DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Application of cold spray to manufacturing of solid-state batteries
No full textMay2025School of EngineeringAs current liquid electrolyte lithium-ion batteries reach their theoretical limit, exploration of solid-state batteries (SSBs) as an energy dense and safe alternative continues to expand. Due to many decades of research the stage is being approached where the main hurdle is how to manufacture high performance batteries at scale. Tantalum doped Lithium Lanthanum Zirconium oxide (LLZTO) is a solid electrolyte material of much interest, due to its high ionic conductivity and stability with lithium metal, which faces manufacturing challenges to its high sintering temperature. It is also commonly mixed with lithium cobalt oxide (LCO), a much-used cathode material, to increase ionic transport. In this study cold spray (CS), a kinetic spraying technique with a low process temperature capable of producing dense films, is investigated for its ability to create dense films of LLZTO and LCO. Single layer dense films of LLZTO with a thickness of around 12 µm were produced on an aluminium substrate. Attempts to produce a thicker film by depositing further layers of LLZTO were unsuccessful and caused large surface perturbations which allowed aluminium to come to the surface. LCO was likewise successfully deposited on aluminium and was able to form a 1µm thick film on Alumina. Both films experienced changes in crystallographic structure during deposition, and LLZTO displayed an increase in lattice strain which could affect electrochemical performance. Despite the immaturity of using this technique with solid state batteries, this study shows that with further development, cold spray could become a viable option to produce SSBs at large scale.M
Influence of surface confinement and curvature on the formation of loop brushes
No full textMay2025School of EngineeringPolymer chain conformation and behavior under geometric confinement have been widely explored due to their importance in applications such as nanotechnology, biomaterials, and surface functionalization. If long flexible macromolecule is covalently anchored on a substrate surface, polymer brush is obtained. Depending on molecular weight of the polymer and the number of grafting sites (graft density), surface properties such as adhesion, wetting, and lubrication can be tuned. One acquires loop brushes either by reating both ends of the chain to the substrate or reacting the free ends of grafted chains. While extensive studies have focused on linear polymer brush behavior on various substrates as a function of curvature, formation of loops and loop chain conformation as a function of curvature remains less explored. In the current study, the static properties of linear polymer chains grafted to flat and spherical surfaces and their tendency to form loop grafts were investigated using Monte Carlo simulations. The end-to-end distance (R), radius of gyration (Rg), distance between grafting sites (S), and distance between the free ends of the grafted chains (L) were analyzed to understand how confinement by an impenetrable surface and its curvature influences grafted chain conformation, providing insights into the probability of forming loop brushes on the surface.Our simulations reveal distinct power law scaling behaviors of polymer conformations influenced by surface curvature: the power scaling exponent for R decreases from 0.48 for polymers grown in free space, to 0.45 for polymers grafted onto a flat surface, and further reduces to 0.42 for polymers grafted onto a spherical surface, reflecting increased geometric constraint on polymer conformation due to surface curvature. Additionally, surface curvature significantly impacts the orientation of grafted polymers, with average angles between chain end-to-end vectors being ap- proximately 62 − 65◦ on a flat surface and 90◦ on a spherical surface. This difference in the average angle affects the distance between graft chain ends, L, explaining why polymer chains on spherical surfaces display greater average chain end separation, ⟨L⟩, despite having smaller ⟨R⟩ values.
An unexpected observation is the appearance of bimodal distributions in L histograms of short polymer chains confined to a spherical surface. It is believed that this bimodal distribution is an artifact of the way L is calculated. While L is calculated as the shortest distance between graft chain ends, S is calculated as the arc length between grafting sites. At low chain lengths, the end-to-end distance of the grafted chain is comparable to the radius of the spherical surface or less than it and the L vector might span through the sphere itself and become artificially smaller than the corresponding value of S. This is especially true when the grafting sites are far apart from each other – at large S values.
These findings advance our understanding of how geometric confinement influences polymer conformations and provide a basis for future investigations into loop brush formation and polymer surface interactions relevant to nanotechnology and biomaterials engineering.M
A Community-driven vision for a new Knowledge Resource for AI
No full textThe long-standing goal of creating a comprehensive, multi-purpose knowledge resource, reminiscent of the 1984 Cyc project, still persists in AI. Despite the success of knowledge resources like WordNet, ConceptNet, Wolfram|Alpha and other commercial knowledge graphs, verifiable, general-purpose widely available sources of knowledge remain a critical deficiency in AI infrastructure. Large language models struggle due to knowledge gaps; robotic planning lacks necessary world knowledge; and the detection of factually false information relies heavily on human expertise. What kind of knowledge resource is most needed in AI today? How can modern technology shape its development and evaluation? A recent AAAI workshop gathered over 50 researchers to explore these questions. This paper synthesizes our findings and outlines a community-driven vision for a new knowledge infrastructure. In addition to leveraging contemporary advances in knowledge representation and reasoning, one promising idea is to build an open engineering framework to exploit knowledge modules effectively within the context of practical applications. Such a framework should include sets of conventions and social structures that are adopted by contributors
Characterizing the conformational landscape of arf1 and mutant altering equilibria
No full textAugust2025School of ScienceThe adenosine diphosphate (ADP) ribosylation factor (Arf) family of small guanosine triphosphatases (GTPases) function as molecular switches regulating membrane dynamics and intracellular trafficking in eukaryotic organisms. Among this family, Arf1 serves as the founding member, offering a structurally and functionally conserved framework for investigation into their regulatory mechanisms. Central Arf1’s switch-like behavior is the regulated exchange of guanosine diphosphate (GDP) for guanosine triphosphate (GTP), a process that requires an extensive reorganization within its central -sheet and switch regions. To elucidate the structural and energetic basis of this switch mechanism, we combined pressure perturbation with nuclear magnetic resonance spectroscopy (NMR), Fourier Transform infrared spectroscopy (FTIR), and small-angle X-ray scattering, and computational modeling. Application of high hydrostatic pressure revealed that Arf1 populates a molten globule ensemble of partially unfolded conformers characterized by retained secondary structure but increased conformational flexibility. Using Arf1 as a model, we characterized energetic contributions of its autoinhibitory N-terminal helix. Truncation of the first 17 residues destabilizes the GDP-bound state promoting pressure-stabilized conformations that enhance nucleotide exchange, indicating a global stabilizing role for the helix. To further probe determinants of conformational activation, we introduced a corollary mutation from Arf6, the most distantly related Arf isoform, substitution isoleucine 42 with serine (I42S). In both context, this substitution modulated the energetic landscape and shifted the distribution of accessible conformational state within the Arf1 native state ensemble in a pressure and pH- dependent manner. Together these findings establish Arf1 as a reference system for investigating the wide functional divergence that exists within the Arf family of small GTPases.Ph
Evaluation of deep-learning speech recognizer in reverberation
No full textAugust2025School of ArchitectureIn the last decade, automatic speech recognition (ASR) systems have improved dramatically,even achieving human parity or superhuman ability for recognition accuracy in some cases.
