2026-05-14 | | Total: 35
Despite extensive research on chirality-driven spin selectivity, most studies have focused on static magnetic properties, while the influence of chirality on the dynamic magnetic response remains largely unexplored. Here, we investigate how chiral molecular interfaces affect magnetization dynamics in thin Co/Ni multilayers with perpendicular magnetic anisotropy using broadband ferromagnetic resonance spectroscopy. A comparison between bare (reference) films and molecule-functionalized (hybrid) samples reveals no measurable changes in either the resonance field or the linewidth that could be attributed to the presence of the chiral environment. Motivated by our findings we develop a macrospin description that distinguishes equilibrium modifications of the magnetic free-energy landscape (MIPAC-type effects) from non-equilibrium, CISS-induced spin torques. Our analysis shows that equilibrium modifications primarily shift the resonance condition via changes to the free energy landscape and thereby the effective field, whereas damping-like non-equilibrium torques provide a distinct channel for varying the effective damping rate. This approach establishes clear criteria for disentangling chiral-interface-induced energy modifications from torque-driven dynamical effects in ferromagnetic resonance experiments.
Optical spin-orbit interactions (SOI) link photonic spin to momentum, offering a route toward on-chip polarization control and beam steering. Nevertheless, achieving sufficient optical SOI and nonlinearities on sub-micrometer scales - a prerequisite for dense photonic integration - remains an outstanding challenge. Here, we show that highly birefringent van der Waals (vdW) waveguides provide an ideal, chip-compatible platform to address this limitation. We focus on the ferroelectric semiconductor NbOI2, which exhibits record optical nonlinearities and dielectric anisotropy. Using femtosecond optical microscopy, we image light propagation and harmonic conversion beyond the total internal reflection barrier over tens of micrometers in NbOI2 slab waveguides. We report giant optical spin-splitting through the optical spin Hall effect, which facilitates spatial separation of optical spin currents on sub-micrometer scales, in quantitative agreement with a microscopic light-matter interaction model. We further leverage optical spin-momentum locking to realize polarization-controlled waveguide steering. We generalize these observations across various vdW waveguides and empirically confirm a scaling law linking dielectric anisotropy to geometric spin-splitting. Our results establish highly anisotropic vdW waveguides as an ideal platform for densely integrated opto-spintronic technologies.
Fluctuation formulas for elastic and viscoelastic moduli allow their computation from equilibrium molecular dynamics simulations, avoiding explicit nonequilibrium deformation protocols. While such expressions are well established for the quasi-static moduli, and also the linear dynamic moduli, no fluctuation formula exists for the nonlinear time-dependent moduli that govern anharmonic viscoelastic response under finite time-dependent strains. In this work we derive transient-time correlation function expressions for both the linear and the nonlinear dynamic moduli, starting from the DOLLS/SLLOD equations of motion for irrotational motion. The resulting formulas involve equilibrium time correlations of the stress tensor and Born-kinetic terms, and they recover the known quasi-static and linear dynamic results in the appropriate limits.
High-pressure X-ray diffraction experiments were performed on Ti-Al-V alloys to investigate the effect of composition on structural stability, focusing on Ti-3Al-2.5V and comparing with pure Titanium and Ti-6Al-4V. Measurements using different pressure-transmitting media show a phase transition in Ti-3Al-2.5V at 17-19 GPa, intermediate between pure Ti (5-10 GPa) and Ti-6Al-4V (~30 GPa). Despite variations arising from the choice of pressure medium, the transition pressure shows a clear and systematic increase with higher Al and V content. Equation-of-state analysis indicates that the bulk modulus remains nearly unchanged across compositions. This suggests a decoupling between elastic properties and phase stability, with alloying primarily affecting the transition pressure rather than compressibility. These results highlight the role of composition in tuning high-pressure phase transformations in Ti-Al-V based alloys.
In light of the rise of transition metal dichalcogenides as 2D semiconductors for device applications, band engineering becomes very important from an application point of view. In many of these materials, such as the canonical example of MoS$_{2}$, the semiconductor band gap depends on the layer number. It changes from indirect to direct as it evolves from a bulk semiconductor to a monolayer. Interestingly, it was predicted and experimentally confirmed that, by thinning the material from bulk to a bilayer, the indirect transition shows a strong blue-shift. Here, we present the results of scanning tunnelling spectroscopy measurements on MoS$_{2}$ that has been grown \textit{in situ} via molecular beam epitaxy on graphene on Ir(111) at thicknesses ranging from 1 to 5 layers. We find a drastic decrease of the band gap with increasing layer number, to values even below the band gap in bulk. We also observe that the pinning of the conduction band vanishes above 4 layers. Comparing our experimental data with density functional theory and \textit{GW} calculations indicates that an additional screening is introduced by the sample growth conditions.
