Fine-grained Phase Research Articles

Push the Limit of Highly Accurate Ranging on Commercial UWB Devices

Ranging plays a crucial role in many wireless sensing applications. Among the wireless techniques employed for ranging, Ultra-Wideband (UWB) has received much attention due to its excellent performance and widespread integration into consumer-level electronics. However, the ranging accuracy of the current UWB systems is limited to the centimeter level due to bandwidth limitation, hindering their use for applications that require a very high resolution. This paper proposes a novel system that achieves sub-millimeter-level ranging accuracy on commercial UWB devices for the first time. Our approach leverages the fine-grained phase information of commercial UWB devices. To eliminate the phase drift, we design a fine-grained phase recovery method by utilizing the bi-directional messages in UWB two-way ranging. We further present a dual-frequency switching method to resolve phase ambiguity. Building upon this, we design and implement the ranging system on commercial UWB modules. Extensive experiments demonstrate that our system achieves a median ranging error of just 0.77 mm, reducing the error by 96.54% compared to the state-of-the-art method. We also present three real-life applications to showcase the fine-grained sensing capabilities of our system, including i) smart speaker control, ii) free-style user handwriting, and iii) 3D tracking for virtual-reality (VR) controllers.

Intersection control based on Fine-grained Phases for connected and automated vehicles

Traditional traffic signal control is one of the most widely used urban traffic control methods, due to its high efficiency, robustness, and simplicity. However, in the era of connected and automated vehicles (CAVs), it is prospective to enhance traffic control precision and achieve vehicle coordination in a more efficient manner. Focusing on an isolated intersection, this study proposes an innovative concept of Fine-grained Phase (FP) for CAV-enabled traffic control, which realizes more refined right-of-way allocation than traditional signal phases while preserving the simplicity and scalability. The development of FP jointly optimizes traffic control strategy and intersection layout design to fully utilize time-space resources, and multiple forms of FP could be generated in response to various traffic demands; meanwhile, vehicle coordination is achieved in an offline–online manner to balance solution quality and efficiency. Under the FP-based control, each lane is assigned with periodical entry time points using FP design results, then vehicles are supposed to catch up with the appropriate time points to fulfill their trips. Theoretical analyses are conducted under stationary vehicle arrival patterns, where the average delay and the intersection capacity associated with different forms of FP can be analytically acquired. For the implementation of FP-based control, we construct a general offline–online framework, consisting of an offline model that optimizes the geometry of the conflict zone at intersections and the control strategy of vehicles simultaneously, which is formulated as a mixed-integer linear programming problem, and an online FP adoption mechanism that can adapt to varying demand patterns. Under the framework, limited online computation is required, facilitating its practical real-time implementation. Simulation studies under a variety of demand settings show that FP-based control is advantageous over benchmark methods in reducing vehicle delay and also improving intersection capacity in most cases; even for cases with extremely imbalanced demands, FP-based control can admit comparable throughput, while the average delay is much smaller. In addition, results of computation time reveal that FP-based control could realize real-time implementation with negligible computation burden.

Evolution of pegmatite recorded by zoned garnet from the No. 9 dike in the Jiajika Li polymetallic deposit, eastern Tibetan plateau

Granitic pegmatites can be extremely rich in Li, an important metal for modern industries. Being the Asia’s largest hard-rock Li polymetallic deposit, the Jiajika Li deposit is associated with numerous pegmatite dikes in the eastern Tibetan Plateau. The major Li resources in Jiajika are mainly hosted in fine-grained phases of pegmatite dikes. In this study, samples were collected from the typically zoned No. 9 dike, which inwardly consists of fine-grained albite phase, medium-grained albite phase, coarse-grained albite-microline phase, massive microline phase and massive quartz phase. Garnet is one of the common accessory minerals in the fine-grained albite phase. We report the first in-situ LA-ICP-MS U-Pb age of garnet from pegmatite dikes, yielding an age of 216.7 ± 5.9 Ma, which is consistent with the age of the associated granitic pluton. Combined with published isotopic dating results in Jiajika, two age groups can be identified: ∼210 to ∼ 220 Ma and ∼ 190 to ∼ 200 Ma, which are suspected to be resulted from two stages of thermal events, probably separate pulses of pegmatitic melts. The late pulse of melts might be responsible for the weak deformation of garnet grains and Fe-enriched rims along the deformed boundaries. The early pulse of melts derived from the Majingzi-Jiajijiami granitic pluton shows an indistinct time path from granite, to fine-grained phases, to medium- to coarse-grained phases, which resulted in the trace elemental core-rim textures within individual garnet grains. The garnet cores have typically higher trace element concentrations than the rims, and the boundaries between cores and rims are usually sharp and sinuous, which are proposed to record the amalgamation, dissolution and reprecipitation processes. The growth of cores was not in equilibrium with other phases by strong undercooling, and thus inherited the initial Li signature of the incipient pegmatitic melts, which can be up to 373 ppm. Accordingly, it implies that Li and other incompatible element contents of the incipient pegmatitic melts in Jiajiaka could be enriched in a significant amount at the early stage, which may, to some extent, explain why major Li resources are hosted in very-fine to fine-grained phases in Jiajika.

