Nanoscience News (<front>)

Abstract not available

• Speaker: Jelena Ninic, University of Birmingham
• Friday 17 October 2025, 15:00-16:00
• Venue: CivEng Seminar Room (1-33) (Civil Engineering Building).
• Series: Engineering Department Structures Research Seminars; organiser: Shehara Perera.

Add to your calendar or Include in your list

In this work, we develop a class of interacting particle Langevin algorithms to solve inverse problems for partial differential equations (PDEs). We leverage the statistical finite elements formulation to obtain a finite-dimensional statistical model, where the parameter is that of the forward map and the latent variable is the discretised solution of the PDE , assumed to be partially observed. We then adapt a recently proposed expectation-maximisation like scheme, the interacting particle Langevin algorithm (IPLA), for this problem and obtain a joint estimation procedure for the parameters and the latent variables. The estimation of an unknown source term is demonstrated for linear and nonlinear Poisson PDEs, as well as the diffusivity parameter for the linear Poisson PDE . We provide computational complexity estimates for forcing estimation in the linear case, including comprehensive numerical experiments and preconditioning strategies that significantly improve the performance.

• Speaker: Alex Glyn-Davies, University of Cambridge, UK
• Friday 30 May 2025, 15:00-16:00
• Venue: CivEng Seminar Room (1-33) (Civil Engineering Building).
• Series: Engineering Department Structures Research Seminars; organiser: Shehara Perera.

Add to your calendar or Include in your list

Abstract not available

• Speaker: Matthew Santer, Imperial College London, UK
• Friday 10 October 2025, 15:00-16:00
• Venue: CivEng Seminar Room (1-33) (Civil Engineering Building).
• Series: Engineering Department Structures Research Seminars; organiser: Shehara Perera.

Add to your calendar or Include in your list

While model design gets much of the spotlight, subtle data choices, such as which documents are seen and how they’re represented, can profoundly shape the behaviour of language models. Nowadays, training data is the secret sauce behind a language model’s success, yet it remains relatively understudied. In this talk, I will discuss how training data influences a model’s behaviour via two key phenomena: memorisation and tokenisation bias. First, I’ll present our work on memorisation, asking: To what extent does a model remember specific documents it was trained on? Directly answering this question is computationally expensive. Instead, we frame memorisation as a causal question and introduce an efficient method to estimate it without re-training. This reveals how memorisation depends on factors such as data order and model size. Next, I’ll discuss how subword tokenisation, often seen as a preprocessing detail, systematically biases model predictions. We ask: How would a model’s output change if a piece of text were tokenised as one subword instead of two? Using tools from econometrics, we estimate this counterfactual question without re-training the model using a different vocabulary. We show that when a piece of text is tokenised into fewer subwords, it consistently receives a higher probability. Together, these results show that training data profoundly shapes a model’s behaviour. Causal methods let us efficiently estimate and understand these phenomena, offering insight into how to better train language models.

Bio: Pietro Lesci is a final-year PhD student in Computer Science at the University of Cambridge, working with Prof Andreas Vlachos. His research explores how training data shape a model’s behaviour, focusing on memorisation, tokenisation, and generalisation. To study this question, he draws on causal methods from econometrics. His work has been presented at major machine learning conferences such as ICLR , ACL, NAACL , and EMNLP . He has received the Best Paper Award at ACL 2024 , the Paper of the Year Award from Cambridge’s Department of Computer Science and Technology, and funding from Translated’s Imminent Research Grant. Pietro’s experience spans academia and industry, including 3+ years working in research labs, consulting firms, and international institutions. He holds an MSc in Economic and Social Sciences from Bocconi University.

• Speaker: Pietro Lesci (University of Cambridge)
• Friday 30 May 2025, 12:00-13:00
• Venue: Room FW26 with Hybrid Format. Here is the Zoom link for those that wish to join online: https://cam-ac-uk.zoom.us/j/4751389294?pwd=Z2ZOSDk0eG1wZldVWG1GVVhrTzFIZz09.
• Series: NLIP Seminar Series; organiser: Suchir Salhan.