The Materials Genome Initiative catalyzed the proliferation of centralized platforms--SaaS, PaaS, and IaaS--that aggregate computational and experimental resources for accelerated materials discovery. In parallel, breakthroughs in large language models (LLMs) and autonomous agents have created powerful new reasoning capabilities for scientific research. Yet a critical "last mile" problem remains: while we possess world-class models and vast repositories of materials data, we lack the organizational infrastructure to compose these capabilities securely across institutional boundaries. The development of structural and functional materials for harsh service environments--high-temperature alloys, radiation resistant steels, corrosion-resistant coatings--remains characterized by long-term iteration, mechanistic complexity, and high domain expertise--demands that exceed both monolithic agent systems and traditional centralized platforms. To address this gap we propose OpenAaaS, an open-source hierarchical and distributed Agent-as-a-Service framework that enables organized multi-agent collaboration for intelligent materials design. OpenAaaS is built on a single foundational principle: code flows, data stays still. A Master Agent plans and decomposes complex research tasks without requiring direct access to subordinate agents' managed data and computational resources. Sub-agents, deployed as near-data execution nodes, retain full sovereignty over local datasets, proprietary algorithms, and specialized hardware. This architecture guarantees that raw data never leaves its domain of origin while enabling cross-scale, cross-domain secure integration of previously isolated materials intelligence silos. We validate the framework through two representative case studies: (i) AlphaAgent, an evidence-grounded materials literature analysis executor that achieves 4.66/5.0 on deep analytical questions against single-pass RAG baselines; and (ii) an ultra-large-scale hexa-high-entropy alloy descriptor database service that demonstrates secure near-data execution and domain-specific scientific workflows under strict data-sovereignty constraints. OpenAaaS establishes a principled pathway toward "organized research" via agent collectives, offering a scalable foundation for next-generation materials intelligent design platforms. All source code is available at https://github.com/Wolido/OpenAaaS.
This work presents a systematic study of the magnetocaloric effect in the nanostructured perovskite series $La_{0.6}Sr_{0.4}Fe_{1-x}Co_{x}O_3$ (x = 0, 0.2, 0.5, 0.8, and 1.0), synthesized by a pore-wetting method using polymeric membranes with pore diameters of 200 nm and 800 nm. All samples were calcined at 1000°C. Structural characterization was made by X-ray diffraction and confirmed the formation of a single-phase perovskite with distorted rhombohedral symmetry, without detectable secondary phases. We observed significant influence of substitution of Fe by Co on the morphology, as the analysis by scanning electron microscopy revealed a clear evolution from smaller to larger particles and from thin to thicker nanotubes, as the Co content increased. Magnetic measurements showed that the cationic substitution enhances ferromagnetic coupling, increasing both the saturation magnetization (MS) and the Curie temperature (TC). The magnetocaloric properties, determined through the Maxwell relations, exhibit a maximum entropy change of 1.13 J/(kg K) under an applied field of 3 T for the sample with x = 1. These results demonstrate that the combination of Co doping and controlled nanostructuring effectively optimizes the magnetocaloric response.
Dopant distribution in two-dimensional semiconductors is typically assumed to be stochastic, limiting deterministic defect engineering. Here, we show that non-equilibrium growth kinetics can be harnessed to define dopant-driven strain architectures in vanadium-doped WS$_2$ monolayers. Using synchrotron X-ray fluorescence, we identify preferential vanadium incorporation, anti-correlated with tungsten content, along crystallographic bisectors. An adsorption-growth-diffusion model with a single kinetic parameter quantitatively captures the dopant segregation arising from preferential corner adsorption and limited diffusion during chemical vapor deposition growth. Hyperspectral Raman imaging demonstrates mechanically induced vibrational responses, revealing localized tensile strain ($\varepsilon \approx0.70\%$) channels associated with the anisotropic dopant distribution. This regime is marked by the depletion of W-site-sensitive in-plane modes and the emergence of a localized $J2$ mode (210~cm$^{-1}$), which our ab-initio calculations attribute to antiphase V$-$V oscillations. These findings establish kinetic segregation as a route to deterministic chemical and strain architectures in 2D semiconductors, enabling programmable defect landscapes and strain engineering during synthesis.