Experimental free-space continuous-variable quantum key distribution with thermal source.

Passive-state-preparation (PSP) continuous-variable quantum key distribution (CVQKD) protocol explores the intrinsic field fluctuations of a thermal source. Compared with traditional Gaussian-modulated coherent-state CVQKD, it does not need active modulations and has promising applications in chip integration and portable free-space quantum key distribution. In this Letter, we propose and experimentally realize a PSP CVQKD scheme with transmitted local oscillator (LO) through fluctuating transmittance free-space channel using an off-the-shelf amplified spontaneous emission source for the first time. By proposing thermal-state polarization multiplexing transmitted LO, synchronized channel transmittance monitoring and fine-grained phase compensation techniques, secure keys within -15 dB transmittance of simulated free-space channel with turbulence are generated, with a final average secure key rate of 1.015 Mbps asymptotically. Equivalent atmospheric turbulence model analysis shows that the free-space PSP CVQKD scheme provides a promising outlook for high-speed and chip-based CVQKD for kilometer-level atmospheric channel networks.

Intersection Control based on Fine-grained Phases for Connected and Automated Vehicles

Intersection Control based on Fine-grained Phases for Connected and Automated Vehicles

Coarse-to-Fine Output Predictions for Efficient Decoding in Neural Machine Translation

Neural Machine Translation (NMT) systems are undesirably slow as the decoder often has to compute probability distributions over large target vocabularies. In this work, we propose a coarse-to-fine approach to reduce the complexity of the decoding process, using only the information of the weight matrix in the Softmax layer. The large target vocabulary is first trimmed to a small candidate set in the coarse-grained phase, and from this candidate set the final top- k results are generated in the fine-grained phase. Tested on an RNN-based NMT system and a Transformer-based NMT system separately, our GPU-friendly method achieved a significant speed-up without harming the translation quality.

A 10 M Ω, 50 kHz-40 MHz Impedance Measurement Architecture for Source-Differential Flow Cytometry.

A low-power, impedance-based integrated circuit (IC) readout architecture is presented for cell analysis and cytometry applications. A three-electrode layout and source-differential excitation cancels baseline current prior to the sensor front-end, which enables the use of a high-gain readout circuit for the difference current. A lock-in architecture is employed with down-conversion and up-conversion in the feedback loop, enabling high closed-loop gain (up to 10 M Ω) and high bandwidth (up to 40 MHz). A hybrid-RC feedback network mitigates the SNR degradation seen over a wide operating frequency range when using purely capacitive feedback. The effect of phase shift on the closed-loop system gain and noise performance are analyzed in detail, along with optimization strategies, and the design includes fine-grained phase adjustment to minimize phase error. The impedance sensor was fabricated in a 0.18 μ m CMOS process and consumes 9.7 mW with an operating frequency from 50 kHz to 40 MHz and provides adjustable bandwidth. Measurements demonstrate that the impedance sensor achieves 6 pA [Formula: see text] input-referred noise over 200 Hz bandwidth at 0.5 MHz modulation frequency. Combined with a microfluidic flow cell, measured results using this source-differential measurement approach are presented using both monodisperse and polydisperse sample solutions and demonstrate single-cell resolution, detecting 3 μ m diameter particles in solution with 22 dB SNR.

Spatio-Temporal Causal Transformer for Multi-Grained Surgical Phase Recognition.

Automatic surgical phase recognition plays a key role in surgical workflow analysis and overall optimization in clinical work. In the complicated surgical procedures, similar inter-class appearance and drastic variability in phase duration make this still a challenging task. In this paper, a spatio-temporal transformer is proposed for online surgical phase recognition with different granularity. To extract rich spatial information, a spatial transformer is used to model global spatial dependencies of each time index. To overcome the variability in phase duration, a temporal transformer captures the multi-scale temporal context of different time indexes with a dual pyramid pattern. Our method is thoroughly validated on the public Cholec80 dataset with 7 coarse-grained phases and the CATARACTS2020 dataset with 19 fine-grained phases, outperforming state-of-the-art approaches with 91.4% and 84.2% accuracy, taking only 24.5M parameters.

