The glass transition temperature (Tg) of alkali borate glasses is strongly composition-dependent and difficult to predict from first principles due to the structural complexity of the boron network. Here, we apply physics-informed symbolic regression (combining evolutionary search with physically meaningful descriptors) to derive an interpretable closed-form expression for Tg for the xM2O⋅(100−x)B2O3 glass family, with M = Li, Na and K, and x expressed in mol%, followed by extrapolation to M = Rb and Cs. The resulting model achieves a root mean square error of 14–16 K while maintaining clear physical interpretability, explicitly capturing the interplay among Tg, structural dissociation energy, and network packing. Critically, models built on the Rigid Unit Packing Fraction (RUPF) yield substantially more realistic Tg predictions than those using the conventional Atomic Packing Fraction (APF), as APF overestimates structural rigidity at intermediate compositions. The fitted dissociation energies are further validated against the revised Makishima–Mackenzie model, confirming that the inferred parameters are physically consistent, not merely statistically effective, within the alkali borate family. Finally, Monte Carlo uncertainty quantification reveals that prediction uncertainty is highest in the compositional regions associated with the boron anomaly, directly linking model limitations to a known structural transition in these glasses. This result highlights the potential of physics-informed symbolic regression as a transparent and interpretable alternative to black-box models for property prediction in glass systems.
Tests are essential for ensuring code quality in software development. However, poor implementation practices can compromise the maintainability and evolution of test code, leading to additional costs and effort. These practices, known as test smells, have been the focus of numerous studies that expanded the catalog of known test smells. Despite this progress, little attention has been given to examining test smells from the perspective of programming languages, particularly their relationship with specific languages or test frameworks. As a result, researchers and practitioners have proposed various approaches and tools to detect and refactor the same test smells across different languages. In this work, we address this gap by analyzing test smells in the literature from a programming language perspective. We conducted a systematic mapping study, gathering data from papers published up to August 2025, examining 117 papers. We identified the most prevalent languages and frameworks associated with test smells, the most common test smells in each language, refactorings for addressing test smells, datasets of test smells, the criticality of specific test smells, and a collection of tools to detect or refactor test smells for various languages and frameworks. Our results show that while many test smells have been addressed, research remains limited to a few languages and frameworks. We also found datasets and detection/refactoring tools for some languages and analyzed the criticality of frequently mentioned test smells. This study highlights the need to expand research to more languages and develop more universal solutions.
Background Climate change disproportionately affects socially minoritized populations, but migrants' vulnerability to extreme temperatures remains overlooked. We estimated short-term associations between extreme ambient temperatures and mortality among migrants and non-migrants across Brazil. Methods We linked the 100 Million Brazilian Cohort to mortality and temperature data (2000–2020) and conducted a space-time-stratified case-crossover study analysing the differential effects of extreme temperature on mortality among internal migrants (i.e., Individuals whose state of birth was different from their residence at registration or death) and non-migrants. We used conditional quasi-Poisson regression with distributed lag non-linear model (DLNM) to estimate relative risks (RRs) and 95% confidence intervals (CIs) for all-cause non-accidental and cause-specific mortality associated with extreme-cold (1st percentile) and extreme-heat (99th percentile) compared to minimum-mortality-temperature (i.e., the temperature associated with the lowest mortality risk), over 0–20 days in both groups. We further stratified by age, literacy, urbanicity and housing conditions. Findings We included 1,199,636 deaths among internal-migrants (mean age at death = 62.67 years (standard-deviation (SD) = 18.36), with 673,166 (56%) males) and 3,716,900 deaths among non-migrants (57.18 years (SD = 20.31), with 2,135,293 (57%) males). Over 0–20 days, we observed a consistent pattern of higher risks of all-cause, non-accidental mortality among internal-migrants than non-migrants following both extreme-heat (RRMig(95%CI):1.05(1.01–1.10) vs. RRNon-mig:1.01(0.99–1.04)) and extreme-cold (RRMig:1.27(1.22–1.32) vs. RRNon-mig:1.24(1.20–1.28)). Extreme cold-related circulatory mortality (mortality due to diseases of the circulatory system, based on the relevant International Classification of Diseases codes) (I00–I99) was higher among non-migrants (RRMig:1.32(1.20–1.45) vs. RRNon-mig:1.41(1.31–1.51)) while heat-related respiratory mortality (J00–J99) was higher among internal-migrants (RRMig:1.32(1.10–1.58) vs. RRNon-mig:1.06(1.00–1.17)), particularly those illiterate (RRMig:1.84(1.25–2.71) vs. RRNon-mig:1.10(0.89–1.36)). Other subcategories showed greater 95%CI overlap. Interpretation We found internal migrants are likely to experience higher heat-related respiratory mortality than non-migrants particularly low-educated migrants. Our results highlight the need to further consider migrants for targeted climate adaptation policies, with a focus on reducing social inequalities in the place of arrival. Funding The Wellcome Trust.