RuO$_2$ has been regarded as a prototypical candidate for metallic altermagnet, offering a potential platform for high-speed and high-efficiency spintronics. However, the magnetic ground state of RuO$_2$ remains a topic of active debate due to conflicting experimental reports. In this work, we investigate the effect of electron correlations in RuO$_2$ using density functional theory combined with dynamical mean-field theory (DFT+DMFT). In contrast to previous DFT-based studies, DFT+DMFT captures essential dynamical correlation effects, yielding spectral functions and optical conductivities in excellent quantitative agreement with experiments, and further reveals that RuO$_2$ resides in the close vicinity of both the paramagnetic-altermagnetic phase boundary and the itinerant-localized crossover, rendering the magnetic ground state highly susceptible to external perturbations. Indeed, even a minimal compressive strain of $\sim$0.5% is sufficient to drive the system into an altermagnetic phase. These findings elucidate the origin of the conflicting experimental observations and reveal that dynamical correlation effects are the key driving force behind the highly tunable magnetic ground state of RuO$_2$.
Further development of nuclear power plant technology relies heavily on materials durability under operating conditions. Estimating the materials performance in the operando tests is crucial. In this paper, the mechanical behavior of thin amorphous nuclear-dedicated Al2O3 coatings deposited by pulsed laser deposition was investigated by nanoindentation over the temperature range of 25-650C. Experimental nanomechanical analysis was supported by MD simulations. The results indicate that the hardness of the amorphous coating experiences a gradual, constant decrease with temperature, while the Young modulus value remains constant in the whole temperature range. Observed phenomena confirm the increasing plasticity of the material and it is postulated to be related to the bond-switching mechanism that accelerates at high temperatures. The post-mortem transmission electron microscopy characterization confirmed that the loaded material was non-crystalline over the entire range of the indentation temperatures. The thermal stability of the structure was further studied in-situ up to 1050C by X-ray diffraction. The implemented methodology allowed us to follow the dynamic process of phase transitions occurring in the material above 650C. First, thermally activated crystallization was observed at 700C. Intermediate alumina phases were present up to 950C, while above this temperature, exclusively the thermodynamically stable alpha-Al2O3 was observed. The in-situ high-temperature characterization of the evolution of thin films boosts the understanding of the application limits of the coating systems at elevated temperatures. The added value is that the paper demonstrates the potential usefulness of combining high-temperature techniques to characterize the complete behavior of thin films at elevated temperatures.
Charge density wave (CDW) formation in two-dimensional materials is governed by complex competing lattice instabilities that remain incompletely understood. Here, we investigate the structural evolution of monolayer $\mathrm{1T-VSe}_{2}$ using first-principles electronic and phonon calculations. The pristine phase exhibits several imaginary-frequency phonon modes associated with dominant instability wave vectors $\mathrm{Q}_{CDW}$, which generate the first-generation CDW phases. Subsequent phonon analyses reveal that several of these intermediate structures remain dynamically unstable and undergo further symmetry-lowering distortions into larger superstructures. Through iterative phonon-driven relaxations, we identify multiple transformation pathways that converge toward the same low-energy $2\sqrt{3}\times4$ CDW configuration. Although these pathways originate from distinct intermediate CDW states, they ultimately reach nearly degenerate energetically stable phases, demonstrating that different phonon-driven routes can lead to the same ground-state configuration. The results establish a unified phonon-driven cascade mechanism for hierarchical CDW formation in monolayer $\mathrm{1T-VSe}_{2}$ and provide a systematic framework for understanding competing ordered phases in low-dimensional quantum materials.
Yttrium iron garnet (YIG) is a ferrimagnetic insulator valued for its high Curie temperature, very low magnetic damping, and ability to support long-range spin-wave transport. These qualities have established it as a cornerstone material in the field of spintronics and magnonics. Most studies on YIG so far have been focused on bulk crystals, thin films, and nanoparticles, including variants with substitutions at the yttrium or iron site. New morphologies such as sub-micron flakes have drawn interest recently as their geometry and mechanical flexibility might enable different device architectures. However, detailed investigations combining their electronic structure and magnetic behavior remain scarce. In this work, we present a comparative study of the electronic and magnetic properties of a bulk-like YIG film and a sub-micron-sized YIG flake. Our results highlight the distinct behavior that emerges in sub-micron dimensions and point toward future uses for flake-based YIG in compact spintronics devices.
