Search NASA⌕ Search

SEARCH · Search NASA

Results for “Generative modeling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 955 records · Page 53

Trust Your Gut: Comparing Human and Machine Inference from Noisy Visualizations

People commonly utilize visualizations not only to examine a given dataset, but also to draw generalizable conclusions about the underlying models or phenomena. Prior research has compared human visual inference to that of an optimal Bayesian agent, with deviations from rational analysis viewed as problematic. However, human reliance on non-normative heuristics may prove advantageous in certain circumstances. We investigate scenarios where human intuition might surpass idealized statistical rationality. In two experiments, we examine individuals’ accuracy in characterizing the parameters of known data-generating models from bivariate visualizations. Our findings indicate that, although participants generally exhibited lower accuracy compared to statistical models, they frequently outperformed Bayesian agents, particularly when faced with extreme samples. Participants appeared to rely on their internal models to filter out noisy visualizations, thus improving their resilience against spurious data. However, participants displayed overconfidence and struggled with uncertainty estimation. They also exhibited higher variance than statistical machines. Our findings suggest that analyst gut reactions to visualizations may provide an advantage, even when departing from rationality. These results carry implications for designing visual analytics tools, offering new perspectives on how to integrate statistical models and analyst intuition for improved inference and decision-making. The data and materials for this paper are available at https://osf.io/qmfv6

human-machine collaboration↗

Estimation of Melting-Layer Cooling Rate from Dual-Polarization Radar: Spectral Bin Model Simulations

Diabatic cooling from hydrometeor phase changes in the stratiform melting layer is of great interest to both operational forecasters and modelers for its societal and dynamical consequences. Attempts to estimate the melting-layer cooling rate typically rely on either the budgeting of hydrometeor content estimated from reflectivity Z or model-generated lookup tables scaled by the magnitude of Z in the bright band. Recent advances have been made in developing methods to observe the unique polarimetric characteristics of melting snow and the additional microphysical information they may contain. However, to date no work has looked at the thermodynamic information available from the polarimetric radar brightband signature. In this study, a one-dimensional spectral bin model of melting snow and a coupled polarimetric operator are used to study the relation between the polarimetric radar bright band and the melting-layer cooling rate. Simulations using a fixed particle size distribution (PSD) and variable environmental conditions show that the height and thickness of the bright band and the maximum brightband Z and specific differential phase shift [Formula: see text] are all sensitive to the ambient environment, while the differential reflectivity [Formula: see text] is relatively insensitive. Additional simulations of 2700 PSDs based on in situ observations above the melting layer indicate that the maximum Z, [Formula: see text], and [Formula: see text] within the melting layer are poorly correlated with the maximum cooling rate while [Formula: see text] is strongly correlated. Finally, model simulations suggest that, in addition to riming, concurrent changes in aggregation and precipitation intensity and the associated cooling may plausibly cause observed sagging brightband signatures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Microstructure Segmentation with Deep Learning Encoders Pre-Trained on a Large Microscopy Dataset

This study examined the improvement of microscopy segmentation accuracy by transfer learning from a large dataset of microscopy images called MicroNet. Many neural network encoder architectures, including VGG, Inception, and ResNet, were trained on over 100,000 labelled microscopy images from 54 classes. These pre-trained encoders were then embedded into multiple segmentation architectures including U-Net and DeepLabV3+ to evaluate segmentation performance on newly created benchmark microscopy datasets. Compared to ImageNet pre-training, models pre-trained on MicroNet generalized better to out-of-distribution micrographs taken under different imaging and sample conditions and were more accurate with less training data. When training with only a single Ni-superalloy image, pre-training on MicroNet produced a 72.2 percent reduction in relative segmentation error. These results suggest that transfer learning from large in-domain datasets generate models with learned feature representations that are more useful for downstream tasks and will likely improve any microscopy image analysis technique that can leverage pre-trained encoders.