Add to your calendar or Include in your list

Abstract not available

• Speaker: Andrei Paleyes (University of Cambridge)
• Tuesday 24 June 2025, 13:00-14:00
• Venue: Lecture Theatre 2, Computer Laboratory, William Gates Building.
• Series: Artificial Intelligence Research Group Talks (Computer Laboratory); organiser: Mateja Jamnik.

Add to your calendar or Include in your list

In this project (N = 8,525) we longitudinally evaluate the effectiveness of a gamified inoculation intervention that protects people against vaccine misinformation (“Bad Vaxx”). Across 30 days after the initial intervention (Bad Vaxx or Tetris) we re-invited a subset of participants to an item rating task to establish their misinformation discernment skills. Some participants in the control group were exposed to inoculation “diffusion messages”, messages generated by inoculated participants to help protect non-inoculated participants. This is the first inoculation longevity study that maps the inoculation decay curve with daily data points for an entire month. We found that the effects decay completely within the first two weeks, unless a booster intervention was administered to top up the effect. We also establish that a key mechanism for the decay of the effect is forgetting what was learned, rather than a decline in motivation. Meanwhile, inoculation diffusion messages had little to no effects, indicating that the scope for spread immunity is likely limited.

• Speaker: Rakoen Maertens (University of Oxford)
• Wednesday 28 May 2025, 15:00-16:00
• Venue: Ground Floor Lecture Theatre, Department of Psychology, Downing Site, Cambridge.
• Series: Social Psychology Seminar Series (SPSS); organiser: Miriam Remshard.

Add to your calendar or Include in your list

The seminar will be an interdisciplinary journey between applications of Artificial Intelligence (AI) Computer Vision based methods to the field of Forensics and Biomedical Imaging. I will first focus on AI applications for country recognition and city verification, presenting two works in which AI based models have been applied to such forensics task, with a final overview of other forensics applications such as visual sentiment analysis and detection of generated images through AI methods. Subsequently the seminar will focus on further applications of AI vision methods to the field of deep generative models for Biomedical Imaging, with a particular focus on a work on deep generative model for structural brain MRI reconstruction and interpretation.

• Speaker: Giovanna Maria Dimitri, University of Siena (Italy)
• Tuesday 27 May 2025, 13:00-14:00
• Venue: Lecture Theatre 2, Computer Laboratory, William Gates Building.
• Series: Artificial Intelligence Research Group Talks (Computer Laboratory); organiser: Pietro Lio.

Add to your calendar or Include in your list

Traditional theories of Spike Time Dependent Plasticity rely on near coincidence between postsynaptic and presynaptic action potentials to produce plasticity. However, recently, the unique dendritic features of the pyramidal neurons, specifically the segregation of their dendritic trees into basal and apical dendrites, have been investigated as a site for branch specific plasticity – the separation of plasticity ‘rules’ across dendritic locales. In this talk, Carl will first introduce the idea of dendrite specific plasticity and present a recent paper by Wright et al. (2025) investigating the differences between apical and basal plasticity rules in L2/3 pyramidal dendrites. Saeyeon will then present the synaptic plasticity mechanisms occurring in the distal tuft dendrites of hippocampal pyramidal neurons (O’Hare et al, 2025). Hippocampal pyramidal neurons support spatial memory by forming place fields, which are known to arise through a recently described mechanism called Behavioral Timescale Synaptic Plasticity (BTSP). According to BTSP , place field formation is driven by plateau potentials in distal dendrites that lead to widespread depolarization across the entire dendritic arbor. However, the relationship between the dendritic and somatic activity in vivo, and the cellular mechanisms that initiate these plateau potentials, remain unclear. Using in vivo recording of both somatic and distal dendritic signals in a single CA1 neuron during a virtual reality-based spatial navigation task in mice, the authors expand on the BTSP model. They demonstrate that the timing and magnitude of dendritic activity predict key properties of the resultant somatic place fields, and that distal tuft dendrites undergo local plasticity during place field formation. Papers Distinct synaptic plasticity rules operate across dendritic compartments in vivo during learning Distal tuft dendrites predict properties of new hippocampal place fields

• Speaker: Carl Ashworth; Saeyeon Kwon
• Tuesday 27 May 2025, 11:00-12:45
• Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
• Series: Computational Neuroscience; organiser: .