Understanding defect evolution is essential to the structural stability of quasicrystals, yet the kinetics of defect repair remain poorly understood. Here, by combining the string method and the spring pair method, we determine the minimum energy path from defective to defect-free dodecagonal quasicrystals using a particle model with the Lennard-Jones-Gauss potential. We find that defect annihilation proceeds via three stages: phason flip, aggregation and decomposition of shield-like defects. These sequential transformations are driven by potential energy gradients and accompanied by an increase in structural symmetry. The three stages act synergistically in promoting defect annihilation, offering new insights into the microscopic repair mechanisms of quasicrystals.
A recently developed approach, conditional probability density functional theory (CP-DFT), yields direct access to the exchange-correlation hole of a system, an important correlation function that is not available from any standard DFT calculation. We present the first results for extended materials with periodic boundary conditions. We demonstrate that CP-DFT works on weakly correlated materials (Na, Si). When applied to the prototypical Kagome material $CsV_3Sb_5$, we find $d$-orbital correlations that are not captured by standard DFT. Such distribution leads to a positive finding probability between two separated electrons and an enhanced charge density wave signal, suggesting a useful approach for strongly correlated systems.
A strategy to overcome sluggish hydrogenation/dehydrogenation of magnesium is demonstrated by creating magnesium-graphene interphase boundaries via high-pressure torsion (HPT). HPT reduces the grain size of pure magnesium from 1 mm to 850 nm, with 70% of grain boundaries having high misorientation angles. Graphene addition leads to even finer grain sizes of 10-500 nm with a bimodal morphology. The magnesium-graphene composites exhibit superior kinetics at 623 K while maintaining high air resistance. Kinetic modeling reveals that the rate-controlling mechanism transits from interfacial reaction in coarse-grained magnesium to atomic diffusion in magnesium-graphene nanocomposites. Kissinger analysis shows that the activation energy for hydrogen desorption remains unchanged at 145 +/- 2 kJ/mol, regardless of the presence of grain or interphase boundaries. However, the frequency factor (number of successful attempts to overcome the activation energy) increases with the generation of interfaces, which serve as sites for hydrogen diffusion and heterogeneous metal/hydride nucleation. These findings highlight the impact of interphase boundary engineering via severe plastic deformation for enhancing the kinetics and air resistance of hydrogen storage materials.
Unlike cubic GeSn, which requires a high Sn concentration to undergo an indirect-to-direct bandgap transition, lonsdaleite (2H) germanium is an intrinsic direct-gap semiconductor. We employ first-principles density functional theory to investigate the structural, electronic, and optical properties of 2H-Ge$_{1-x}$Sn$_{x}$ random alloys in the dilute Sn regime ($x \le 0.10$). The extended alloy disorder is modeled using 48-atom special quasirandom structure (SQS) supercells, and the coherent effective band structure is recovered via spectral band unfolding. We show that 2H-Ge$_{1-x}$Sn$_{x}$ maintains a direct bandgap at the $Γ$ point across the studied composition range, exhibiting a strong bandgap bowing that shifts the fundamental absorption edge into the mid-infrared. Evaluation of the optical transition matrix elements reveals a giant polarization anisotropy dictated by spin-orbit coupling. The fundamental transition is strongly dipole-allowed for light polarized perpendicular to the crystal $c$-axis, an optical selection rule that is robustly preserved despite the random alloy disorder breaking the symmetry. These results demonstrate that hexagonal GeSn bypasses the compositional threshold limitations of the cubic phase, providing a highly tunable direct-gap system for infrared optoelectronics.
The extremely low solubility of helium in liquid metals may lead to rapid supersaturation, promoting spontaneous formation of helium bubbles by nucleation. Once nucleated, the stability of these bubbles is governed by the properties of the helium liquid metal interface. In particular, interfacial tension between the immiscible phases controls bubble interactions and induces local pressure inhomogeneities. This work is motivated by the need of a better understanding of helium bubble formation in liquid Pb Li alloys, which are of particular relevance for the design of breeding blankets in the future nuclear fusion reactors. We employ classical molecular dynamics simulations to investigate helium segregation in a range of lead lithium systems, including the limiting cases of pure lead and pure lithium. Changes in local pressure are evaluated from direct mechanical calculations, enabling the characterization of interfacial properties. Interfacial tension and radius of the bubble are subsequently determined across multiple thermodynamic conditions, spanning temperatures starting near the melting points of the constituent metals up to 1021 K. The impact of curvature and composition of the alloy on the interfacial behaviour are also investigated.