How fluid infiltrates dry crustal rocks during progressive eclogitization and shear zone formation: insights from H2O contents in nominally anhydrous minerals

Granulites from Holsnøy (Bergen Arcs, Norway) maintained a metastable state until fluid infiltration triggered the kinetically delayed eclogitization. Interconnected hydrous eclogite-facies shear zones are surrounded by unreacted granulites. Macroscopically, the granulite–eclogite interface is sharp and there are no significant compositional changes in the bulk chemistry, indicating the fluid composition was quickly rock buffered. To better understand the link between deformation, fluid influx, and fluid–rock interaction one cm-wide shear zone at incipient eclogitization is studied here. Granulite and eclogite consist of garnet, pyroxene, and plagioclase. These nominally anhydrous minerals (NAMs) can incorporate H2O in the form of OH groups. H2O contents increase from granulite to eclogite, as documented in garnet from ~ 10 to ~ 50 µg/g H2O, pyroxene from ~ 50 to ~ 310 µg/g H2O, and granulitic plagioclase from ~ 10 to ~ 140 µg/g H2O. Bowl-shape profiles are characteristic for garnet and pyroxene with lower H2O contents in grain cores and higher at the rims, which suggest a prograde water influx into the NAMs. Omphacite displays a H2O content range from ~ 150 to 425 µg/g depending on the amount of hydrous phases surrounding the grain. The granulitic plagioclase first separates into a hydrous, more albite-rich plagioclase and isolated clinozoisite before being replaced by new fine-grained phases like clinozoisite, kyanite and quartz during ongoing fluid infiltration. Results indicate a twofold fluid influx with different mechanisms to act simultaneously at different scales and rates. Fast and more pervasive proton diffusion is recorded by NAMs that retain the major element composition of the granulite-facies equilibration where hydrogen decorates pre-existing defects in the crystal lattice and leads to OH increase. Contemporaneously, slower grain boundary-assisted aqueous fluid influx enables element transfer and results in progressive formation of new minerals, e.g., hydrous phases. Both mechanisms lead to bulk H2O increase from ~ 450 to ~ 2500 µg/g H2O towards the shear zone and convert the system from rigid to weak. The incorporation of OH groups reduces the activation energy for creep, promotes formation of smaller grain sizes (phase separation of plagioclase), and synkinematic metamorphic mineral reactions. These processes are part of the transient weakening, which enhance the sensitivity of the rock to deform.

Episodic emplacement of the Lingshan Granitic Complex and related two-stage molybdenum mineralization in the Dabie orogenic belt

The Lingshan Granitic Complex consists of a granitic batholith and its proximal Mushan and Xiaofan porphyry stocks, and hosts stockwork-type Mo mineralization within the porphyry stocks and the fine-grained monzogranite, and several vein-type Mo occurrences in the coarse-grained porphyritic monzogranite. All these rocks show similar values of the whole-rock initial 87Sr/86Sr (0.70442–0.70987), εNd(t) (–17.2 to –11.0), and zircon εHf(t) (–27.2 to –9.3), indicating their derivation from partial melting of the Paleoproterozoic-Neoarchean recycled continental crust induced by basaltic underplating at the crust-mantle boundary. The zircon U-Pb and molybdenite Re-Os geochronological results indicate that the Lingshan Granitic Complex was formed by at least two pulses of magmatism at 140–137 Ma and 128–125 Ma, accompanied by two periods of Mo mineralization, which is the epitome of the peak granitic magmatism and related Mo mineralization in the whole Dabie orogenic belt. Combined with previous data, it can be concluded that the discontinuous and episodic nature of magma emplacement in the whole Dabie Mo metallogenic belt has significant consequences for large-scale, multi-staged Mo mineralization. Although Mo mineralization can occur in various lithofacies of the Lingshan Granitic Complex, the industrial Mo enrichment mainly lies within the small shallow emplacement granite porphyry stocks and subordinately within the fine-grained phase of the pluton. These features indicate that the Mo industrial enrichment is not only related to the highly evolved magma, but also more closely related to mechanism of magma emplacement. Rapid decompression induced by rapid ascent of the Mo-bearing granite porphyry stocks, and rapid crystallization of magma induced by quickly cooling due to conductive heat loss to the sidewalls of the fine-grained phase of the outer zone of the granite batholith, are favorable for the ore-fluid exsolution responsible for the magmatic-hydrothermal Mo mineralization.