The magnon dispersion spectra, as well as the reflectance and transmittance of nanojunctions composed of fcc Fe-Ni alloys sandwiched between Ni(001) leads at Fe concentrations of 20% and 30% are investigated. First-principles calculations based on relativistic density functional theory and the Korringa-Kohn-Rostoker (KKR) Green's function method are employed to determine the magnetic exchange constants between pairs of nonequivalent atoms. Phase-field matching theory (PFMT) is applied to match spin waves incident on the bulk Ni layers with those propagating within the Fe-Ni alloy surface of the nanojunction. In contrast to previous models that considered only one non-equivalent Ni atom per Nilayer, the present approach includes all non-equivalent atoms across all layers of the nanojunction, yielding 26 and 42 magnon dispersion branches for c = 0.2 and c = 0.3, respectively. Magnonic dispersion spectra calculated across the first Brillouin zone reveal the presence of non-propagating, evanescent waves, which, depending on the propagation orientation, can dominate the highest magnonic mode or affect both the lowest and highest modes within the nanojunctions. Local density of states (LDOS) analysis indicates that Fe atoms dominate high-frequency magnonic modes, whereas Ni atoms dominate low-frequency modes. For the c = 0.3 case, the system exhibits threshold-induced localization via two high-frequency Ni peaks: a quasi-bound resonance just below the leads' propagation threshold and a true bound, evanescent state above it, arising from total internal reflection. These states are absent at c = 0.2, indicating that a critical iron concentration is required to sufficiently enhance magnetic exchange. Using the Landauer-B & uuml;ttiker formalism for distinct incidence directions, the reflectance and transmittance spectra arising from the interactions of incident spin waves with the nanojunctions' magnonic modes are analyzed. The results highlight the presence of resonance-assisted maxima of both the Fabry-Perot and Fano types. Compared with the previous simpler model, the present model identifies transmission states that were previously inaccessible and provides a more robust framework for analyzing spin transport across complex magnetic interfaces.. These findings can be crucial for the experimental selection of suitable nanojunctions and incident spin-wave directions for magnonic devices and spin-wave filters.
The exponential growth of global data traffic has intensified the adoption of multi-band transmission (MBT) in elastic optical networks (EONs). However, defining the worst-case quality of transmission (QoT) for safety-margin dimensioning remains challenging because heterogeneous spectral configurations and the coupled effects of inter-channel stimulated Raman scattering (ISRS) and nonlinear interference (NLI) make maximum occupancy correspond to many distinct spectral configurations rather than a single one. This paper proposes a methodology to identify worst-case generalized signal-to-noise ratio (GSNR) configurations in C+L+S EONs. The Per-Band Homogeneous (Hom. Band) spectral-filling strategy is evaluated against a stochastic incremental-allocation model and a genetic algorithm (GA) that searches for near-worst-case conditions. Simulations over an 80km standard single-mode fiber (SSMF) span using the GNPy library show that the conventional assumption of uniformly filling the spectrum with noise-like 50GHz channels does not bound the sampled stochastic worst-case GSNR at any of the nine evaluated channel positions. In contrast, the Hom. Band strategy remains conservative relative to all sampled stochastic configurations. The GA finds configurations with GSNR at most 0.075dB below the Hom. Band estimate, a difference that is small relative to practical engineering margins; Hom. Band therefore provides an empirical conservative estimate of worst-case GSNR and a near-worst-case predictor. Under the single-span scenario, worst-case GSNR is primarily limited by amplified spontaneous emission (ASE) noise and global Raman tilt. Requiring only a few thousand simulations per channel under test (CUT), the Hom. Band method provides an efficient predictor for worst-case dimensioning of single-span multi-band EONs.
