TDTR实验、热物性调控、机器学习

Silicon carbide silicon carbide (SiC SiC) composites are often used in oxidizing environments at high temperatures. Measurements of the thermal conductance of the oxide layer provide a way to better understand the oxidation process with high spatial resolution. We use time domain thermoreflectance (TDTR) to map the thermal conductance of the oxide layer and the thermal conductivity of the SiC SiC composite with a spatial resolution of 3 μm. Heterodyne detection using a 50 kHz modulated probe beam and a 10 MHz modulated pump suppresses the coherent pick-up and enables faster data acquisition than what has previously been possible using sequential demodulation. By analyzing the noise of the measured signals, we find that in the limit of small integration time constants or low laser powers, the dominant source of noise is the input noise of the preamplifier. The thermal conductance of the oxide that forms on the fiber region is lower than the oxide on the matrix due to small differences in thickness and thermal conductivity.

Time-domain thermoreflectance (TDTR) is a well-established pump/probe method for measuring thermal conductivity and interface conductance of multilayers. Interpreting signals in a TDTR experiment requires a thermal model.In standard front/front TDTR experiments, both pump and probe beams typically irradiate the surface of a multilayer. As a result, existing thermal models for interpreting thermoreflectance experiments assume the pump and probe beams both interact with the surface layer. Here, we present a frequency-domain solution to the heat-diffusion equation of a multilayer in response to nonhomogenous laser heating. This model allows analysis of experiments where the pump and probe beams irradiate opposite sides of a multilayer. We call such a geometry a front/back experiment to differentiate such experiments from standard TDTR experiments. As an example, we consider a 60nm amorphous Si film. We consider how signals differ in a front/front vs. front/back geometry and compare thermal model predictions to experimental data.

We have developed a transient thermoreflectance technique using picosecond pulsed and cw laser to study thermal conductivity and interface conductance in both thin-films and bulk materials. A real time-resolved system observes a thermal transport along the cross-plane direction of the sample during a single pulse excitation. The suggested TTR technique can measure thermal conductivity in up to a few hundred nm of thin films with a reasonable uncertainty by carefully selecting metal transducer thickness. In this paper, we examine thermal conductivity in several substrates including Si, GaAs, Sapphire, and Glass after depositing Au thin film as metal transducer and compare with reported values to validate our technique. For further study on our method, MoS2 thin-films with different thicknesses are prepared via exfoliating, and their thermal conductivity are measured as average value of 3.4 W/mK. Compared to TDTR technique, TTR is a simpler and inexpensive method to study thermophysical properties and can also measure in-plane thermal property using a grating imaging technique. TTR can be one of the available options for observing thermal transport phenomena in both horizontal and vertical directions with a simple and inexpensive preparation.

Silicon carbide (SiC) is a wide bandgap (WBG) semiconductor with promising applications in high-power and high-frequency electronics. Among its many useful properties, the high thermal conductivity is crucial. In this letter, the anisotropic thermal conductivity of three SiC samples: n-type 4H-SiC (N-doped 1x10^19 cm-3), unintentionally doped (UID) semi-insulating (SI) 4H-SiC, and SI 6H-SiC (V-doped 1x10^17 cm-3), is measured using femtosecond laser based time-domain thermoreflectance (TDTR) over a temperature range from 250 K to 450 K. We simultaneously measure the thermal conductivity parallel to (k_r) and across the hexagonal plane (k_z) for SiC by choosing the appropriate laser spot radius and the modulation frequency for the TDTR measurements. For both k_r and k_z, the following decreasing order of thermal conductivity value is observed: SI 4H-SiC > n-type 4H-SiC > SI 6H-SiC. This work serves as an important benchmark for understanding thermal transport in WBG semiconductors.

Thermoreflectance techniques, including time-domain thermoreflectance (TDTR), frequency-domain thermoreflectance (FDTR), and the square-pulsed source (SPS) method, are powerful tools for characterizing the thermal properties of bulk and thin-film materials. However, accurately interpreting their signals remains challenging due to intricate interdependencies among experimental variables. This study introduces a systematic framework based on singular value decomposition (SVD) to decouple these interdependent parameters and enhance the reliability of thermal property extraction. By applying SVD to the sensitivity matrix, we identify key parameter combinations and establish essential dimensionless numbers that govern thermoreflectance signals. The framework is applied to a GaN/Si heterostructure, where the performance of TDTR, FDTR, and SPS is evaluated and compared. The results demonstrate a high degree of consistency across all three techniques. Notably, with the intricate relationships of parameters unraveled, TDTR, FDTR, and SPS demonstrate significant potential to simultaneously and accurately extract five to seven key thermal properties, including thermal conductivity, heat capacity, and interfacial thermal conductance of the GaN/Si multilayer system. This framework not only improves the precision of thermoreflectance measurements but also lays a foundation for advanced thermal metrology in research and industrial applications.