machine learning↗

Modeling Spatial Asymmetries in Teleconnected Extreme Temperatures

Abstract Combining strengths from deep learning and extreme value theory can help describe complex relationships between variables where extreme events have significant impacts (e.g., environmental or financial applications). Neural networks learn complicated nonlinear relationships from large datasets under limited parametric assumptions. By definition, the number of occurrences of extreme events is small, which limits the ability of the data-hungry, nonparametric neural network to describe rare events. Inspired by recent extreme cold winter weather events in North America caused by atmospheric blocking, we examine several probabilistic generative models for the entire multivariate probability distribution of daily boreal winter surface air temperature. We propose metrics to measure spatial asymmetries, such as long-range anticorrelated patterns that commonly appear in temperature fields during blocking events. Compared to vine copulas, the statistical standard for multivariate copula modeling, deep learning methods show improved ability to reproduce complicated asymmetries in the spatial distribution of ERA5 temperature reanalysis, including the spatial extent of in-sample extreme events.

Krock, Mitchell L.↗

Multiscale modeling of polyisoprene on graphite

The local dynamics and the conformational properties of polyisoprene next to a smooth graphite surface constructed by graphene layers are studied by a multiscale methodology. First, fully atomistic molecular dynamics simulations of oligomers next to the surface are performed. Subsequently, Monte Carlo simulations of a systematically derived coarse-grained model generate numerous uncorrelated structures for polymer systems. A new reverse backmapping strategy is presented that reintroduces atomistic detail. Finally, multiple extensive fully atomistic simulations with large systems of long macromolecules are employed to examine local dynamics in proximity to graphite. Polyisoprene repeat units arrange close to a parallel configuration with chains exhibiting a distribution of contact lengths. Efficient Monte Carlo algorithms with the coarse-grain model are capable of sampling these distributions for any molecular weight in quantitative agreement with predictions from atomistic models. Furthermore, molecular dynamics simulations with well-equilibrated systems at all length-scales support an increased dynamic heterogeneity that is emerging from both intermolecular interactions with the flat surface and intramolecular cooperativity. This study provides a detailed comprehensive picture of polyisoprene on a flat surface and consists of an effort to characterize such systems in atomistic detail.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Microstructure Segmentation With Deep Learning Encoders Pre-Trained on a Large Microscopy Dataset

This study examined the improvement of microscopy segmentation intersection over union accuracy by transfer learning from a large dataset of microscopy images called MicroNet. Many neural network encoder architectures were trained on over 100,000 labeled microscopy images from 54 material classes. These pre-trained encoders were then embedded into multiple segmentation architectures including UNet and DeepLabV3+ to evaluate segmentation performance on created benchmark microscopy datasets. Compared to ImageNet pre-training, models pre-trained on MicroNet generalized better to out-of-distribution micrographs taken under different imaging and sample conditions and were more accurate with less training data. When training with only a single Ni-superalloy image, pre-training on MicroNet produced a 72.2% reduction in relative intersection over union error. These results suggest that transfer learning from large in-domain datasets generate models with learned feature representations that are more useful for downstream tasks and will likely improve any microscopy image analysis technique that can leverage pre-trained encoders.

machine learning↗

Optical bandgap of single- and multi-layered amorphous germanium ultra-thin films

Accurate optical methods are required to determine the energy bandgap of amorphous semiconductors and elucidate the role of quantum confinement in nanometer-scale, ultra-thin absorbing layers. Here, we provide a critical comparison between well-established methods that are generally employed to determine the optical bandgap of thin-film amorphous semiconductors, starting from normal-incidence reflectance and transmittance measurements. First, we demonstrate that a more accurate estimate of the optical bandgap can be achieved by using a multiple-reflection interference model. We show that this model generates more reliable results compared to the widely accepted single-pass absorption method. Second, we compare two most representative methods (Tauc and Cody plots) that are extensively used to determine the optical bandgap of thin-film amorphous semiconductors starting from the extracted absorption coefficient. Analysis of the experimental absorption data acquired for ultra-thin amorphous germanium (a-Ge) layers demonstrates that the Cody model is able to provide a less ambiguous energy bandgap value. Finally, we apply our proposed method to experimentally determine the optical bandgap of a-Ge/SiO{sub 2} superlattices with single and multiple a-Ge layers down to 2 nm thickness.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

