticles (MFe2O4, where M = Fe, Co, Mn) were synthesized through a novel one-step aqueous coprecipitation method based on the use of a new type of alkaline agent: the alkanolamines isopropanolamine and diisopropanolamine. The role played by the bases on the particles’ size, chemical composition, and magnetic properties was investigated and compared directly with the effect of the traditional inorganic base NaOH. The novel MFe2O4 nanomaterials exhibited high colloidal stability, particle sizes in the range of 4–12 nm, and superparamagnetic properties. More remarkably, they presented smaller particle sizes (up to 6 times) and enhanced saturation magnetization (up to 1.3 times) relative to those prepared with NaOH. Furthermore, the nanomaterials exhibited improved magnetic properties when compared with nanoferrites of similar size synthesized by coprecipitation with other bases or by other methods reported in the literature. The alkanolamines were responsible for these achievements by acting both as alkaline agents and as complexing agents that controlled the particle size during the synthesis process and improved the spin rearrangement at the surface (thinner magnetic “dead” layers). These results open new horizons for the design of water-dispersible MFe2O4 nanoparticles with tuned properties through a versatile and easily scalable coprecipitation route.

(MRSA). 23 of these complexes have not been reported for their antimicrobial properties before. This work reveals the vast diversity that metal-containing compounds can bring to antimicrobial research. It is important to raise awareness of these types of compounds for the design of truly novel antibiotics with potential for combatting antimicrobial resistance.

Loïc Barrault, Ondřej Bojar, Marta R. Costa-jussà, Christian Federmann, Mark Fishel, Yvette Graham, Barry Haddow, Matthias Huck, Philipp Koehn, Shervin Malmasi, Christof Monz, Mathias Müller, Santanu Pal, Matt Post, Marcos Zampieri. Proceedings of the Fourth Conference on Machine Translation (Volume 2: Shared Task Papers, Day 1). 2019.

Perfect, broadband and asymmetric sound absorption is theoretically, numerically and experimentally reported by using subwavelength thickness panels in a transmission problem. The panels are composed of a periodic array of varying crosssection waveguides, each of them being loaded by Helmholtz resonators (HRs) with graded dimensions. The low cut-off frequency of the absorption band is fixed by the resonance frequency of the deepest HR, that reduces drastically the transmission. The preceding HR is designed with a slightly higher resonance frequency with a geometry that allows the impedance matching to the surrounding medium. Therefore, reflection vanishes and the structure is critically coupled. This results in perfect sound absorption at a single frequency. We report perfect absorption at 300 Hz for a structure whose thickness is 40 times smaller than the wavelength. Moreover, this process is repeated by adding HRs to the waveguide, each of them with a higher resonance frequency than the preceding one. Using this frequency cascade effect, we report quasi-perfect sound absorption over almost two frequency octaves ranging from 300 to 1000 Hz for a panel composed of 9 resonators with a total thickness of 11 cm, i.e., 10 times smaller than the wavelength at 300 Hz.

The orogenic architecture of the world's largest ultrahigh‐pressure exposure, the Hong'an‐Dabie Mountains of the Triassic Qinling‐Dabie orogenic belt, is dominated by Cretaceous and Cenozoic structures that contributed to its exhumation from ≤30 km depth. Cretaceous magmatic crustal recycling (≥50% for the entire Dabie) and heating (>250° to >700°C) were most prominent in Dabie, and exhumation, magmatism, and cooling were all controlled by Cretaceous transtension. Exhumation was accomplished principally by an asymmetric Cordilleran‐type extensional complex in the northern Dabie (Northern Orthogneiss unit) between 140 and 120 Ma, at rates as fast as 2 mm/yr and average horizontal stretching rates of up to 6 mm/yr. Cretaceous reactivation occurred within a regional transtensional strain field as a result of far‐field collisions and Pacific subduction. The onset of crustal extension was preceded and possibly facilitated by a reheating of the Hong'an‐Dabie crust (∼140 Ma) coeval with the onset of voluminous magmatism in eastern China (∼145 Ma), which resulted from a change in Pacific subduction from highly oblique to orthogonal. The Tan‐Lu continental‐scale fault was a normal fault zone in the mid‐Cretaceous (∼110‐90 Ma) and underwent ≥5.4 km dip slip and ≥4 km throw in the Cenozoic. During the India‐Asia collision the Qinling‐Dabie belt acted as the structural discontinuity between the strike‐slip‐dominated escape tectonics south of the Qilian‐Qinling‐Dabie belt and the rifting‐dominated tectonism north of it. The most prominent Cretaceous and Cenozoic structures of the Hong'an‐Dabie, the Xiaotian‐Mozitang and the Jinzhai fault zones, respectively, reactivated major lithospheric structures of the Triassic orogen, i.e., the Huwan detachment zone and the suture.