Fe$_3$GeTe$_2$ is a prototypical metallic van der Waals ferromagnet with itinerant magnetism and a highly tunable Curie temperature, yet how electronic excitations couple to spin and lattice degrees of freedom across its magnetic transition remains largely unexplored. Here, we use two-color pump-probe reflectivity to investigate the coupled electronic, spin, and lattice dynamics. The time-resolved reflectivity exhibits a tri-exponential relaxation, in which the intermediate component shows an anomaly near the Curie temperature due to enhanced interlayer spin-lattice interactions, while the slowest component displays pronounced critical slowing down with an exponent of ~ 0.3, revealing non-universal relaxation dynamics associated with intralayer spin correlations. Furthermore, we observe an emergent nonequilibrium A1g phonon Fano asymmetry that is suppressed in the ferromagnetic phase but anomalously enhanced in the paramagnetic regime, driven by thermally activated anharmonic decay pathways that bridge the kinematic gap to a hot electronic continuum. The pronounced enhancement of the acoustic strain pulse amplitude near T$_c$ further evidences robust magnetoelastic coupling. Overall, our results reveal how magnetic order governs the interplay among critical spin dynamics, electronic continuum excitations, and lattice response in metallic van der Waals ferromagnets
Extra components in series to non-ferroelectric capacitance can cause apparent ferroelectric D-E hysteresis loops even with the double-wave method (DWM). Characteristics of fake loops are studied withsimple circuit models, and suspicious loops of actual materials are found in some papers using the DWM. Inverse piezoelectric strain also reverses along with the polarization by the electric field, and S-E loops are considered more reliable to prove ferroelectricity than the D-E loops. But measurement of S-E loops usually requires expensive instruments in contrast to D-E loops with a simple circuit. A very simple method is developed to observe S-E loops of bulk samples with an inexpensive small electret microphone and a little expansion to the circuit for D-E loops. S-E loops of a commercial ceramic capacitor by this method reveal that its D-E loops apparent.
We report the synthesis of the Cr analogue of sugilite, KNa$_2$Cr$_2$Li$_3$Si$_{12}$O$_{30}$, in the milarite-type framework. Rietveld refinement of a composition-conserving antisite model gives $x = 0.0024(18)$ in KNa$_2$[Cr$_{2-x}$Li$_x$][Li$_{3-x}$Cr$_x$]Si$_{12}$O$_{30}$, corresponding to a T2-site Cr occupancy of $0.0008(6)$. X-ray MEM analysis shows no detectable Cr-like density at T2. Magnetic susceptibility indicates weak antiferromagnetic interactions with $θ_{\mathrm{W}} = -4.78(7)$ K and no ordering above 1.8 K.
This work reports the fabrication and optical characterization of ultraviolet-emitting plastic scintillators based on polystyrene [-(CH2-CH(C6H5))n-] and polyvinyltoluene [-(CH2-CH(C6H4CH3))n-] doped with different concentrations of naphthalene (Naph). Photoluminescence (PL) measurements show that Naph incorporation enhances the emission intensity, inducing a red shift in the emission wavelength of the polystyrene (PS), while Naph-doped polyvinyltoluene (PVT) exhibits a bimodal emission with peaks around 335 and 350 nm. Time-resolved photoluminescence (TRPL) reveals fast decay components for the undoped matrices and a dominant slow component of approximately 87 ns in the doped samples. Scintillation light yield measurements indicate moderate performance for the undoped polymers and a significant enhancement upon Naph doping. Proton irradiation experiments reveal a reduction in light yield for all samples, with Naph-doped PVT retaining a larger fraction of its initial light yield compared to PS-based scintillators, indicating improved radiation tolerance. Overall, these results demonstrate the effectiveness of Naph as a UV emission enhancer and highlight Naph-doped PVT as a promising candidate for compact and radiation-resistant scintillation detectors.