Reaction-induced phase mixing and the formation of ultramylonitic bands

As for most upper mantle shear zones, mylonitic and ultramylonitic peridotites exposed in the northern Lanzo shear zone display thoroughly mixed, fine-grained (< 25 μm) bands. The weak crystallographic preferred orientations (CPO) of all their phases are indicative for deformation by diffusion creep. Although interpreted as significantly decreasing the strength of the upper mantle if interconnected, the formation process of fine-grained phase mixtures is still debated. Microstructural analysis of gradually higher strained peridotites of the northern Lanzo shear zone revealed the formation of fine-grained (< 25 μm) polymineralic tails adjacent to clino- and orthopyroxene porphyroclasts in mylonitic samples. Phase mixtures (av. 65% phase boundaries) are dominated by neoblasts derived from the porphyroclast (Opx/Cpx) and olivine, with minor abundances (< 10 area-%) of plagioclase, spinel ± amphibole. Neoblast microstructures and mixing of phases directly at the border of the pyroxenes, indicate a reaction of pyroxene porphyroclasts to neoblast phase mixtures. Neoblast temperature estimations (~860 °C), their systematic change in composition, and plagioclase/spinel abundances suggest that continuous net-transfer reactions, enhanced by the spinel lherzolite to plagioclase lherzolite transition are likely the main driving forces for the formation of ultramylonitic phase mixtures in the Lanzo shear zone. These reactions were potentially enhanced by the presence of fluid as evidenced by pargasitic amphibole crystallization. Comparing these results to other upper mantle shear zones and deformed xenoliths, the importance of reactions (melt-rock, fluid-rock and/or metamorphic) for the formation of polymineralic, ultramylonitic bands is confirmed.

Slow crystal settling controls the diversity of high-silica granites of the Late Cretaceous Shengsi Pluton at northeastern tip of southeast China

The intrinsic mechanism on geochemical diversity of high-silica (>70 wt% SiO2) granites (HSGs) still remains unclear. In this paper, we focus on the magmatic differentiation of the Shengsi Pluton at the northeastern tip of SE China by an integrated study of geochronology, whole-rock geochemistry and mineral chemistry of zircon and apatite. The Shengsi granites crystallized at ca. 97 Ma, and can be divided into early medium-grained phase (MGP) and late fine-grained phase (FGP). The FGP granites have higher SiO2 (75.0–77.1 wt%) contents and lower amount of plagioclase, apatite and titanite. The MGP granites likely represent the primary granitic melts and they show transitional geochemical features between I- and A-type granites. In contrast, the FGP granites may represent the melts differentiated from the MGP magma mush and they are characterized by continuously varied to strongly depleted middle rare earth elements (MREEs). Geochemical modeling demonstrates that the varied REEs of the FGP granites are likely controlled by the crystallization/accumulation of titanite, plagioclase and minor zircon during magmatic differentiation. Such a crystal fractionation process in HSGs could be explained by slow “crystal settling” (mainly titanite) in crystal mush induced by dripping solid–liquid mixtures from the magma roof. This work exemplifies the effective crystal fractionation of >70 wt% SiO2 melts to generate >75 wt% SiO2 melts in the present coastal area of SE China, and it also highlights the role of slow settling of early crystallized minerals in crystal mush on geochemical diversity of HSGs.

Coarse-to-Fine Few-Shot Defect Recognition With Dynamic Weighting and Joint Metric

Deep learning-based methods have been widely used in the defect recognition and achieved great success. However, deep learning-based methods need a large-scale dataset. While in the real industrial scenarios, the training samples are always insufficient since the defect data acquisition is difficult and time-consuming. Therefore, in this paper, the few-shot learning theory is introduced to address the challenge. We propose to achieve the few-shot defect recognition in a coarse-to-fine manner with dynamic weighting and joint metric. In the coarse-grained phase, following the feature embedding, we propose an affine dynamic weighting module to control the embedding output of all channels according to the global context. By the dynamic weighting, the model can extract discriminative features better with learnable affine parameters. In the fine-grained phase, we propose a joint metric method which contains a K-L divergence based covariance metric module (KLCM) and cosine classifier. In this method, KLCM exploits the covariance matrix from the local descriptors to represent the distribution of a special defect class and then measures the similarity between the support and query defect images. A novel few-shot defect recognition dataset FSDR which contains a variety of defects from four different surfaces is constructed to evaluate our method. The results show the state-of-the-art performance compared to the mainstream methods.