Spatial mapping of thermal properties is critical for unveiling the structure-property relation of materials, heterogeneous interfaces, and devices. These property images can also serve as datasets for training artificial intelligence models for material discoveries and optimization. Here we introduce a high-throughput thermal property imaging method called compressive sensing frequency domain thermoreflectance (CS-FDTR), which can robustly profile thermal property distributions with micrometer resolutions while requiring only a random subset of pixels being experimentally measured. The high-resolution thermal property image is reconstructed from the raw down-sampled data through L_1-regularized minimization. The high-throughput imaging capability of CS-FDTR is validated using the following cases: (a) the thermal conductance of a patterned heterogeneous interface, (b) thermal conductivity variations of an annealed pyrolytic graphite sample, and (c) the sharp change in thermal conductivity across a vertical aluminum/graphite interface. With less than half of the pixels being experimentally sampled, the thermal property images measured using CS-FDTR show nice agreements with the ground truth (point-by-point scanning), with a relative deviation below 15%. This work opens the possibility of high-throughput thermal property imaging without sacrificing the data quality, which is critical for materials discovery and screening.

Cubic boron arsenide (c-BAs) has been theoretically predicted to exhibit thermal conductivity \k{appa} comparable to that of diamond, yet experimental measurements have plateaued at ~1300W/mK. We report room-temperature \k{appa} exceeding 2000W/mK in c-BAs, on par with single-crystal diamond. This finding is enabled by high-quality single crystals and a newly developed nanosecond, transducer-less time-domain thermoreflectance technique that allows spatial mapping of \k{appa} without metal transducers. Thermal conductivity correlates with crystal quality, as evidenced by stronger photoluminescence and longer photoluminescence lifetimes. However, the observed nanosecond lifetimes remain shorter than expected for an indirect bandgap semiconductor, suggesting room for further crystal quality improvement and higher \k{appa}. These results challenge current theoretical models and highlight c-BAs as a promising material for next-generation electronics.

In this work we present results of the analysis of thermal properties of high-power laser bars obtained by spatially resolved thermoreflectance (TR) spectroscopy. Thermoreflectance is a modulation technique relying on periodic facet temperature modulation induced by pulsed current supply of the laser. The periodic temperature change of the laser induces variation of the refractive index and consequently modulates probe beam reflectivity. The technique has a spatial resolution of about ~1 $μ$m and can be used for temperature mapping over 300 $μ$m x 300 $μ$m area. Information obtained in these experiments provide an insight into thermal processes occurring at devices' facets and consequently lead to increased reliability and substantially longer lifetimes of such structures.

Whether diffuse mismatch model for electrons (DMMe) hold true in more general cases remains largely unexplored, especially in cases where at least one material does not behave like a free-electron metal and/or the interface is smooth enough to allow non-diffuse transmission of electrons. In this study, DMMe was proposed to predict the thermal conductance of metal-metal interfaces. A set of aluminum-X samples (X = Cu, Ag, Fe, Ni) were grown and the time domain thermoreflectance (TDTR) technique was used to measure the metal-metal interface conductance. It was then compared to the two variants of the DMMe-using both a crude theory based on free-electron metals and accurate band structures provided by density functional theory.

It is challenging to characterize thermal conductivity of materials with strong anisotropy. In this work, we extend the time-domain thermoreflectance (TDTR) method with a variable spot size approach to simultaneously measure the in-plane (Kr) and the through-plane (Kz) thermal conductivity of materials with strong anisotropy. We first determine Kz from the measurement using a larger spot size, when the heat flow is mainly one-dimensional along the through-plane direction, and the measured signals are sensitive to only Kz. We then extract the in-plane thermal conductivity Kr from a second measurement using the same modulation frequency but with a smaller spot size, when the heat flow becomes three-dimensional, and the signal is sensitive to both Kr and Kz. By choosing the same modulation frequency for the two sets of measurements, we can avoid potential artifacts introduced by the frequency-dependent Kz, which we have found to be non-negligible, especially for some two-dimensional layered materials like MoS2. After careful evaluation of the sensitivity of a series of hypothetical samples, we provided a guideline on choosing the most appropriate laser spot size and modulation frequency that yield the smallest uncertainty, and established a criterion for the range of thermal conductivities that can be measured reliably using our proposed variable spot size TDTR approach. We have demonstrated this variable spot size TDTR approach on samples with a wide range of in-plane thermal conductivity, including fused silica, rutile titania (TiO2 [001]), zinc oxide (ZnO [0001]), molybdenum disulfide (MoS2), hexagonal boron nitride (h-BN), and highly ordered pyrolytic graphite (HOPG).