Add to your calendar or Include in your list

Covalent organic framework membranes featuring abundant acidic groups and matching channel size are constructed. The strong membrane-ion interactions between the acidic sites and multivalent cations create a dual-nanoconfinement effect within rigid channels. This mechanism accelerates monovalent ions transport while hindering multivalent ions, enabling superior mixed ion separations.

Ion transport in environments with mixed ions is crucial for numerous real-world applications. However, most membranes suffer from reduced selectivity in the mixed ion environment. Herein, the concept of dual-nanoconfined channels in covalent organic frameworks (COFs) is demonstrated for superior mixed mono-/multivalent ion separation. By incorporating acidic functionalities into the pore wall and coupling with a suitable pore size in the COF membranes, multivalent ions interact strongly with the acidic groups. This shields the interactions of the wall with monovalent ions, leaves sufficient space for fast monovalent ion passage, and in the meantime blocks multivalent ions, creating well-defined dual-nanoconfinement for mixed ion separation. Consequently, mixed Li+/Mg2+ ion selectivity exceeding 1,300 and mono-/trivalent ion selectivity over 9,000 is achieved. Molecular dynamics simulation and experiment with multiple ion pairs further confirm the mechanisms. The study sheds light on a fundamentally new design principle for mixed ion separation, offering the potential to reach remarkable selectivity in practical applications.

Surprisingly, a cell can bind to itself to make a self-adhesion, which engineered here to improve how cells attach to biomaterials. Nanoprinting are used to make 3D structures smaller than cells–called Self-Adhesion-Tunnels (SATs)–around which cells can wrap and bind to themselves. These self-adhesions improve cell attachment stability and offer a new approach for biomaterial design.

A cell can bind to itself and form a self-adhesion that can be engineered and harnessed as a new way to adhere cells to engineered materials–a key challenge for biomaterials are demonstrated. Here, a 3D structure smaller is developed than a single cell, that a Self-Adhesion-Tunnel (SAT) is called, that causes cells to wrap around it and bind to themselves. This process is driven through the cadherin proteins that regulate cell-cell adhesion, and it is shown that many of the key elements of a normal cell-cell adhesion are found in self-adhesions. Size and shape of the SAT determine the efficiency of self-adhesion formation, and >90% efficient formation of self-adhesions are observed in both kidney and skin cells per SAT. Self-adhesions can persist for at least 24 hrs and act to stabilize the cell-material interface and reduce migration. Overall, this ability to co-opt the native cell-cell adhesion machinery in cells and use it as an attachment strategy can provide new approaches for soft-tissue implant integration and tissue engineering scaffolds where stable tissue-material interfaces are critical.

This study proposed that doping indium (In) in p-type SnSe introduced resonant levels near the Fermi level in the valence bands, effectively enhancing the local density of states and significantly boosting the Seebeck coefficient. These improvements result in extraordinary near-room-temperature thermoelectric performances. Specifically, the optimal SnSe-0.7%In sample achieved a PF of ≈55 µWcm−1K−2, and a ZT of ≈1.0 at 300 K.

Tin selenide (SnSe) has emerged as a promising thermoelectric cooling candidate, exhibiting room-temperature performance comparable to that of commercial bismuth telluride (Bi2Te3). However, the Seebeck coefficient of p-type SnSe crystals remains significantly lower than that of n-type Bi₂(Te, Se)₃ (BTS), and the resulting mismatches hinder effective utilization of its excellent cooling potential. To address this limitation, resonant levels are introduced in the valence bands of hole-doped SnSe through indium-doping, which increased the density of states and thereby boosted the Seebeck coefficient. This strategy enable the power factor to reach ≈55 µWcm−1K−2 and ZT value of ≈1.0 at 300 K, with a more matching Seebeck coefficient of ≈211 µVK−1. Furthermore, a full-scale thermoelectric cooler incorporating the p-type SnSe paired with n-type BTS demonstrated a maximum cooling temperature difference (ΔT max) of ≈81.1 K at 343 K. A SnSe-based single-leg device achieve a conversion efficiency of ≈7.0% under a ΔT of 250 K. These findings highlight that matching thermoelectric parameter of p-type and n-type materials is crucial for enhancing the cooling efficiency of devices, and engineering resonant energy levels constitutes a robust strategy for solving the inherent performance limitations of p-type SnSe in practical thermoelectric applications.

