Hongxiang Xu

Mixed nonionic/anionic surfactant systems are widely used to regulate complex oil–water/solid interfaces, yet the molecular role of anionic structural variation in interfacial coassembly and Winsor III microemulsion behavior remains insufficiently understood. Herein, Tween 80/SDBS (TS) and Tween 80/AES (TA) were selected as representative Tween 80-based mixed surfactant systems to clarify how different anionic structures influence oil–water interfacial organization. Molecular dynamics simulations were combined with middle-phase (Winsor III) microemulsion experiments to examine interfacial electrostatics, molecular packing, local hydration, local hydrogen-bond associations, counterion association, and the resulting microemulsion response. The simulations showed that electrostatic interactions made the dominant contribution to the nonbonded interfacial interactions, while local hydration and local hydrogen-bond associations further modulated interfacial organization. At a nonionic/anionic molar ratio of 4:7 under the investigated conditions, TS tended to exhibit relatively compact Tween 80 conformations and a locally concentrated interfacial organization, whereas TA exhibited a more hydrated and spatially coordinated interfacial arrangement. These differences were associated with the benzenesulfonate structure of SDBS and the EO-containing sulfate structure of AES, which gave rise to distinct charge distributions, counterion association, local hydration, and packing characteristics. Although the theoretical interfacial tensions of the TS and TA Winsor III microemulsions were both very low and comparable, their formulation-dependent behaviors remained distinct. Overall, the results establish a structure–interaction–response relationship linking anionic structural variation with local interfacial organization and macroscopic Winsor III microemulsion behavior.

Low-rank ultrafine flotation clean coal typically yields filter cake moisture above 20% due to abundant oxygen-containing functional groups and strong surface hydrophilicity. Conventional polyacrylamide (PAM) flocculants are hydrophilic and improve dewatering only by altering cake porosity, not by reducing particle surface hydrophilicity, so they remove little adsorbed water. In this study, hydrophobically modified anionic polyacrylamides (HMAPAM) were synthesized by grafting lauryl acrylate onto APAM. FTIR, ¹H NMR, XPS, and SEM confirmed the grafting and progressive enrichment of hydrophobic alkyl chains on the surface. Moderate hydrophobic modification markedly improved solid–liquid separation. HMAPAM-D (APAM/LA = 4.5:0.5) achieved a settling velocity of 0.817 cm/s at 9 mg/L, 50.2% higher than APAM, and reduced filter cake moisture to 16.64% at 1 mg/L under 0.6 MPa versus 19.39% for unmodified APAM. Excessive modification (HMAPAM-E, 4:1) promoted intramolecular self-association, producing heterogeneous flocs and higher filtration resistance that degraded dewatering efficiency. The performance gain stems from hydrophobic association combined with adsorption bridging. These results clarify how hydrophobic group content controls flocculation and dewatering, informing the design of better flocculants for this type of coal slurry.

The ash content of coal, defined as the inorganic residue remaining after complete combustion, serves as a key indicator of coal quality, with lower values signifying superior quality (Y. Liu et al...

Near-infrared (NIR) spectroscopy offers a rapid approach for coal ash estimation, which is critical for intelligent coal preparation. However, quantitative analysis of high-dimensional raw coal spectra is often hindered by overlapping overtones and weak mineral absorptions. Furthermore, purely data-driven models are prone to severe overfitting under industrial small-sample scenarios due to the massive parameter search space. To overcome this, we propose DDAFMNet, a dual-domain spectral architecture integrating mineralogical prior knowledge. A hierarchical backbone comprising multi-layer convolutions and bidirectional memory connections first extracts local morphological and global sequential features. Crucially, the characteristic NIR absorption bands of primary ash-forming minerals are transformed into learnable inductive biases and explicitly injected into a frequency-domain branch. This prior knowledge injection substantially reduces the feature dimensionality and explicitly restricts the parameter search space, effectively mitigating overfitting while enhancing sensitivity to weak mineral vibrations. A dynamic weight generator subsequently achieves instance-level adaptive fusion of frequency-domain energy and wavelength-domain dependencies. Evaluated on a raw coal dataset, DDAFMNet achieves an RMSE of 1.284% and an RPD of 11.631, outperforming state-of-the-art prior-coupling algorithms. Attaining an 83.3% compliance rate under a 1.0% industrial error tolerance, this knowledge-driven framework provides a highly robust and interpretable methodology for complex coal quality monitoring.