We present experimental and theoretical investigations of phonon thermal transport in (LaMnO$_3$)$_m$/(SrMnO$_3$)$_n$ superlattices (LMO/SMO SLs) with the thickness of individual layers $m,n = 3 - 10\;$ u.c. and the thickness ratio $m/n = 1, 2$. Optical transient thermal reflectivity measurements reveal a pronounced difference in the thermal conductivity between SLs with $m/n = 1$, and SLs with $m/n = 2$. State-of-the art electron microscopy techniques and ab-initio density functional calculations enables us to assign the origin of this difference to the absence ($m/n = 1$) or presence ($m/n = 2$) of spatially periodic, static oxygen octahedral rotation (OOR) inside the LMO layers. The experimental data analysis shows that the effective thermal conductance of the LMO/SMO interfaces strongly changes from $0.3$ GW/m$^2$K for $m/n = 2$ SLs with OOR to a surprisingly large value of $1.8$ GW/m$^2$K for $m/n = 1$ SLs without OOR. An instructive lattice dynamical model rationalizes our experimental findings as a result of coherent phonon transmission for $m/n = 1$ versus coherent phonon blocking in SLs with $m/n = 2$. We briefly discuss the possibilities to exploit these results for atomic-scale engineering of a crystalline phonon insulator. The thermal resistivity of this proposal for a thermal metamaterial surpasses the amorphous limit, although phonons still propagate coherently.

To develop effective thermal management strategies for GaN transistors, it is essential to accurately predict the device junction temperature. Since the width of the heat generation in the devices is comparable to phonon mean free paths of GaN, phonon ballistic transport exists and can significantly affect the heat transport process, which necessitates a thorough understanding of the influence of the phonon ballistic effects in GaN transistors. In this paper, the ballistic-diffusive phonon transport in GaN-on-SiC devices is examined by measuring the hotspot temperature using the thermoreflectance thermal imaging TTI combined with the hybrid phonon Monte Carlo-diffusion simulations. A series of Au heaters are fabricated on the top of the GaN layer to quantitatively mimic the different heat source distributions during device operation. The experimental and simulation results show a good consistency and both indicate that the phonon ballistic effects can significantly increase the hotspot temperature. With the size of the heat source decreasing, the errors of Fourier's law-based predictions increase, which emphasizes the necessity to carefully consider the phonon ballistic transport in device thermal simulations.

Inverse problems arise anywhere we have indirect measurement. As, in general they are ill-posed, to obtain satisfactory solutions for them needs prior knowledge. Classically, different regularization methods and Bayesian inference based methods have been proposed. As these methods need a great number of forward and backward computations, they become costly in computation, in particular, when the forward or generative models are complex and the evaluation of the likelihood becomes very costly. Using Deep Neural Network surrogate models and approximate computation can become very helpful. However, accounting for the uncertainties, we need first understand the Bayesian Deep Learning and then, we can see how we can use them for inverse problems. In this work, we focus on NN, DL and more specifically the Bayesian DL particularly adapted for inverse problems. We first give details of Bayesian DL approximate computations with exponential families, then we will see how we can use them for inverse problems. We consider two cases: First the case where the forward operator is known and used as physics constraint, the second more general data driven DL methods. keyword: Neural Network, Variational Bayesian inference, Bayesian Deep Learning (DL), Inverse problems, Physics based DL.

Measuring thermal properties of materials is not only of fundamental importance in understanding the transport processes of energy carriers (electrons and phonons) but also of practical interest in developing novel materials with desired thermal conductivity for applications in energy, electronics, and photonic systems. Over the past two decades, ultrafast laser-based time-domain thermoreflectance (TDTR) has emerged and evolved as a reliable, powerful, and versatile technique to measure the thermal properties of a wide range of bulk and thin film materials and their interfaces. This tutorial discusses the basics as well as the recent advances of the TDTR technique and its applications in the thermal characterization of a variety of materials. The tutorial begins with the fundamentals of the TDTR technique, serving as a guideline for understanding the basic principles of this technique. A diverse set of TDTR configurations that have been developed to meet different measurement conditions are then presented, followed by several variations of the TDTR technique that function similarly as the standard TDTR but with their own unique features. This tutorial closes with a summary that discusses the current limitations and proposes some directions for future development.

Hexagonal boron nitride (h-BN) has received great interest in recent years as a wide bandgap analog of graphene-derived systems. However, the thermal transport properties of h-BN, which can be critical for device reliability and functionality, are little studied both experimentally and theoretically. The primary challenge in the experimental measurements of the anisotropic thermal conductivity of h-BN is that typically sample size of h-BN single crystals is too small for conventional measurement techniques, as state-of-the-art technologies synthesize h-BN single crystals with lateral sizes only up to 2.5 mm and thickness up to 200 μm. Recently developed time-domain thermoreflectance (TDTR) techniques are suitable to measure the anisotropic thermal conductivity of such small samples, as it only requires a small area of 50x50 μm2 for the measurements. Accurate atomistic modeling of thermal transport in bulk h-BN is also challenging due to the highly anisotropic layered structure. Here we conduct an integrated experimental and theoretical study on the anisotropic thermal conductivity of bulk h-BN single crystals over the temperature range of 100 K to 500 K, using TDTR measurements with multiple modulation frequencies and a full-scale numerical calculation of the phonon Boltzmann transport equation starting from the first principles. Our experimental and numerical results compare favorably for both the in-plane and through-plane thermal conductivities. We observe unusual temperature-dependence and phonon-isotope scattering in the through-plane thermal conductivity of h-BN and elucidate their origins. This work not only provides an important benchmark of the anisotropic thermal conductivity of h-BN but also develops fundamental insights into the nature of phonon transport in this highly anisotropic layered material.