Covalent organic framework nanosheets are engineered via ternary monomer design to achieve programmable interlayer offsets. Non-planar alkyl-functionalized monomers induce interlayer torsion, transitioning AA to AA' and AB stacking for angstrom-precise pore tuning. The mechanically robust, blade-cast membranes exhibit exceptional H₂/CO₂ selectivity up to 60 in gas mixtures from methanol-reforming.

Hydrogen purification by membrane technology offers a sustainable path to meet the escalating demands of green energy. However, conventional polymeric membranes are constrained by permeability-selectivity trade-off and instability under real-world operating conditions. While covalent organic framework (COF) membranes hold promise, their overlarge pores and poor film-processibility are to be imperatively solved. Herein, a ternary building unit system is designed for synthesizing imine-based COF nanosheets with programmable interlayer offsets. By synergizing a planar aldehyde monomer as the basic structural unit and a none-planar alkyl-functionalized aldehyde monomer as the structure regulation unit, we induce layer distortion that disrupts π–π dominated AA stacking, enabling angstrom-precise pore tuning (1.4–0.6 nm) via controlled transitions to AB stacking while retaining crystallinity. The mechanically robust nanosheets are easily assembled into large-area membranes via a facile blade casting, overcoming the processability bottleneck associated with binary building unit systems. The resulting membranes demonstrate an exceptional H2/CO2 selectivity of 60, surpassing existing benchmarks. When treating gas mixtures from methanol steam reforming, a two-stage membrane process achieves 99.5% H2 purity and 94.0% recovery. Precise modulation of pore architecture and mechanical flexibility through building units engineered stacking affords a platform for microporous organic membranes.

In this work, a novel floatable hydrogen-freshwater cogeneration hybrid hydrogel featuring a dynamic gas-solid interface toward photothermal-assisted photocatalytic hydrogen evolution reaction is reported. The synergy of photothermal materials and 3D nanostructured hydrogel on covalent triazine framework facilitates solar interfacial evaporation and enhanced hydrogen production, highlighting its great potential for scalable and sustainable solar energy utilization.

Organic photocatalysts generally suffer from insufficient near-infrared light absorption and undesirable photogenerated charge transport properties, resulting in unfavorable hydrogen evolution performance from water splitting. Hydrogen evolution reaction (HER) is also known to be significantly influenced by the interfacial charge and mass transfer in a catalyst/H2O biphase system. Herein, for the first time, a highly stable and floating hydrogen-water cogeneration hybrid hydrogel that utilizes photothermal-induced interface microenvironment variation to accelerate sluggish photocatalytic water splitting reaction is reported. Supported by solar-powered interfacial evaporation and efficient vapor generation, the rationally designed hydrogel effectively transforms the conventional liquid-solid interface into a gas-solid photocatalytic interface. The presence of gas-liquid coexistence state offers a disordered and loose hydrogen-bond network while preserving the proton transfer channel, greatly reducing reaction activation energy and interfacial energy barriers. The improved heat and mass transfer together with optimized charge transfer pathways suppress electron-hole recombination, the integrated photothermal-coupled solar photocatalytic hydrogel exhibits excellent operational stability and self-adaptive rotation in seawater, mitigating salt accumulation and achieving an exceptional vapor generation rate of 4.71 kg m−2 h−1 and a hydrogen-evolving rate of 1961.25 µmol g−1 h−1 under one sun illumination.

In this study, an innovative myocardial infarction repair strategy by creating a “pre-neuralized” scaffold that combines strontium silicate microparticles with neural stem cells (NSCs) is introduced. This scaffold exhibits neuroregulatory capabilities that facilitates myocardial repair and improves cardiac function in vivo by modulating genes linked to circadian rhythm, underscoring the strategic benefit of neural-induced regulation in tissue repair.

