Structure and elasticity of CaC2O5 suggests carbonate contribution to the seismic anomalies of Earth's mantle.

Knowledge of carbonate compounds under high pressure inside Earth is key to understanding the internal structure of the Earth, the deep carbon cycle and major geological events. Here we use first-principles simulations to calculate the structure and elasticity of CaC2O5-minerals with different symmetries under high pressure. Our calculations show that CaC2O5-minerals represent a group of low-density low-seismic-wave velocity mantle minerals. Changes in seismic wave velocity caused by the phase transformation of CaC2O5-Cc to CaC2O5-I[Formula: see text]2d (CaC2O5-C2-l) agree with wave velocity discontinuity at a depth of 660 km in the mantle transition zone. Moreover, when CaC2O5-Fdd2 transforms into CaC2O5-C2 under 70 GPa, its shear wave velocity decreases by 7.4%, and its density increases by 5.8%, which is consistent with the characteristics of large low-shear-velocity provinces (LLSVPs). Furthermore, the shear wave velocity of CaC2O5-I[Formula: see text]2d is very similar to that of cubic Ca-perovskite, which is one of the main constituents of the previously detected LLSVPs. Therefore, we propose that CaC2O5 and its high-pressure polymorphs may be a main component of LLSVPs.

Sweet Gradient matters: Designing consistent and efficient estimator for Zero-shot Architecture Search.

Zero-shot Neural Architecture Search has garnered attention due to its training-free nature and rapid search speed. However, existing zero-shot estimators commonly suffer from low consistency, which hampers their practicality. In this work, we theoretically analyze that network generalization and convergence are highly correlated with Sweet Gradient of Parameter, i.e., the number of parameters whose gradient absolute values are within a certain interval. Empirical results indicate that Sweet Gradient of Parameter brings a higher consistency than the overall number of parameters. Additionally, we demonstrate a positive correlation between the network depth and the proportion of parameters with sweet gradients in each layer. Based on the analysis, we propose a training-free method to find the Sweet Gradient interval and obtain an estimator, named Sweetimator. Furthermore, Sweet Gradient can be an effective and general approach to promote the consistency of zero-shot estimators. Experiments show that Sweetimator and Sweet-enhanced estimators have significant consistency improvement in multiple benchmarks. Our method achieves state-of-the-art performance with 256x speedup in NAS-Bench-201 and maintains high competitiveness in DARTS, MobileNet, and Transformer search spaces. The source code is available at https://github.com/xingxing-123/SweetGradient.

A first-principles study of the structural, electronic and elastic properties of the FeO2-FeO2He system under high pressure.

The origin of the wave velocity anomalies at the core-mantle boundary (CMB) has been controversial. The primordial helium reservoir in the deep lower mantle remains elusive even with geochemical evidence for its existence. Here, we calculated the density and wave velocity of the FeO2-FeO2He system under the CMB conditions using first principles. The FeO2 and FeO2He of pyrite-type can exist stably under the CMB conditions without melting, and the incorporated helium increases the stability of the system. The electrical properties of FeO2 and FeO2He are not related to pressure. Doped helium reduces the density of the system but increases the elastic modulus. Our results suggest that FeO2 can be used as a viable material composition of ultra-low velocity zones (ULVZs), and FeO2He can explain the D'' seismic discontinuity instead of ULVZs. The primordial helium reservoir possibly formed by the accumulation of FeO2He, the only stable solid helium-bearing compound under the CMB conditions, may coincide with the location of the D'' layer.