Super-aligned carbon nanotube (CNT) films have intriguing anisotropic thermal transport properties due to the anisotropic nature of individual nanotubes and the important role of nanotube alignment. However, the relationship between the alignment and the anisotropic thermal conductivities was not well understood due to the challenges in both the preparation of high-quality super-aligned CNT film samples and the thermal characterization of such highly anisotropic and porous thin films. Here, super-aligned CNT films with different alignment configurations are designed and their anisotropic thermal conductivities are measured using time-domain thermoreflectance (TDTR) with an elliptical-beam approach. The results suggest that the alignment configuration could tune the cross-plane thermal conductivity k_z from 6.4 to 1.5 W/mK and the in-plane anisotropic ratio from 1.2 to 13.5. This work confirms the important role of CNT alignment in tuning the thermal transport properties of super-aligned CNT films and provides an efficient way to design thermally anisotropic films for thermal management.

We present the time-dependent complete-active-space self-consistent-field (TD-CASSCF) method to simulate multielectron dynamics in ultrafast intense laser fields from the first principles. While based on multiconfiguration expansion, it divides the orbital space into frozen-core (tightly bound electrons with no response to the field), dynamical-core (electrons tightly bound but responding to the field), and active (fully correlated to describe highly excited and ejected electrons) orbital subspaces. The subspace decomposition can be done flexibly, conforming to phenomena under investigation and desired accuracy. The method is gauge invariant and size extensive. Infinite-range exterior complex scaling in addition to mask-function boundary is adopted as an efficient absorbing boundary. We show numerical examples and illustrate how to extract relevant physical quantities such as ionization yield, high-harmonic spectrum, and photoelectron spectrum from our full-dimensional implementation for atoms. The TD-CASSCF method will open a way to the ab initio simulation study of ultrafast intense laser science in realistic atoms and molecules.

Inverse problems are ubiquitous in nature, arising in almost all areas of science and engineering ranging from geophysics and climate science to astrophysics and biomechanics. One of the central challenges in solving inverse problems is tackling their ill-posed nature. Bayesian inference provides a principled approach for overcoming this by formulating the inverse problem into a statistical framework. However, it is challenging to apply when inferring fields that have discrete representations of large dimensions (the so-called "curse of dimensionality") and/or when prior information is available only in the form of previously acquired solutions. In this work, we present a novel method for efficient and accurate Bayesian inversion using deep generative models. Specifically, we demonstrate how using the approximate distribution learned by a Generative Adversarial Network (GAN) as a prior in a Bayesian update and reformulating the resulting inference problem in the low-dimensional latent space of the GAN, enables the efficient solution of large-scale Bayesian inverse problems. Our statistical framework preserves the underlying physics and is demonstrated to yield accurate results with reliable uncertainty estimates, even in the absence of information about underlying noise model, which is a significant challenge with many existing methods. We demonstrate the effectiveness of proposed method on a variety of inverse problems which include both synthetic as well as experimentally observed data.

Digital lasers control the laser beam by dynamically updating the phase patterns of the spatial light modulator (SLM) within the laser cavity. Due to the presence of nonlinear effects, such as mode competition and gain saturation in digital laser systems, it is often necessary to rely on specifically manually tailored approach or iteration processes to find suitable loaded phases in Digital lasers. This study proposes a model based on Conditional Generative Adversarial Networks (cGAN) and a modified U-Net architecture, with designed loss functions to inverse design the loaded phases. In this work, we employ deep neural networks to learn the nonlinear effects in simulated L-shape digital lasers, enabling the prediction of SLM-loaded phases for both analytical and non-analytical arbitrary structured light fields. The results demonstrate superior performance on non-analytical light fields compared to the current methods in L-shape Digital lasers. Furthermore, a transfer learning strategy is introduced, allowing knowledge obtained from one class of structured beams to be effectively reused for another, thereby enhancing generalization and improving performance under limited training data. Importantly, this method, the first proposed learning framework for digital lasers, is not limited to the L-shaped digital lasers discussed in this study, providing an efficient alternative for generating structured light in other digital laser systems.

Lead island films were obtained via vacuum vapor deposition on glass and ceramic substrates at 80 K. Electrical conductance was measured during vapor condensation and further annealing of the film up to room temperature. The resistance behavior during film formation and atomic force microscopy of annealed films were used as information sources about their structure. A model for the quenched growth, based on ballistic aggregation theory, was proposed. The nanostructure, responsible for chemiresistive properties of thin lead films and the mechanism of sensor response are discussed.