Neural networks are found to play an important role in monitoring and coordinating cardiac physiological activities. However, the clinical use of neuroregulatory strategies for repairing infarcted myocardium, such as vagus nerve stimulation and pharmacological activation, confronts the challenges of managing stimulation signals and potential drug side effects. In this study, an innovative myocardial infarction repair strategy by creating a “pre-neuralized” scaffold that combines strontium silicate microparticles with neural stem cells (NSCs) is introduced. Strontium silicate promotes NSCs differentiation, resulting in a scaffold enriched with mature neurons. This scaffold exhibits neuroregulatory capabilities that enhance the maturation and synchronized contraction of cardiomyocytes, facilitating myocardial repair and improving cardiac function in vivo. The findings indicate that the pre-neuralized scaffold aids myocardial recovery by modulating genes linked to circadian rhythm, underscoring the strategic benefit of neural-induced regulation in tissue repair. In conclusion, this study presents a promising approach to repairing infarcted myocardium using inorganic biomaterial-induced scaffolds with neuromodulatory properties from the perspective of systemically physiological regulation. This work may offer a new perspective for addressing complex tissue and organ injuries.

By intentionally introducing H2O to treat the ZnMgO and post-annealing the devices, the ZnMgO defects can be fully passivated soon after the devices are fabricated. The developed in situ H2O treatment strategy, which enables the realization of high performance QLEDs and the elimination of uncontrollable positive aging, could promote the industrialization process of QLEDs.

Quantum dot light-emitting diodes (QLEDs) usually exhibit a positive aging phenomenon with their efficiency and brightness increase after storage for several days. Although positive aging enhances device performance, its random and uncontrollable process presents substantial challenges for QLED industrialization. Here, It is identified that the H2O, produced by the reaction between the acidic resin and ZnMgO, can slowly passivate the defects of ZnMgO and is thus responsible for the positive aging. By intentionally introducing H2O to treat the ZnMgO and post-annealing the devices, the ZnMgO defects can be fully passivated soon after the devices are fabricated. As a result, the fresh devices exhibit a high efficiency of 55.2 lm W−1 (26%), which is 2.04 (1.67)-fold higher than those of untreated devices and 1.21 (1.02)-fold higher than those of untreated devices after positive aging. Moreover, the performance of the treated devices remains very stable even after storage for several days, indicating the undesired positive aging is completely eliminated. The developed in situ H2O treatment strategy, which enables the realization of high performance QLEDs and the elimination of uncontrollable positive aging, can promote the industrialization process of QLEDs.

This work leverages a conductance-reinforced attenuation mechanism via enhanced Metal-N coordination to boost electron mobility and optimize electromagnetic response. The π-conjugated MOFs are 3D-printed into a stealth drone propeller that achieves both radar and infrared stealth.

π-conjugated metal-organic frameworks (MOFs) have emerged as promising candidates for electromagnetic wave (EMW) absorption, owning to their high conductivity and versatile structural tunability. Nevertheless, the effective control over their dielectric properties is a challenge. Herein, the charge carrier migration in π-conjugated MOFs is harnessed to significantly amplify the electromagnetic response, where the strengthened atom coordination can activate a distinctive conductance-reinforced attenuation mechanism. This results in finely calibrated EMW absorption characteristics, including a wide effective absorption bandwidth of 6.0 GHz at mere 2 mm, a minimum reflection loss of −46.7 dB at 3.5 mm, and a substantial reduction in radar cross-section (RCS) up to −23.3 dBm2. Furthermore, the seamless integration of the π-conjugated MOF hybrids within ultraviolet (UV)-curable 3D printing technology has enabled the fabrication of a stealth-enabled drone propeller prototype, which exhibits a remarkably low infrared emissivity of 0.205. Additionally, when the propeller device is subjected to a 100 °C heating platform for 30 min, its surface temperature remains below 50 °C, demonstrating exceptional thermal management and stability under elevated temperature conditions. This work underscores the immense potential of these cutting-edge absorbers to shape the future of advanced military stealth technologies.

The high entropy design effectively inhibits the structural degradation in Li-rich cathode materials during cycling, including the oxygen vacancy formation, volume expansion, and Mn coordination environment disordering, thus significantly improving the voltage retention and energy output. This migration retardation effect is attributed to the synergistic enhancement of covalency and ionicity, which makes Mn ion migration thermodynamically infeasible.

