[en] The future High Luminosity LHC (HL-LHC) is expected to deliver about 5 times higher instantaneous luminosity than the present LHC, resulting in pile-up up to 200 interactions per bunch crossing (PU200). As part of the phase-II upgrade program, the CMS collaboration is developing a new endcap calorimeter system, the High Granularity Calorimeter (HGCAL), featuring highly-segmented hexagonal silicon sensors and scintillators with more than 6 million channels. For each event, the HGCAL clustering algorithm needs to group more than 10⁵ hits into clusters. As consequence of both high pile-up and the high granularity, the HGCAL clustering algorithm is confronted with an unprecedented computing load. CLUE (CLUsters of Energy) is a fast fullyparallelizable density-based clustering algorithm, optimized for high pile-up scenarios in high granularity calorimeters. In this paper, we present both CPU and GPU implementations of CLUE in the application of HGCAL clustering in the CMS Software framework (CMSSW). Comparing with the previous HGCAL clustering algorithm, CLUE on CPU (GPU) in CMSSW is 30x (180x) faster in processing PU200 events while outputting almost the same clustering results.

[en] Kondo semimetal CeRu₄Sn₆ is attracting renewed attention due to the theoretically predicted nontrivial topology in its electronic band structure. We report hydrostatic and chemical pressure effects on the transport properties of single- and poly-crystalline samples. The electrical resistivity ρ(T) is gradually enhanced by applying pressure over a wide temperature range from room temperature down to 25 mK. Two thermal activation gaps estimated from high- and low-temperature windows are found to increase with pressure. A flat ρ(T) observed at the lowest temperatures below 300 mK appears to be robust against both pressure and field. This feature as well as the increase of the energy gaps calls for more intensive investigations with respect to electron correlations and band topology. (paper)

[en] Highlights: • The eXtreme Gradient Boosting Regression (XGBR) algorithm was first applied to build a robust and predictive machine learning (ML) model for perovskite materials. • The method of combining machine learning and DFT calculation used in this article can greatly save computing resources and time, and accelerate the discovery of renewable energy materials. • Two lead-free perovskite structures screened out by the combination of machine learning and DFT calculation have reasonable band gap values, high environmental stability and good optical absorption properties. To accelerate the application of perovskite materials in photovoltaic solar cells, developing novel lead-free perovskite materials with suitable band gaps and high stability is vital. However, laborious experiment and density functional theory (DFT) calculation are time-consuming and incapable to screen promising perovskites rapidly and efficiently. Here, we proposed a novel search strategy combining machine learning and DFT calculation to screen 5,796 inorganic double perovskites. The eXtreme Gradient Boosting Regression (XGBR) algorithm was first applied to build a robust and predictive machine learning (ML) model for perovskite materials. XGBR algorithm yielded a lower mean square error (MSE) than both Artificial Neural Network (ANN) algorithm and Support Vector Regression (SVR) algorithm. From the ML model, two novel lead-free inorganic double perovskites: Na₂MgMnI₆, K₂NaInI₆, were obtained, suitable direct bandgaps of 1.46 eV for K₂NaInI₆ and 1.89 eV for Na₂MgMnI₆, which are similar to the organic–inorganic perovskite (MAPI3) CH₃NH₃PbI₃ (E_g = 1.6 eV), high thermal stability and good optical properties were also confirmed by DFT calculation. The method of combining ML and DFT exhibits high accuracy and significantly speeds up the discovery of promising perovskite materials.