AGR-1, AGR-2 and AGR-3/4 Release-to-Birth Ratio Data Analysis

A series of Advanced Gas Reactor (AGR) irradiation tests is being conducted in the Advanced Test Reactor at Idaho National Laboratory in support of development and qualification of tristructural isotropic (TRISO) low enriched fuel used in the High Temperature Gas-cooled Reactor (HTGR). Each AGR test consists of multiple independently controlled and monitored capsules containing fuel compacts placed in a graphite cylinder shrouded by a steel shell. These capsules are instrumented with thermocouples embedded in the graphite enabling temperature control. AGR configuration and irradiation conditions are based on prismatic HTGR technology distinguished primarily by the use of helium coolant, a low-power-density ceramic core capable of withstanding very high temperatures, and TRISO coated particle fuel. Thus, these tests provide valuable irradiation performance data to support fuel process development, qualify fuel for normal operating conditions, and support development and validation of fuel performance and fission product transport models and codes. The release-rate-to-birth-rate ratio (R/B) for each of fission product isotopes (i.e., krypton and xenon) is calculated from release rates in the sweep gas flow measured by the germanium detectors used in the AGR Fission Product Monitoring (FPM) System installed downstream from each irradiated capsule. Birth rates are calculated based on the fission power in the experiment and fission product generation models. Thus, this R/B is a measure of the ability of fuel kernel, particle coating layers, and compact matrix to retain fission gas atoms preventing their release into the sweep gas flow, especially in the event of particle coating failures that occurred during AGR-2 and AGR-3/4 irradiations. The major factors that govern gaseous radioactive decay, diffusion, and release processes are found to be material diffusion coefficient, temperature, and isotopic decay constant. For each of all AGR capsules, ABAQUS-based three-dimensional finite-element thermal models are created to predict daily averages of fuel compact temperatures for the entire irradiation period, which are used in establishing the R/B correlation with temperature and decay constant. This correlation can be used by reactor designers to estimate fission gas release from postulated failed fuel particles in HTGR cores, which is the key safety factor for fuel performance assessment.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

AGR-2 and AGR-3/4 Release-to-Birth Ratio Data Analysis

A series of Advanced Gas Reactor (AGR) irradiation tests is being conducted in the Advanced Test Reactor at Idaho National Laboratory in support of development and qualification of tristructural isotropic (TRISO) low enriched fuel used in the High Temperature Gas-cooled Reactor (HTGR). Each AGR test consists of multiple independently controlled and monitored capsules containing fuel compacts placed in a graphite cylinder shrouded by a steel shell. These capsules are instrumented with thermocouples embedded in the graphite enabling temperature control. AGR configuration and irradiation conditions are based on prismatic HTGR technology distinguished primarily by the use of helium coolant, a low-power-density ceramic core capable of withstanding very high temperatures, and TRISO coated particle fuel. Thus, these tests provide valuable irradiation performance data to support fuel process development, qualify fuel for normal operating conditions, and support development and validation of fuel performance and fission product transport models and codes. The release-rate-to-birth-rate ratio (R/B) for each of fission product isotopes (i.e., krypton and xenon) is calculated from release rates in the sweep gas flow measured by the germanium detectors used in the AGR Fission Product Monitoring (FPM) System installed downstream from each irradiated capsule. Birth rates are calculated based on the fission power in the experiment and fission product generation models. Thus, this R/B is a measure of the ability of fuel kernel, particle coating layers, and compact matrix to retain fission gas atoms preventing their release into the sweep gas flow, especially in the event of particle coating failures that occurred during AGR-2 and AGR-3/4 irradiations. The major factors that govern gaseous radioactive decay, diffusion, and release processes are found to be material diffusion coefficient, temperature, and isotopic decay constant. For each of all AGR capsules, ABAQUS-based three-dimensional finite-element thermal models are created to predict daily averages of fuel compact temperatures for the entire irradiation period, which are used in establishing the R/B correlation with temperature and decay constant. This correlation can be used by reactor designers to estimate fission gas release from postulated failed fuel particles in HTGR cores, which is the key safety factor for fuel performance assessment.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Chromium-doped uranium dioxide fuels: A review