We investigated the thermal conductivity of surface plasmon polaritons (SPPs) propagating along thin Au and Ag films on a SiO$_2$ substrate with a Ti adhesive layer. To determine the propagation length and skin depth of SPPs along Au and Ag thin films, we numerically solved the dispersion relation while considering the size effect of the permittivity of metal. Additionally, we derived the spatial distribution of SPPs along the film thickness to analyze the effect of the Ti adhesive layer on the plasmon thermal conductivity of Au and Ag thin films. Our theoretical predictions revealed a decrease of approximately 30\% in plasmon thermal conductivity when considering the size effect of the permittivity of thin metal films. Furthermore, this causes the film thickness at which maximum thermal conductivity occurs to increase by about 30\%. Taking these factors into account, we calculated the optimal thickness of Au and Ag films, along with Ti adhesive layers, on SiO$_2$ substrates to be approximately 20 nm. By fabricating a sample with the optimal thickness of Au and Ag films, we experimentally demonstrated that the plasmon thermal conductivity of Au and Ag films can be as high as about 20\% of their electron contribution. This research will broaden the thermal design applications of ballistic thermal transport by SPPs propagating along thin metal coatings in microelectronics.

Due to their extremely long propagation lengths compared to the wavelengths, surface plasmon polaritons (SPPs) have been considered as a key in enhancing thermal conductivity in thin metal films. This study explores the conditions at which the plasmon thermal conductivity is maximized, considering the thickness-dependent metal permittivity. We derived the analytical solutions for the plasmon thermal conductivity in both the thin-film and thick-film limits to analyze the effect of the permittivities of metals and substrates. From the analytical solutions of plasmon thermal conductivity, we deduced that the plasmon thermal conductivity is proportional to the electron thermal conductivity based on the Wiedemann-Franz law. Additionally, we analyzed the conditions where the enhancement ratio of the thermal conductivity via SPPs is maximized. Metals with high plasma frequency and low damping coefficient are desirable for achieving the maximum plasmon thermal conductivity as well as the maximum enhancement ratio of thermal conductivity among metals. Significantly, 10-cm-long and 14-nm-thick Al film demonstrates most superior in-plane heat transfer via SPPs, showing a 53.5\% enhancement in thermal conductivity compared to its electron thermal counterpart on a lossless glass substrate.

Organic thin film materials with molecular ordering are gaining attention as they exhibit semiconductor characteristics. When using them for electronics, the thermal management becomes important, where heat dissipation is directional owning to the anisotropic thermal conductivity arising from the molecular ordering. However, it is difficult to evaluate the anisotropy by simultaneously measuring in-plane and cross-plane thermal conductivities of the film on a substrate, because the film is typically as thin as tens to hundreds of nanometers and its in-plane thermal conductivity is low. Here, we develop a novel bidirectional 3ω system that measures the anisotropic thermal conductivity of thin films by patterning two metal wires with different widths and preparing the films on top, and extracting the in-plane and cross-plane thermal conductivities using the difference in their sensitivities to the metal-wire width. Using the developed system, the thermal conductivity of spin-coated poly(3,4-ethylenedioxythiophene) olystyrene sulfonate (PEDOT:PSS) with thickness of 70 nm was successfully measured. The measured in-plane thermal conductivity of PEDOT:PSS film was as high as 2.9 W m-1 K-1 presumably due to the high structural ordering, giving anisotropy of 10. The calculations of measurement sensitivity to the film thickness and thermal conductivities suggest that the device can be applied to much thinner films by utilizing metal wires with smaller width.

We experimentally demonstrate a boosted in-plane thermal conduction by surface plasmon polaritons (SPPs) propagating along a thin Ti film on a glass substrate. Owing to a lossy nature of metal, SPPs can propagate over centimeter-scale distance even with a supported metal film, and resulting ballistic heat conduction can be quantitatively validated. Further, for a 100-nm-thick Ti film on glass substrate, a significant enhancement of in-plane thermal conductivity compared to bulk value ($\sim 35\%$) is experimentally shown. This study will provide a new avenue to employ SPPs for heat dissipation along a supported thin film, which can be readily applied to mitigate hot-spot issues in microelectronics.

Characterizing spatially varying thermal conductivities is significant to unveil the structure–property relation for a wide range of thermal functional materials such as chemical-vapor-deposited (CVD) diamonds, ion-irradiated materials, nuclear materials under radiation, and battery electrode materials. Although the development of thermal property microscopy based on time/frequency-domain thermoreflectance (TDTR/FDTR) enabled in-plane scanning of a thermal conductivity profile, measuring depth-dependent profiles remains challenging. This work proposed a machine-learning-based reconstruction method for extracting depth-dependent thermal conductivity [Formula: see text] directly from pump–probe phase signals. We demonstrated that the simple supervised-learning algorithm kernel ridge regression can reconstruct [Formula: see text] without requiring pre-knowledge about the functional form of the profile. The reconstruction method can not only accurately reproduce typical [Formula: see text] distributions such as the exponential profile of CVD diamonds and the Gaussian profile of ion-irradiated materials but also complex profiles artificially constructed by superimposing Gaussian, exponential, polynomial, and logarithmic functions. In addition to FDTR, the method also shows excellent performance of reconstructing [Formula: see text] of ion-irradiated semiconductors from TDTR signals. This work demonstrates that combining machine learning with pump–probe thermoreflectance is an effective way for depth-dependent thermal property mapping.

