N,N-dimethylformamide (DMF) has excellent chemical stability and is widely used as an aprotic polar solvent. In order to reduce production costs and reduce pollution to the surrounding environment, it is necessary to recycle and reuse DMF. Previous research has found that the thin film composite nanofiltration membrane prepared from liquefied walnut shells exhibited a high rejection rate in DMF, but relatively low permeance and mechanical strength. In order to increase permeance without compromising the separation performance, ethylenediamine (EDA) is used as a modifier to graft onto the structure of liquefied walnut shell through the Mannich reaction. Then, modified liquefied walnut shell as an aqueous monomer reacts with trimesoyl chloride (TMC) via the interfacial polymerization method on the EDA-crosslinked polyetherimide (PEI) membrane. The results show that the permeance of the prepared membrane is significantly improved by an order of magnitude, demonstrating a rejection rate of 98% for crystal violet (CV), and a permeance of 3.53 L m⁻² h⁻¹ bar⁻¹ in DMF. In conclusion, this study reveals the potential of utilizing liquefied walnut shells as raw materials for preparing high-performance separation membranes and demonstrates that surface modification is a feasible approach to enhance permeance of membranes without sacrificing the rejection rate.

We conducted integrated thermodynamic analyses for the two-step method of CsPbBr₃ growth. Deriving from Gibbs free energy and heterogeneous nucleation theory, a thermodynamic reaction model equation was proposed and verified, from which we could find that the grainsizes of the CsPbBr₃ can be precisely controlled by the CsBr concentration. The impact of temperature of the solvent was also studied. Finally, a best cell was fabricated showing a power conversion efficiency of 4.75%, with open circuit voltage of 1.13 V, short circuit current of 6.68 mA/cm² and fill factor of 63%. Our results will help other researchers to determine the appropriate condition of the growth of CsPbBr₃ by the two-step method.

Recommender systems face fundamental challenges, including extreme data sparsity and noisy rating observations. We propose L21RSVD, a robust latent factor model that employs L21 norm regularization within a Singular Value Decomposition framework to mitigate the impact of outliers while preserving low-rank structure. Unlike conventional L2-regularized approaches, our formulation induces group sparsity in the latent factor space, yielding more discriminative user and item representations. We derive three optimization variants: standard L21RSVD, L21RSVD without the squared term, and coefficient-free adaptive L21RSVD. Building upon these, we introduce a fusion strategy that adaptively aggregates predictions based on local data density. Extensive experiments on benchmark datasets demonstrate that L21RSVD substantially outperforms classical collaborative filtering and the SVD-type model. The proposed fusion model achieves state-of-the-art performance, reducing RMSE by up to 31.21% and MAE by up to 42.99% relative to baseline methods.