Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
CellScout: Visual Analytics for Mining Biomarkers in Cell State Discovery
No full textCell state discovery is crucial for understanding biological systems and enhancing medical outcomes. A key aspect of this process is identifying distinct biomarkers that define specific cell states. However, difficulties arise from the co-discovery process of cell states and biomarkers: biologists often use dimensionality reduction to visualize cells in a two dimensional space. Then they usually interpret visually clustered cells as distinct states, from which they seek to identify unique biomarkers. However, this assumption is often invalid due to internal inconsistencies in a cluster, making the process trial-and error and highly uncertain. Therefore, biologists urgently need effective tools to help uncover the hidden association relationships between different cell populations and their potential biomarkers. To address this problem, we first designed a machine-learning algorithm based on the Mixture-of-Experts (MoE) technique to identify meaningful associations between cell populations and biomarkers. We further developed a visual analytics system CellScout-in collaboration with biologists, to help them explore and refine these association relationships to advance cell state discovery. We validated our system through expert interviews, from which we further selected a representative case to demonstrate its effectiveness in discovering new cell states.</p
City-scale rooftop retrofit prioritization: linking thermal performance, energy impacts, urban morphology, and population exposure
No full textUrban rooftops strongly influence heat exchange and cooling demand, yet most studies oversimplify energy fluxes, neglect district morphology, or lack decision-ready metrics. This study develops a city-scale rooftop modeling framework to evaluate six roof types—concrete, asphalt, cool coatings, ceramic tiles, radiative surfaces, and green roofs—across Hong Kong. The framework captures the full surface energy balance, quantifies morphology effects, and translates rooftop physics into citywide outcomes on energy, emissions, costs, and retrofit priorities. High-resolution LiDAR and building footprint data were integrated with ArcGIS solar radiation modeling to resolve district-specific shading and exposure. Results show clear contrasts: asphalt drives persistent overheating, while radiative and cool roofs sustain near-ambient states and achieve passive cooling. If all rooftops adopted radiative cooling, compared with current conditions, citywide cooling electricity demand would fall by 11.8 %, imported electricity by 13.5 %, cooling costs by 11.8 %, and total CO₂ emissions by 1.7 %. The Composite Retrofit Priority Index reveals two priorities: suburban districts with extensive rooftop areas achieve the largest energy and carbon reductions, while dense urban cores deliver the greatest equity benefits by reaching large populations. Together, these insights provide clear guidance for climate-resilient and socially equitable retrofit planning
Pathways to authoritarian capitalism
No full textThe global emergence of authoritarian capitalism warrants analysis of its features, economic pathways, and class basis. We argue that the timing of capitalism’s establishment relative to authoritarianism affects the extent of state capture by business and the regime’s stance towards labor. When capitalism preceded authoritarianism, autocracy arose from a “crisis of capitalism” in flawed democracies, either from fear of communist redistribution or backlash against neoliberalism in post-communist countries. These governments appealed to business and labor interests for electoral survival. Conversely, when authoritarianism preceded capitalism, late industrializing states with weak business and working classes pursued developmental agendas by exercising autonomy over capital and repressing labor. Understanding these historical pathways to authoritarian capitalism provides insight on contemporary democratic backsliding and right-wing populism in the U.S. and Europe.</p
Influence of particle morphology on hysteretic K<sub>0</sub> of granular soils