Lithium niobate (LiNbO3, LN) has unique electro‐optic and piezoelectric properties, making it widely used in optical devices, telecommunications, sensors, and acoustic systems. Thermal conductivity κ is a critical property influencing the performance and reliability of these applications. Point defects commonly exist in LN and can significantly reduce its κ. However, the effects of point defects on thermal transport in LN remain poorly understood. In this work, LN crystals are prepared through thermal reduction at 600–800 °C, inducing a depth‐dependent distribution of oxygen vacancies (VO) that increases in concentration with increasing reduction temperature. Time‐domain thermoreflectance and square‐pulsed source measurements reveal a significant suppression and a notable gradient in κ, attributed to the depth‐dependent distribution of VO. A machine learning potential with ab initio accuracy is developed to simulate the impact of typical point defects on thermal transport in LN, demonstrating that VO predominantly suppresses κ by affecting the transport of low‐frequency phonons below 6 THz. Notably, niobium vacancies and antisite defects exhibit similar effects, whereas lithium vacancies show minimal impact. This work highlights the dominant role of VO in modulating κ and provides insights into defect engineering for advanced LN‐based devices and similar ferroelectric crystals.

Accurate prediction of the thermal conductivity (κ) of polymers is generally challenging due to their complex structures. Currently available ab initio methods (e.g., DFT-BTE) are prohibitively expensive, and the classical force fields used in molecular dynamics lack accuracy. In this study, we combine ab initio hybrid machine learning (ML)/multipolar polarizable potential (i.e., PhyNEO) with ML-facilitated heat flux calculation. This approach provides reliable heat flux trajectories, which are then used to predict polymer κ quantitatively. Using poly(ethylene oxide) as an example, we compare our calculation results with reliable experimental reference obtained from time-domain thermoreflectance measurement, reaching excellent agreement. This work enables the quantitative prediction of bulk polymer κ starting from only small cluster quantum data, warranting broad applications in future.

Lattice thermal conductivity is a critical parameter for assessing the thermal transport properties of materials. When confined to the monolayer limit, two‐dimensional materials display unique thermal characteristics distinct from their three‐dimensional counterparts. This article first provides a concise summary of three widely applied experimental methodologies—Raman thermometry, suspended microbridge techniques, and time‐domain thermoreflectance—and their utility in validating theoretical predictions. It subsequently delves into recent advancements in theoretical modeling, encompassing both equilibrium and nonequilibrium molecular dynamics studies; first‐principles calculations grounded in the phonon Boltzmann transport equation that account for higher‐order scattering phenomena such as four‐phonon processes and phonon–electron interactions; emerging methods based on normal mode analysis for detailed phonon contribution decomposition; and novel approaches employing the Wigner transport equation to unify the description of phonon coherence and wave‐like heat transport phenomena beyond conventional theoretical frameworks. In addition, the advent of machine learning has expanded the scope of direct thermal conductivity prediction and the development of high‐precision interatomic potentials, paving the way for high‐throughput screening and extensive simulations. This review contrasts the advantages and drawbacks of these methodologies, identifies key challenges facing the field, and sketches future directions for 2D thermal transport research, emphasizing the integration of multiscale modeling, data‐driven innovation, and the synergy between experiments and theoretical insights.

Data science has become increasingly essential for the production of official statistics, as it enables the automated collection, processing, and analysis of large amounts of data. With such data science practices in place, it enables more timely, more insightful and more flexible reporting. However, the quality and integrity of data-science-driven statistics rely on the accuracy and reliability of the data sources and the machine learning techniques that support them. In particular, changes in data sources are inevitable to occur and pose significant risks that are crucial to address in the context of machine learning for official statistics. This paper gives an overview of the main risks, liabilities, and uncertainties associated with changing data sources in the context of machine learning for official statistics. We provide a checklist of the most prevalent origins and causes of changing data sources; not only on a technical level but also regarding ownership, ethics, regulation, and public perception. Next, we highlight the repercussions of changing data sources on statistical reporting. These include technical effects such as concept drift, bias, availability, validity, accuracy and completeness, but also the neutrality and potential discontinuation of the statistical offering. We offer a few important precautionary measures, such as enhancing robustness in both data sourcing and statistical techniques, and thorough monitoring. In doing so, machine learning-based official statistics can maintain integrity, reliability, consistency, and relevance in policy-making, decision-making, and public discourse.

Modern biology frequently relies on machine learning to provide predictions and improve decision processes. There have been recent calls for more scrutiny on machine learning performance and possible limitations. Here we present a set of community-wide recommendations aiming to help establish standards of supervised machine learning validation in biology. Adopting a structured methods description for machine learning based on data, optimization, model, evaluation (DOME) will aim to help both reviewers and readers to better understand and assess the performance and limitations of a method or outcome. The recommendations are formulated as questions to anyone wishing to pursue implementation of a machine learning algorithm. Answers to these questions can be easily included in the supplementary material of published papers.