Electron irradiation is essential for materials characterization and modification, though the fundamental interactions between incident electrons and host materials remain under investigation. Here, we employ first-principles simulations to study electron dynamics under external electron irradiation. We quantify differences in key observables, including kinetic energy loss, secondary electron emission, and backscattered electrons, between classical and quantum mechanical descriptions of the incident electron. Around 400 eV incident energy, we identify significant differences in backscattered electron yields between classical point-charge and quantum wave-packet descriptions, whereas the quantum-mechanical effects diminish at incident energies above 600 eV. These differences highlight the critical importance of quantum effects in electron irradiation phenomena that occur in a specific energy range of the incident electron. Our results provide clear guidance for selecting appropriate incident, electron descriptions based on kinetic-energy regimes, identify a targeted experimental window for isolating quantum-only backscattering, and enable the rational design of 2D materials for nanofabrication and high-resolution electron-beam technologies.
Control of light polarization and propagation in sub-wavelength architectures is foundational to nanophotonic technologies. A frontier direction is to leverage strong optical spin-orbit interactions to realize polarization-selective light steering, known as the photonic spin Hall effect. In this context, hyperbolic plasmon polaritons (HPPs) are of particular interest as they offer large optical spin-orbit coupling from strong confinement and dielectric anisotropy, as well as ray-like propagation. Despite theoretical predictions, however, the hyperbolic spin Hall effect in natural materials has remained elusive. Here, we demonstrate the hyperbolic spin Hall effect in the visible and near-infrared range in the natural hyperbolic van der Waals metal MoOCl2. Enabling this discovery is a novel far-field pump-probe microscope that facilitates the launching and imaging of HPPs with exceptional sensitivity through interference with a high-momentum reference field. This approach preserves excellent control over light polarization, overcoming a key barrier to polarization-selective interrogation of hyperbolic materials. We show that both hyperbolic and surface plasmons in MoOCl2 display chiral fields, and that their propagation direction can be completely switched upon light helicity reversal. Our results demonstrate on-demand steering of chiral plasmons, firmly establishing natural hyperbolic materials as ideal components for reconfigurable nanophotonics and chiral light-matter coupling.
We present a variational reduced-order model for three-dimensional coplanar propagation of sharp cracks in heterogeneous perfectly brittle solids under mixed-mode I+II+III loading. The approach connects the variational fracture formulation of Francfort and Marigo (1998) and the perturbation theory of Rice (1985) by computing equilibrium crack-front configurations through minimization of the total energy defined as the sum of (i) the elastic potential energy, evaluated asymptotically from front deformations, and (ii) the dissipated energy, set by the fracture energy field. The potential energy and its derivatives are evaluated efficiently using the Fast Fourier Transform. The resulting nonconvex box-constrained minimization problem is solved with a matrix-free Newton conjugate gradient algorithm with a trust region and physics-based preconditioning, enforcing irreversibility while resolving energy barriers and long-range elastic interactions. We validate our implementation against newly derived analytical solutions. We then perform 116,000 large-scale simulations of tensile and shear crack propagation in disordered media to quantify the impact of finite-size effects, disorder intensity, and mode mixity. The simulations reproduce the transition from smooth to intermittent crack growth, and show that mode mixity has limited influence on the onset of intermittency but induces quasi-elliptic fronts in mixed II+III loading. They reveal a size-dependent crossover from disorder-induced weakening to toughening controlled by the emergence of depinning instabilities.
Scientific machine learning reports predictive performance. It does not report whether the same prediction would survive a different draw of training data. Across $9$ chemistry benchmarks, two classifiers trained on independent bootstraps of the same training set agree on aggregate accuracy to within $1.3\text{--}4.2$ percentage points but disagree on the class label of $8.0\text{--}21.8\%$ of test molecules. We call this gap \emph{cross-sample prediction churn}. The standard parameter-side techniques (deep ensembles, MC dropout, stochastic weight averaging) do not reduce this gap; two data-side methods do. The first is $K$-bootstrap bagging, which cuts the rate $40\text{--}54\%$ on every dataset at no accuracy cost ($K{\times}$-ERM compute). The second is \emph{twin-bootstrap}, our proposal: two networks trained jointly on independent bootstraps with a sym-KL consistency loss between their predictions, which at matched $2{\times}$-ERM compute reduces churn a further median $45\%$ beyond bagging-$K{=}2$. Cross-sample prediction churn deserves a column alongside predictive performance in scientific-ML benchmark reports, because without it the parameter-side and data-side methods are indistinguishable on the metric they actually differ on.