No full textIn many applications of geotechnical analysis and design, correct determination of the coefficient of earth pressure at rest, K0, is essential. The K0 in soils is typically estimated using empirical equations that depend on the soil's friction angle and stress history in terms of the overconsolidation ratio (OCR). However, this estimation ignores any effects of particle morphology in granular soils, potentially leading to unrealistically large hysteresis upon loading-unloading cycles. Existing experimental and numerical studies only reported K0 of granular soils with limited ranges of particle morphology during loading. This study conducted a comprehensive experimental campaign to systematically study K0 of soils with controlled morphology along loading and unloading paths. An oedometer equipped with flexible tactile sensors to precisely measure horizontal and vertical stresses during loading-unloading cycles was developed to obtain K0 in five granular soils with a wide range of particle shapes but similar particle size and initial density. Results showed that K0 during loading and unloading were strongly associated with particle morphology. Soils with a lower particle regularity (REG, i.e., average of roundness and sphericity) or overall regularity (OR, i.e., average of aspect ratio, convexity, and sphericity) mobilized more friction angles during loading, returning smaller K0 values. They also hold most of the locked-in stresses developed upon loading at different unloading stages, causing greater hysteresis. When employing REG and OR, equations were obtained for K0 during loading that are identical to Jaky's equation, even though these overall descriptors have different definitions and ranges of value. New empirical equations that incorporate REG and OR into the K0-OCR relationship were developed to markedly improve the estimation of hysteretic K0 after validating against high-quality laboratory data of 10 sand types from the literature.</p
Protic ionic liquid-based extraction of potentially toxic elements from digested sewage sludges with ecotoxicological and bioavailability assessment of residues
No full textAgricultural use of sewage sludges (SSLs) supports circular economy objectives but is often limited by high levels of potentially toxic elements (PTEs). The aim of this study was to use protic ionic liquids (PILs), characterized by different properties (1-methylimidazolium chloride ([H1Cim]Cl), triethylammonium hydrogen sulfate ([TEA][HSO4]), and 1-methylimidazolium hydrogen sulfate ([H1Cim][HSO4])) to remove PTEs from digested SSLs. Apart from assessing removal efficiency, the mechanisms of action of specific PILs were identified as well as the bioavailability of the PTEs and the toxicity of the residue after the removal of the PTEs were determined. [H1Cim]Cl was characterized by the highest PTE extraction efficiency. By using this solvent, from 11 % to 89 % of the PTEs was removed, among which Zn, Cd, and Cu were removed most effectively. However, the efficiency of the hydrogen sulfate-based PILs was relatively lower. The extraction efficiency was influenced by temperature, SSL-to-PIL ratio, ethanol washing, and SSL type. The treatment of the SSLs with [H1Cim]Cl significantly reduced the contents of some bioavailable PTEs, including Zn and Ni, while retaining key nutrients such as phosphorus. The leachates toxicity varied by SSL types and organisms, it decreased overall by 14–87 %, particularly for bacteria (78–87 %). The statistical analysis found that Ni and Zn could have been the critical factors determining the SSLs’ toxicity to bacteria, whereas As towards plants. The results confirm the high efficacy of the proposed method, which can be a promising solution supporting circular economy-based management of SSLs.</p
Atomically Dispersed Pt-Ru Dual-Atom Catalysts for Efficient Low-Temperature CO Oxidation Reaction
No full textWe successfully fabricated an atomically dispersed dual-atom catalyst featuring Pt1-Ru1 sites anchored on defective graphene (Pt1Ru1/ND@G). Pt1Ru1/ND@G achieves a high turnover frequency of 17.6 × 10−2 s−1 for CO oxidation at 30 °C, which is 10 times higher than Pt1/ND@G and demonstrates outstanding performance compared with the previous reports. Pt-Ru bond enhances the metallicity of both Pt and Ru atoms, facilitating the simultaneous adsorption and activation of CO and O2 and overcoming the limitations of single-atom catalysts.</p
Faithful in Steps: Improving Generalization and Citation in RAG via Query Decomposition
No full textRetrieval-augment generation is a prevalent strategy to mitigate hallucinations of LLMs. The attributable RAG (RAGQ) generates quotes for its answers. The quotes indicate which input contexts support the RAG to derive the answers, enhancing the answer's verifiability and trustworthiness. However, existing RAGQs exhibit significant degradation when dealing with questions that require multi-hop reasoning and multi-modal understanding, suffering from over-citation, implicit entity identification failure, and poor generalization. In this paper, we propose a novel RAGQ framework, namely QDRAG. QDRAG breaks down the input question into atomic subquestions to identify the implicit entities. Then, the reranker prunes context distractors to eliminate the downstream over-citation. To facilitate query decomposition, we propose two zero-shot approaches: QD-C and QD-R, which guide the QD MLLM to decompose the question based on context knowledge and retrieval rewards, respectively. One interesting finding is that finetuning on the QD task shows better generalizability compared to directly finetuning on the downstream RAGQ task. Experiments on four multi-modal QA benchmarks demonstrate QDRAG's efficacy in grounding answers and generating faithful citations. The framework significantly outperforms all the baselines on both in-domain and out-of-domain tests, even surpassing Gemini-Pro