Learning curves are a concept from social sciences that has been adopted in the context of machine learning to assess the performance of a learning algorithm with respect to a certain resource, e.g., the number of training examples or the number of training iterations. Learning curves have important applications in several machine learning contexts, most notably in data acquisition, early stopping of model training, and model selection. For instance, learning curves can be used to model the performance of the combination of an algorithm and its hyperparameter configuration, providing insights into their potential suitability at an early stage and often expediting the algorithm selection process. Various learning curve models have been proposed to use learning curves for decision making. Some of these models answer the binary decision question of whether a given algorithm at a certain budget will outperform a certain reference performance, whereas more complex models predict the entire learning curve of an algorithm. We contribute a framework that categorises learning curve approaches using three criteria: the decision-making situation they address, the intrinsic learning curve question they answer and the type of resources they use. We survey papers from the literature and classify them into this framework.

Online active learning is a paradigm in machine learning that aims to select the most informative data points to label from a data stream. The problem of minimizing the cost associated with collecting labeled observations has gained a lot of attention in recent years, particularly in real-world applications where data is only available in an unlabeled form. Annotating each observation can be time-consuming and costly, making it difficult to obtain large amounts of labeled data. To overcome this issue, many active learning strategies have been proposed in the last decades, aiming to select the most informative observations for labeling in order to improve the performance of machine learning models. These approaches can be broadly divided into two categories: static pool-based and stream-based active learning. Pool-based active learning involves selecting a subset of observations from a closed pool of unlabeled data, and it has been the focus of many surveys and literature reviews. However, the growing availability of data streams has led to an increase in the number of approaches that focus on online active learning, which involves continuously selecting and labeling observations as they arrive in a stream. This work aims to provide an overview of the most recently proposed approaches for selecting the most informative observations from data streams in real time. We review the various techniques that have been proposed and discuss their strengths and limitations, as well as the challenges and opportunities that exist in this area of research.

The ability to explain decisions made by machine learning models remains one of the most significant hurdles towards widespread adoption of AI in highly sensitive areas such as medicine, cybersecurity or autonomous driving. Great interest exists in understanding which features of the input data prompt model decision making. In this contribution, we propose a novel approach to identify relevant features of the input data, inspired by methods from the energy landscapes field, developed in the physical sciences. By identifying conserved weights within groups of minima of the loss landscapes, we can identify the drivers of model decision making. Analogues to this idea exist in the molecular sciences, where coordinate invariants or order parameters are employed to identify critical features of a molecule. However, no such approach exists for machine learning loss landscapes. We will demonstrate the applicability of energy landscape methods to machine learning models and give examples, both synthetic and from the real world, for how these methods can help to make models more interpretable.

Machine Learning (ML) has recently shown tremendous success in modeling various healthcare prediction tasks, ranging from disease diagnosis and prognosis to patient treatment. Due to the sensitive nature of medical data, privacy must be considered along the entire ML pipeline, from model training to inference. In this paper, we conduct a review of recent literature concerning Privacy-Preserving Machine Learning (PPML) for healthcare. We primarily focus on privacy-preserving training and inference-as-a-service, and perform a comprehensive review of existing trends, identify challenges, and discuss opportunities for future research directions. The aim of this review is to guide the development of private and efficient ML models in healthcare, with the prospects of translating research efforts into real-world settings.

The proliferation of fake news and its propagation on social media has become a major concern due to its ability to create devastating impacts. Different machine learning approaches have been suggested to detect fake news. However, most of those focused on a specific type of news (such as political) which leads us to the question of dataset-bias of the models used. In this research, we conducted a benchmark study to assess the performance of different applicable machine learning approaches on three different datasets where we accumulated the largest and most diversified one. We explored a number of advanced pre-trained language models for fake news detection along with the traditional and deep learning ones and compared their performances from different aspects for the first time to the best of our knowledge. We find that BERT and similar pre-trained models perform the best for fake news detection, especially with very small dataset. Hence, these models are significantly better option for languages with limited electronic contents, i.e., training data. We also carried out several analysis based on the models' performance, article's topic, article's length, and discussed different lessons learned from them. We believe that this benchmark study will help the research community to explore further and news sites/blogs to select the most appropriate fake news detection method.

This article provides the first survey of computational models of emotion in reinforcement learning (RL) agents. The survey focuses on agent/robot emotions, and mostly ignores human user emotions. Emotions are recognized as functional in decision-making by influencing motivation and action selection. Therefore, computational emotion models are usually grounded in the agent's decision making architecture, of which RL is an important subclass. Studying emotions in RL-based agents is useful for three research fields. For machine learning (ML) researchers, emotion models may improve learning efficiency. For the interactive ML and human-robot interaction (HRI) community, emotions can communicate state and enhance user investment. Lastly, it allows affective modelling (AM) researchers to investigate their emotion theories in a successful AI agent class. This survey provides background on emotion theory and RL. It systematically addresses 1) from what underlying dimensions (e.g., homeostasis, appraisal) emotions can be derived and how these can be modelled in RL-agents, 2) what types of emotions have been derived from these dimensions, and 3) how these emotions may either influence the learning efficiency of the agent or be useful as social signals. We also systematically compare evaluation criteria, and draw connections to important RL sub-domains like (intrinsic) motivation and model-based RL. In short, this survey provides both a practical overview for engineers wanting to implement emotions in their RL agents, and identifies challenges and directions for future emotion-RL research.