UO 2 doped with parts per million CR 2 O 3 powder is considered a potential near term accident tolerant fuel candidate. Here, the results of decades of industry and academic research into Cr-doped UO 2 are analyzed and their shortcomings are critiqued. Focusing on the incorporation mechanisms of Cr into the fuel matrix, we explore a mechanistic understanding of the characteristic properties of Cr-doped UO 2 , notably, enhanced fission gas retention attributed to enlarged grain sizes following sintering, along with marginal improvements in the thermophysical properties. The findings of recent X-ray Adsorption Near Edge Spectroscopy studies were compared and put into conversation with historic data regarding the incorporation of Cr in UO 2 . On the basis of defect mechanisms, the case is made for the substitutional incorporation of Cr governing the lattice solubility but not the enhanced U diffusivity. Instead, Cr/CR 2 O 3 redox chemistry in a well-defined oxygen potential explains the differences in the U diffusivity and O/M ratio. The primary mechanism of doping enhanced grain growth is found to be liquid assisted sintering due to a CRO (1) eutectic phase at the grain boundaries. The role of inhomogeneities in Cr concentration in UO 2 at various length scales across the materials microstructure is highlighted and connected to promising experimental and modeling work to fill in the gaps in the current understanding of Cr-doped UO 2 . The review considers both the open scientific questions and engineering applications to illustrate the deep connections between the practice and theory in the design of accident tolerant nuclear fuels. In conclusion, the review ends with an outline of future works that combine meticulous irradiation studies and high resolution experiments with next generation modeling and simulations techniques empowered by machine learning advances to accelerate the fabrication and adoption of Cr-doped UO 2 light water reactors.

Cleveland, Mack Wesley [Massachusetts Inst. of Tec↗

Using PAFEC as a preprocessor for COSMIC/NASTRAN

Programs for Automatic Finite Element Calculations (PAFEC) is a general purpose, three dimensional linear and nonlinear finite element program (ref. 1). PAFEC's features include free format input utilizing engineering keywords, powerful mesh generating facilities, sophisticated data base management procedures, and extensive data validation checks. Presented here is a description of a software interface that permits PAFEC to be used as a preprocessor for COSMIC/NASTRAN. This user friendly software, called PAFCOS, frees the stress analyst from the laborious and error prone procedure of creating and debugging a rigid format COSMIC/NASTRAN bulk data deck. By interactively creating and debugging a finite element model with PAFEC, thus taking full advantage of the free format engineering keyword oriented data structure of PAFEC, the amount of time spent during model generation can be drastically reduced. The PAFCOS software will automatically convert a PAFEC data structure into a COSMIC/NASTRAN bulk data deck. The capabilities and limitations of the PAFCOS software are fully discussed in the following report.

Gray, W. H.↗

Statistical Approach on Utilizing Ground-Based Experiments to Model Break-Up Events

Two laboratory-based impact tests have been conducted to develop and extend the capability of NASA satellite breakup models. The first experiment, the Satellite Orbital Debris Characterization Impact Test (SOCIT), was conducted by the U.S. Department of Defense and NASA in 1992. It employed a fully functional U.S. Navy Transit spacecraft, fabricated from materials commonly used in the 1960s. SOCIT fragmentation data formed the basis of the current NASA Standard Satellite Breakup Model (SSBM). A second experiment, DebriSat, was conducted in 2014 by the DebriSat consortium: the NASA Orbital Debris Program Office, the United States Space Force Space Systems Command, formerly the Air Force Space and Missile Systems Center, the Air Force Arnold Engineering Development Complex, and the University of Florida. This impact test was performed on a high-fidelity mock-up satellite assembled from modern components. Data from both experiments are expected to contribute to the next-generation models for on-orbit breakup analyses, long-term environment predictions, and debris risk assessment. This paper uses the direct statistical sampling of the SOCIT and DebriSat data ensembles to model an on-orbit breakup event, rather than the analytic expressions estimated from the samples in the SSBM and its predecessor. This direct method involves drawing fragments (i.e., sampling with replacement) from each fragmentation data-subset containing mass, shape category, material density, characteristic length, mass, and cross-sectional area. As part of the sampling process, the properties of sampled fragment data (e.g., area-to-mass ratios) are numerically checked to ensure they do not contain unrealistic quantities. The process to simulate a breakup cloud composed of fragments from SOCIT and DebriSat datasets is discussed. The methodology for ensuring that conservation of overall mass of the sampled fragment cloud under this sampling approach is highlighted. Finally, the results are compared with simulated clouds generated from SSBM for specific historical breakup events.