The Acoustic Black Hole (ABH) is a technique for passive vibration control that was recently developed within the Structural Dynamics and Vibroacoustics communities. From a general perspective, the ABH effect is achieved by embedding a local inhomogeneity in a thin-walled structure, typically a beam or a plate. This inhomogeneity is characterized by a variation of the geometric properties (although material variations are also possible) according to a spatial power law profile. The combination of a local stiffness reduction, due to the power law variation of the wall thickness, and of a local increase in damping, provided by the concurrent application of viscoelastic layers, gives rise to a significant reduction of the wave speed and to a remarkable enhancement of the attenuation properties. As an elastic wave travels within an ABH, its speed experiences a smooth and continuous decrease. In the ideal case, that is when the wall thickness vanishes at the ABH center, the wave speed decreases to zero. In the non-ideal case, that is when the ABH has a non-zero residual thickness at its center, the wave speed still decreases smoothly but it never vanishes. In this latter case, which is of great importance for practical applications, the ABH is typically combined with lossy media (e.g. viscoelastic layers) in order to achieve significantly enhanced structural loss factors. If the speed of an incoming wave can vanish inside the ABH, it follows that this object behaves as a wave trap that extracts elastic energy from the host medium without, in principle, ever releasing it. Several characteristic properties are generally observed in structures with embedded ABHs: significant reduction in vibration and acoustic radiation levels, low reflection coefficient at the ABH location, localized vibration and trapped modes, and existence of cut-on frequencies. Contrarily to passive vibration methods based on viscoelastic materials, the ABH was developed and applied to reduce vibrations and structure-radiated noise without increasing the total mass of the system. More recently, applications to other areas including elastic metastructures, energy harvesting, vibro-impact systems, and cochlear systems were also investigated. This review is intended to provide a comprehensive summary of the state-of-the-art of ABH technology, spanning from theoretical and numerical contributions to practical applications.

Nanostructured ZnFe2O4 ferrites with different grain sizes were prepared by high energy ball milling for various milling times. Both the average grain size and the root mean square strain were estimated from the x-ray diffraction line broadening. The lattice parameter initially decreases slightly with milling and it increases with further milling. The magnetization is found to increase as the grain size decreases and its large value is attributed to the cation inversion associated with grain size reduction. The Fe-57 Mossbauer spectra were recorded at 300 K and 77 K for the samples with grain sizes of 22 and 11 nm. There is no evidence for the presence of the Fe2+ charge state. At 77 K the Mossbauer spectra consist of a magnetically ordered component along with a doublet due to the superparamagnetic behaviour of small crystalline grains with the superparamagnetic component decreasing with grain size reduction. At 4.2 K the sample with 11 nm grain size displays a magnetically blocked state as revealed by the Mossbauer spectrum. The Mossbauer spectrum of this sample recorded at 10 K in an external magnetic field of 6 T applied parallel to the direction of gamma rays clearly shows ferrimagnetic ordering of the sample. Also, the sample exhibits spin canting with a large canting angle, maybe due to a spin-glass-like surface layer or grain boundary anisotropies in the material.