Development of CO2-Selective Polyimide-Based Gas Separation Membranes Using Crown Ether and Polydimethylsiloxane.

A series of CO2-selective polyimides (CE-PDMS-PI-x) was synthesized by copolymerizing crown ether diamine (trans-diamino-DB18C6) and PDMS-diamine with 4,4′-(hexafluoroisopropylidene) di-phthalic anhydride (6FDA) through the polycondensation reaction. The structural characteristics of the copolymers and corresponding membranes were characterized by nuclear magnetic resonance (NMR), infrared spectroscopy (ATR-FTIR), thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), X-ray diffraction (XRD), and gel permeation chromatography (GPC). The effect of PDMS loading on the CE-PDMS-PI-x copolymers was further analyzed and a very good structure–property relationship was found. A well-distributed soft PDMS unit played a key role in the membrane’s morphology, in which improved CO2-separation performance was observed at a low PDMS content (5 wt %). In contrast, the fine-grained phase separation adversely affected the separation behavior at a certain level of PDMS loading, and the PDMS was found to provide a flexible gas-diffusion path, affecting only the permeability without changing the selective gas-separation performance for the copolymers with a PDMS content of 20% or above.

STED nanoscopy - A novel way to image the pore space of geological materials.

STED nanoscopy (Stimulated Emission Depletion). which can resolve details far below the diffraction barrier has been applied hitherto preferentially to life sciences. The method is however also ideal for the investigation of geological matrices containing transparent minerals, an application tested here, to our knowledge, for the first time. The measurements on altered granitic rock and sedimentary clay rock, both containing very fine-grained phases, were conducted successfully. The STED fluorophore was dissolved in C-14-labelled methylmethacrylate (C-14-MMA) monomer which was polymerised within the rock matrix, thereby labelling the pore space in the geomaterials. Double labelling provided by the C-14-labelled MMA enables autoradiography and scanning electron microscopy (SEM), providing necessary complementary information for characterisation and quantification of porosity distributions and mineral and structure identification. Promising perspectives for further investigations of geological matrices by using different fluorophores and the optimisation of measuring procedures or even higher resolution are discussed. The combination of these different methods enlarges the observation scale of porosity from nanometre to centimetre scale.

Microstructure Evolution and Mechanical Properties of AQ80 Alloy During Forward Extrusion and Twist Deformation

A new severe plastic deformation technique, forward extrusion and twist deformation, was adopted to manufacture magnesium (Mg) alloys at different temperatures. The microstructure evolution and mechanical properties of extruded AQ80 alloy were investigated. The strain distribution in FETD extrusion was simulated with finite element simulation software. The extruded samples exhibit finer grain sizes, weaker basal textures, higher recrystallized volume fractions and higher strengths than the traditional forward extrusion samples. This technology can refine the microstructure and improve the mechanical properties of Mg alloys. At lower temperatures, the grain size is smaller with a finer second phase, which is detected as Mg17AL12. Mg17AL12 has a particle-stimulated nucleation effect and a retarding effect on grain growth, resulting in much finer DRXed grains. Due to the effects of fine-grained strengthening and second phase strengthening, the strength and elongation of AQ80 alloy can be improved significantly.

Effect of Ti element on the microstructure and properties of high chromium surfacing layer

The self-made high chromium content flux-cored wire was used to surfacing low carbon steel under self-protection. The effect of Ti contents on the microstructure, hardness and wear resistance of surfacing layer was investigated. The results showed that the surfacing layer was mainly composed of martensite, [Formula: see text] carbide and residual austenite. As Ti content in surfacing layer increased, the shape of carbides in surfacing layer varied from lath to granular and the amount of carbides was gradually increased. The hardness of the surfacing layer was slightly increased while the wear resistance was significantly enhanced. The improved wear property could be ascribed to the increasing Ti content and the corresponding increasing amount of Ti (C, N). Ti (C, N) takes the role of heterogenous nucleation particles in the surfacing layer and promotes the formation of fine-grained hard phase (e.g., [Formula: see text] carbide). With the increase of the density of fine-grained hard phase in the surfacing layer, the resistance of cutting motion during abrasive wear test also increased, resulting in a decreased wear loss and significantly enhanced wear performance.