Andrew Vavrin↗

Statistical Approach on Utilizing Ground-based Experiments to Model Break-up Events

Two laboratory-based impact tests have been conducted to develop and extend the capability of NASA satellite breakup models. The first experiment, the Satellite Orbital Debris Characterization Impact Test (SOCIT), was conducted by the U.S. Department of Defense and NASA in 1992. It employed a fully functional U.S. Navy Transit spacecraft, fabricated from materials commonly used in the 1960s. SOCIT fragmentation data formed the basis of the current NASA Standard Satellite Breakup Model (SSBM). A second experiment, DebriSat, was conducted in 2014 by the DebriSat consortium: the NASA Orbital Debris Program Office; the United States Space Force Space Systems Command, formerly the Air Force Space and Missile Systems Center, the Air Force Arnold Engineering Development Complex, and the University of Florida. This impact test was performed on a high-fidelity mock-up satellite assembled from modern components. Data from both experiments are expected to contribute to the next-generation models for on-orbit breakup analyses, long-term environment predictions, and debris risk assessment. This paper uses the direct statistical sampling of the SOCIT and DebriSat data ensembles to model an on-orbit breakup event, rather than the analytic expressions estimated from the samples as with as with the NASA SSBM and its predecessor. This direct method involves drawing fragments (i.e., sampling with replacement) from each fragmentation data-subset containing mass, shape category, material density, characteristic length, mass, and cross-sectional area. As part of the sampling process, the properties of sampled fragment data (e.g., area-to-mass ratios) are numerically checked to ensure they do not contain unrealistic quantities. The process to simulate a breakup cloud composed of fragments from SOCIT and DebriSat datasets is discussed. The methodology for ensuring that conservation of overall mass of the sampled fragment cloud under this sampling approach is highlighted. Finally, the results are compared with simulated clouds generated from SSBM for specific historical breakup events.

Andrew Vavrin↗

Prediction of Carbon Dioxide Adsorption via Deep Learning

Abstract Porous carbons with different textural properties exhibit great differences in CO 2 adsorption capacity. It is generally known that narrow micropores contribute to higher CO 2 adsorption capacity. However, it is still unclear what role each variable in the textural properties plays in CO 2 adsorption. Herein, a deep neural network is trained as a generative model to direct the relationship between CO 2 adsorption of porous carbons and corresponding textural properties. The trained neural network is further employed as an implicit model to estimate its ability to predict the CO 2 adsorption capacity of unknown porous carbons. Interestingly, the practical CO 2 adsorption amounts are in good agreement with predicted values using surface area, micropore and mesopore volumes as the input values simultaneously. This unprecedented deep learning neural network (DNN) approach, a type of machine learning algorithm, exhibits great potential to predict gas adsorption and guide the development of next‐generation carbons.

Zhang, Zihao↗

Assessing Temporal Effect of Economic Activity on Freight Volumes with Two-Period Cross-Sectional Data

The most comprehensive publicly available freight databases are the Commodity Flow Survey (CFS) and the FHWA s Freight Analysis Framework (FAF). These two sources contain dollar value and weight of freight movements at high geographic levels, such as state or metropolitan areas. Due to the difficulty in obtaining freight data at lower geographies various practitioners and researchers have been suggesting to estimate freight models based on aggregate data. Following these recent practices, a methodology to estimate a nationwide production and attraction models for U.S. domestic trade of goods is presented. To this end, a CFS s data set provided by U.S. Census Bureau and composed of two-nonconsecutive year period (2002 and 2007) of movements of goods between U.S. states for 27 industry sectors was used. The state payroll by industry sector, obtained from the County Business Patterns of the U.S. Census, was the variable used to estimate freight generation models. The main objective of this paper is to analyze the temporal stability and predictability of the proposed aggregate models. The results indicate that the payroll alone explains a significant portion of the freight production and attraction at the state level. However, such simplification in the model process did not result in reasonable predictions of freight for a future year horizon. It is recommended that time-dependent factors (e.g. variables related to changes industry productivity) affecting freight demand should be considered in the modeling process.