The development of advanced Li- and Mn-rich layered cathodes (LRO) is essential for high-energy lithium-ion batteries (LIBs). However, LRO exhibits large voltage hysteresis and rapid voltage decay with irreversible TM migration upon prolonged cycling. Given that high-entropy oxides have expanded the potential for retarding the harmful phase transition and regulating the site energies, therefore a high-entropy Li1.17Mn0.50Ni0.12Co0.12Mg0.03Cu0.02Ti0.02Nb0.02O2 cathode is synthesized (HELRO) for LIBs in the present study, demonstrated significantly improved voltage retention and energy output. In addition, this work unveils the sluggish degradation of superlattice and local structure in HELRO during long charge–discharge cycles and explains the “migration retardation effect.” The higher configurational entropy contributes to the higher energy barriers for in-plane, out-of-plane, and continuous Mn migrations due to the synergistic ionic–covalent enhancement of Mn─O bonds. This work provides new insights for understanding the improvement mechanisms of high entropy cathodes and demonstrates the feasibility of suppressing long-standing voltage decay by high entropy design combining covalent and ionic elements.

Compliance current can be utilized to control the volatile and nonvolatile resistive switching behaviors of γ-Cu2S, which results from the γ- to β-Cu2S phase transition and orientation-dependent formation (along γ-Cu2S [201]) of high-digenite Cu1.8S conductive channels, respectively. The Cu1.8S channels form from an intermediate δ-Cu2S phase through the directional Cu+ migration along δ-Cu2S [001].

Achieving multiple types of resistive switching in a single material with controlled ionic motion is a key challenge in neuromorphic computing, traditionally addressed by combining materials with distinct switching behaviors. Here, Cu2-xS is identified as a promising candidate to overcome this limitation due to its hierarchical phase transitions. Using in situ biasing experiments, reversible and non-reversible phase transitions (and resistive switching) are demonstrated in γ-Cu2S by controlling the compliance current. The formation of parallel high-digenite Cu1.8S channels, orientated along the γ-Cu2S [201] crystallographic direction, drives the nonvolatile resistive switching. These channels emerge via an intermediate δ-Cu2S phase and are stabilized at room temperature by residual strains, alongside β-Cu2S phase. The work clarifies the complex, electrically triggered phase transformations in γ-Cu2S, and highlights the potential of Cu2-xS as a versatile material for neuromorphic computing.

Cephalopods use chromatophore organs (muscle-actuated pigment sacs) to alter their skin color and pattern. Synthetic chromatophores, which closely mimic the mechano-optical process found in cephalopods using stimuli-responsive microscale hydrogel actuators, are reported. Hierarchical organization and morphological control of these synthetic chromatophores enable inherently stretchable skins with color and pattern morphing capabilities and rapid response times.

The dynamic optical and mechanical properties of cephalopod skin cannot be mimicked using traditional display technologies. Soft materials (and systems thereof) have the potential to realize cephalopod-like color switching capabilities synthetically. This report describes the fabrication of stretchable arrays of microstructured, stimuli-responsive hydrogels, “synthetic chromatophores,” that emulate the mechano-dynamic action of color change found in cephalopods. By combining multiple layers of these synthetic chromatophores, soft skins with color and pattern morphing capabilities that leverage halftone absorption, optical interference, and microlensing are demonstrated. These skins, made entirely of soft materials, are inherently stretchable and can be programmed to respond to specific environmental stimuli, making them well-suited for applications in soft robotics and human-machine interfaces.

We introduce a unified Bayesian anomaly-detection framework for Cosmology, applied to the REACH global 21cm probe and also Type Ia supernovae. This approach embeds data-integrity beliefs directly into the inference process. Rather than excising contaminated or anomalous data points, the method employs a piecewise likelihood constrained by a Bernoulli prior and an Occam penalty, allowing anomalies to be down-weighted automatically while performing numerical sampling for parameter inference. When applied to supernova light curves, the framework yields precise estimates of brightness scaling, stretch, and colour, while also automating supernova sample and band selection. In the context of global 21 cm cosmology, it offers a principled way to mitigate radio-frequency interference (RFI), particularly within the band of interest. We also discuss additional potential applications of this methodology.

• Speaker: Samuel Leeney (University of Cambridge)
• Tuesday 27 May 2025, 11:15-12:00
• Venue: Martin Ryle Seminar Room, Kavli Institute.
• Series: Hills Coffee Talks; organiser: Charles Walker.

Add to your calendar or Include in your list