Energy-generating smart windows based on reversible metal electrodeposition
No full textThe energy consumption and carbon emissions of buildings account for 34% and 37% of the global totals, respectively; thus, the development of energy-saving and carbon-neutral technologies in buildings is crucial for addressing the global energy crisis and reducing the carbon footprint. This work demonstrates a novel energy-generating smart window integrating a laminated semitransparent organic photovoltaic (STOPV) and a reversible metal electrodeposition-based electrochromic device (ECD). The smart window in the low-transmittance (tinted) state shows a significant light reflection property, which not only generates ∼16% more electricity relative to the high-transmission (clear) state but also shows a good heat insulation effect, reducing the temperature by 1.6 °C. Building energy simulations demonstrate that the smart window achieves an energy gain (sum of energy saving and energy generation) ranging from 10.1% to 16.2% in different cities around the world, indicating that the reflection-based smart window has great potential in energy-efficient and carbon-neutralized buildings.</p
Seasonal variability and sources of brown carbon in Wuhan: insights from Bayesian modeling and organic tracers-based source apportionment
No full textBrown carbon (BrC) is an important constituent of atmospheric carbonaceous aerosols, and accurately measuring its mass concentration and optical property is of great importance to reduce its uncertainties in model parametrization. This study deployed the novel Bayesian inference (BI) model to determine the BrC mass concentration and absorption on an hour-by-hour basis, using online measurements of multi-wavelength light absorption and total carbon (TC) data over June and December 2023 in Wuhan, central China. BrC mass contributed to 42 % of the TC aerosol and showed comparable level in two months, while its absorption showed distinct seasonal contrast with much higher values in December. Important BrC compositions including various organic tracers were measured, taking advantage of the concurrently collected daily 12-h offline filter samples. BrC mass and absorption showed good correlations with important components such as oxygenated-polycyclic aromatic hydrocarbons (PAHs), light PAHs and biomass burning tracer-saccharides. Organic-tracer based source apportionment was conducted for both BrC mass and absorption, and nine distinct source factors were resolved. Secondary sources especially secondary organic aerosol (SOA) factor dominated BrC mass contributions in June, while biomass burning emerged as the primary source in December. The primary sources contributing to BrC absorption were SOA and biomass burning in June, while biomass burning dominated in December. Coal combustion, which was a minor source to BrC mass, was a non-negligible contributor to BrC absorption in both months. Different control measures should be implemented to effectively control BrC mass for improved air quality and to mitigate BrC absorption for its climate effect. The results from this study broaden our understandings of the relationship among chemical, optical and sources of BrC, offering significant implications for environmental management.</p
Structural tuning of nanoporous metal via electrodeposition
No full textNanoporous metal, fabricated via the selective dissolution of an alloy (i.e., dealloying), can be filled with another metal via electrodeposition to create unique, functional structures unattainable via just dealloying. In this work, by controlling the charge of Ni deposition, we finetune the porosity and the pore width of nanoporous copper. At a sufficiently low rate, the deposition proceeds uniformly under interface control, until the porosity approaches a percolation threshold, which also governs the smallest attainable pore width. Via microscopic characterizations, we determine that we can tune down the porosity from 57.5% to 15.8% and the pore width from 89 nm to 35 nm, while retaining the structural bi-continuity. The tuned structure rejects 80% KCl from a 1 mM solution, a function not available in the pristine structure but enabled by the narrowed pores