Oliveira Neto, Francisco Moraes↗

3D reconstruction of carbon nanotube networks from neutron scattering experiments

Structure reconstruction from statistical descriptors, such as scattering data obtained using x-rays or neutrons, is essential in understanding various properties of nanocomposites. Scattering based reconstruction can provide a realistic model, over various length scales, that can be used for numerical simulations. In this study, 3D reconstruction of a highly loaded carbon nanotube (CNT)-conducting polymer system based on small and ultra-small angle neutron scattering (SANS and USANS, respectively) data was performed. These light-weight and flexible materials have recently shown great promise for high-performance thermoelectric energy conversion, and their further improvement requires a thorough understanding of their structure-property relationships. The first step in achieving such understanding is to generate models that contain the hierarchy of CNT networks over nano and micron scales. The studied system is a single walled carbon nanotube (SWCNT)/poly (3,4-ethylenedioxythiophene): poly (styrene sulfonate) (PEDOT: PSS). SANS and USANS patterns of the different samples containing 10, 30, and 50 wt% SWCNTs were measured. These curves were then utilized to calculate statistical two-point correlation functions of the nanostructure. These functions along with the geometrical information extracted from SANS data and scanning electron microscopy images were used to reconstruct a representative volume element (RVE) nanostructure. Generated RVEs can be used for simulations of various mechanical and physical properties. This work, therefore, introduces a framework for the reconstruction of 3D RVEs of high volume faction nanocomposites containing high aspect ratio fillers from scattering experiments.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Prediction of Carbon Dioxide Adsorption via Deep Learning

Porous carbons with different textural properties exhibit great differences in CO 2 adsorption capacity. It is generally known that narrow micropores contribute to higher CO 2 adsorption capacity. However, it is still unclear what role each variable in the textural properties plays in CO 2 adsorption. Herein, a deep neural network is trained as a generative model to direct the relationship between CO 2 adsorption of porous carbons and corresponding textural properties. The trained neural network is further employed as an implicit model to estimate its ability to predict the CO 2 adsorption capacity of unknown porous carbons. Interestingly, the practical CO 2 adsorption amounts are in good agreement with predicted values using surface area, micropore and mesopore volumes as the input values simultaneously. This unprecedented deep learning neural network (DNN) approach, a type of machine learning algorithm, exhibits great potential to predict gas adsorption and guide the development of next-generation carbons.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Quantifying discordance in the 2015 Planck CMB spectrum

We examine the internal consistency of the Planck 2015 cosmic microwave background (CMB) temperature anisotropy power spectrum. We show that tension exists between cosmological constant cold dark matter (ΛCDM) model parameters inferred from multipoles ℓ<1000 (roughly those accessible to Wilkinson Microwave Anisotropy Probe), and from ℓ⩾1000, particularly the CDM density, Ω{sub c}h{sup 2}, which is discrepant at 2.5σ for a Planck -motivated prior on the optical depth, τ=0.07±0.02. We find some parameter tensions to be larger than previously reported because of inaccuracy in the code used by the Planck Collaboration to generate model spectra. The Planck ℓ⩾1000 constraints are also in tension with low-redshift data sets, including Planck ’s own measurement of the CMB lensing power spectrum (2.4σ), and the most precise baryon acoustic oscillation scale determination (2.5σ). The Hubble constant predicted by Planck from ℓ⩾1000, H{sub 0}=64.1±1.7 km s{sup −1} Mpc{sup −1}, disagrees with the most precise local distance ladder measurement of 73.0±2.4 km s{sup −1} Mpc{sup −1} at the 3.0σ level, while the Planck value from ℓ<1000, 69.7±1.7 km s{sup −1} Mpc{sup −1}, is consistent within 1σ. A discrepancy between the Planck and South Pole Telescope high-multipole CMB spectra disfavors interpreting these tensions as evidence for new physics. We conclude that the parameters from the Planck high-multipole spectrum probably differ from the underlying values due to either an unlikely statistical fluctuation or unaccounted-for systematics persisting in the Planck data.

79 ASTRONOMY AND ASTROPHYSICS↗