We introduce the Fourier Learning Machine (FLM), a neural network (NN) architecture designed to represent a multidimensional nonharmonic Fourier series. The FLM uses a simple feedforward structure with cosine activation functions to learn the frequencies, amplitudes, and phase shifts of the series as trainable parameters. This design allows the model to create a problem-specific spectral basis adaptable to both periodic and nonperiodic functions. Unlike previous Fourier-inspired NN models, the FLM is the first architecture able to represent a multidimensional Fourier series with a complete set of basis functions in separable form, doing so by using a standard Multilayer Perceptron-like architecture. A one-to-one correspondence between the Fourier coefficients and amplitudes and phase-shifts is demonstrated, allowing for the translation between a full, separable basis form and the cosine phase-shifted one. Additionally, we evaluate the performance of FLMs on several scientific computing problems, including benchmark Partial Differential Equations (PDEs) and a family of Optimal Control Problems (OCPs). Computational experiments show that the performance of FLMs is comparable, and often superior, to that of established architectures like SIREN and vanilla feedforward NNs.

Efficient approximation lies at the heart of large-scale machine learning problems. In this paper, we propose a novel, robust maximum entropy algorithm, which is capable of dealing with hundreds of moments and allows for computationally efficient approximations. We showcase the usefulness of the proposed method, its equivalence to constrained Bayesian variational inference and demonstrate its superiority over existing approaches in two applications, namely, fast log determinant estimation and information-theoretic Bayesian optimisation.

This research paper delves into the innovative integration of Shannon entropy and rough set theory, presenting a novel approach to generalize the evaluation approach in machine learning. The conventional application of entropy, primarily focused on information uncertainty, is extended through its combination with rough set theory to offer a deeper insight into data's intrinsic structure and the interpretability of machine learning models. We introduce a comprehensive framework that synergizes the granularity of rough set theory with the uncertainty quantification of Shannon entropy, applied across a spectrum of machine learning algorithms. Our methodology is rigorously tested on various datasets, showcasing its capability to not only assess predictive performance but also to illuminate the underlying data complexity and model robustness. The results underscore the utility of this integrated approach in enhancing the evaluation landscape of machine learning, offering a multi-faceted perspective that balances accuracy with a profound understanding of data attributes and model dynamics. This paper contributes a groundbreaking perspective to machine learning evaluation, proposing a method that encapsulates a holistic view of model performance, thereby facilitating more informed decision-making in model selection and application.

We seek to enable classic processing of continuous ultra-sparse spatiotemporal data generated by event-based sensors with dense machine learning models. We propose a novel hybrid pipeline composed of asynchronous sensing and synchronous processing that combines several ideas: (1) an embedding based on PointNet models -- the ALERT module -- that can continuously integrate new and dismiss old events thanks to a leakage mechanism, (2) a flexible readout of the embedded data that allows to feed any downstream model with always up-to-date features at any sampling rate, (3) exploiting the input sparsity in a patch-based approach inspired by Vision Transformer to optimize the efficiency of the method. These embeddings are then processed by a transformer model trained for object and gesture recognition. Using this approach, we achieve performances at the state-of-the-art with a lower latency than competitors. We also demonstrate that our asynchronous model can operate at any desired sampling rate.

We propose unsupervised representation learning and feature extraction from dendrograms. The commonly used Minimax distance measures correspond to building a dendrogram with single linkage criterion, with defining specific forms of a level function and a distance function over that. Therefore, we extend this method to arbitrary dendrograms. We develop a generalized framework wherein different distance measures and representations can be inferred from different types of dendrograms, level functions and distance functions. Via an appropriate embedding, we compute a vector-based representation of the inferred distances, in order to enable many numerical machine learning algorithms to employ such distances. Then, to address the model selection problem, we study the aggregation of different dendrogram-based distances respectively in solution space and in representation space in the spirit of deep representations. In the first approach, for example for the clustering problem, we build a graph with positive and negative edge weights according to the consistency of the clustering labels of different objects among different solutions, in the context of ensemble methods. Then, we use an efficient variant of correlation clustering to produce the final clusters. In the second approach, we investigate the combination of different distances and features sequentially in the spirit of multi-layered architectures to obtain the final features. Finally, we demonstrate the effectiveness of our approach via several numerical studies.

International economics has a long history of improving our understanding of factors causing trade, and the consequences of free flow of goods and services across countries. The recent shocks to the free trade regime, especially trade disputes among major economies, as well as black swan events, such as trade wars and pandemics, raise the need for improved predictions to inform policy decisions. AI methods are allowing economists to solve such prediction problems in new ways. In this manuscript, we present novel methods that predict and associate food and agricultural commodities traded internationally. Association Rules (AR) analysis has been deployed successfully for economic scenarios at the consumer or store level, such as for market basket analysis. In our work however, we present analysis of imports and exports associations and their effects on commodity trade flows. Moreover, Ensemble Machine Learning methods are developed to provide improved agricultural trade predictions, outlier events' implications, and quantitative pointers to policy makers.