Rock magnetic signature of a Miocene playa cycle in Central Asia and environmental implications

The increased aridification of Central Asia during the Miocene coincides in time with lake formations and the evolution of playa environments in the region. However, Miocene continental climate dynamics and the forcing of aridification are still not well constrained. Neogene lacustrine mudflat deposits in the Ili Basin in southeast Kazakhstan provide a well-exposed paleoclimate archive. Here, we present a detailed rock magnetic study of a middle Miocene playa cycle deposited in a closed basin. We use high-resolution rock magnetic parameters, lithological studies and geochemistry to reconstruct the playa environment and the depositional conditions. The rock magnetic mineralogy of the playa cycle is controlled by hematite and two fine-grained magnetite phases. Increased magnetic concentrations occur during dry mudflat conditions, with a lower groundwater table and increased aridity. The underlying processes controlling the observed variation in magnetic concentrations are a complex interplay of diagenetic processes during and after deposition. The data support an authigenic origin of both magnetite phases, one formed before and the other after sediment consolidation. Early diagenetic formation of fine-grained magnetite by microbial activity is followed by post-depositional formation of a secondary fine-grained magnetite phase. The rock magnetic results such as magnetic concentration-dependent parameters, ARM/SIRM and s-ratio indicate a sensitive record of (ground) water availability and aridity changes in the Ili Basin. We suggest that they can serve as an effective proxy for detailed paleo-environment reconstruction of playa evolution, not only in the middle Miocene Ili Basin but also in comparable floodplain/playa lake settings.

Cosmogenic 3He production rate in ilmenite and the redistribution of spallation 3He in fine-grained minerals

Cosmogenic nuclide surface exposure dating and erosion rate measurements in basaltic landscapes rely primarily on measurement of ^3He in olivine or pyroxene. However, geochemical investigations using ^3He have been impossible in the substantial fraction of basalts that lack separable olivine or pyroxene crystals, or where such crystals were present, but have been chemically weathered. Fine-textured basalts often contain small grains of ilmenite, a weathering-resistant mineral that is a target for cosmogenic ^3He production with good He retention and straightforward mineral separation, but with a poorly constrained production rate. Here we empirically calibrate the cosmogenic ^3He production rate in ilmenite by measuring ^3He concentrations in basalts with fine-grained (∼20 μm cross-section) ilmenite and co-existing pyroxene or olivine from the Columbia River and Snake River Plain basalt provinces in the western United States. The concentration ratio of ilmenite to pyroxene and olivine is 0.78 ± 0.02, yielding an apparent cosmogenic ^3He production rate of 93.6 ± 7.7 atom g^(−1) yr^(−1) that is 20–30% greater than expected from prior theoretical and empirical estimates for compositionally similar minerals. The production rate discrepancy arises from the high energy with which cosmic ray spallation reactions emit tritium and ^3He and the associated long stopping distances that cause them to redistribute within a rock. Fine-grained phases with low cosmogenic ^3He production rates, like ilmenite, will have anomalously high production rates owing to net implantation of ^3He from the surrounding, higher ^3He production rate, matrix. Semi-quantitative modeling indicates implantation of spallation ^3He increases with decreasing ilmenite grain size, leading to production rates that exceed those in a large grain by ∼10% when grain radii are <150 μm. The modeling predicts that for the ilmenite grain size in our samples, implantation causes production rates to be ∼20% greater than expected for a large grain, and within uncertainty resolves the discrepancy between our calibrated production rate, theory, and rates from previous work. The redistribution effect is maximized when the host rock and crystals differ substantially in mean atomic number, as they do between whole-rock basalt and ilmenite.

Addressing the Under-Translation Problem from the Entropy Perspective

Neural Machine Translation (NMT) has drawn much attention due to its promising translation performance in recent years. However, the under-translation problem still remains a big challenge. In this paper, we focus on the under-translation problem and attempt to find out what kinds of source words are more likely to be ignored. Through analysis, we observe that a source word with a large translation entropy is more inclined to be dropped. To address this problem, we propose a coarse-to-fine framework. In coarse-grained phase, we introduce a simple strategy to reduce the entropy of highentropy words through constructing the pseudo target sentences. In fine-grained phase, we propose three methods, including pre-training method, multitask method and two-pass method, to encourage the neural model to correctly translate these high-entropy words. Experimental results on various translation tasks show that our method can significantly improve the translation quality and substantially reduce the under-translation cases of high-entropy words.