This is due in part to the widespread adoption of deep-learning techniques. As this technology
continues to better approximate the human auditory system, it becomes more valuable
as a stand-in for human test subjects. In this study, OpenAI’s Whisper, an industry-leading
ASR system with deep learning architecture is assessed in reverberation in an effort to map
predictive speech intelligibility measures onto measured intelligibility, at a scale too big for
running listening tests with humans. Various measured and synthetic room impulse responses
with different levels of early and late reverberant energy are used to test the ASR
systems. These impulse responses are assessed on various established speech intelligibility
prediction approaches used in the architectural acoustics science, including the popular
speech transmission index (STI) and clarity (C50). The word error rates (WERs) produced
by the ASR system give an objective measurement of recognizer accuracy, and allow a step
toward the unification of these predictive metrics onto a much more understandable common
scale. For the measured, binaural room impulse responses, WERs were shown to correlate
very strongly with reverberation time (r_left = 0.9979, r_right = 0.998 and STI (r_left = −0.9783,
r_right = −0.9669). The correlation with clarity C50 was relatively weaker, but still strong
(r_left = −0.7841, r_right = −0.7509). In the synthetic room impulse response test, WERs
correlated less strongly with reverberation time (r = 0.7524), but very strongly with clarity
and STI (r = −0.9896, r = −0.9545).M
Improving Tabular Reusability Through Data Dictionary Descriptions
No full textTables have become a ubiquitous standard for capturing, storing, and sharing data on the web. This is primarily due to the semi-structured nature of tables, where relationships between data are often ambiguously encoded using locality. While this format can be easy for humans to interpret in simple cases, as table complexity increases, so does the difficulty in interpretability. To bridge this context gap, many data publishers provide a data dictionary to capture schema elements' meaning through text descriptions. Existing work compounds the need for data dictionaries to improve tabular interoperability, but few provide detailed requirements for data dictionary descriptions. This paper identifies and defines three common types of data dictionary descriptions in the biomedical domain. We then compare the effectiveness of each description type by normalizing data dictionary descriptions to a single type using large language models and measuring their performance using a semantic tabular interpretation algorithm. Our experiments show that intensional descriptions, which describe the general properties a column member should have, are most effective for tabular alignment and improve the reusability of data dictionaries
Energy and eigenmode analyses in coupled fluid-structure problems using immersed methods
No full textAugust2025School of EngineeringFluid-structure interaction (FSI) simulations often suffer from spurious energy gain or loss and numerical instabilities when the fluid and solid domains are discretized on non-matching meshes. Motivated by the need for robust and energy-consistent coupling schemes in non-boundary-fitted numerical methods, this work presents a comprehensive theoretical and computational framework for analyzing energy behavior and eigenmode in FSI problems using the modified Immersed Finite Element Method (mIFEM). The first part of this work develops a rigorous energy analysis of the mIFEM formulation, deriving both semi-discrete and fully-discrete energy balance that quantify the interplay among kinetic, potential, and numerical energy across non-conforming meshes. A penalty-based interface enforcement strategy is introduced, shown to be dimensionally consistent, and capable of reducing interface velocity errors to solver tolerance with a single tunable parameter, without injecting spurious energy. To quantify transfer errors arising on overlapping meshes, a general bound is established, and a variational projection strategy is proposed. The study compares energy transfer schemes that including nodal interpolation and variational projection to identify sources of conservation error. Numerical verifications confirm the accuracy of the derived energy balance and demonstrate the fidelity of energy exchange in both uncoupled and coupled configurations. In the second part, the first eigenmode analysis of mIFEM is performed for a simplified coupled viscous-elastic system. A coupled eigen-system is derived for a simplified two-dimensional configuration, rigorously enforcing both traction and kinematic continuity. An adaptive argument principle algorithm is implemented to detect unstable roots; grid convergence studies show that it reliably detects all roots while remaining insensitive to contour placement. Numerical studies demonstrate that the mIFEM formulation, in the idealized high penalty limit, shows no unstable eigenmodesfor a broad range of density ratios, thereby providing the theoretical foundation for future eigen analyses of discretized systems. Together, the energy analysis and eigen-system developed in this work furnish a diagnostic suite for assessing algorithmic correctness, guide the design of strictly conservative transfer operators, and open a pathway toward provably stable, high-fidelity simulations of complex FSI phenomena using mIFEM.DEn
Efficient post-processing techniques for foundation models
No full textMay2025School of ScienceFoundation models, such as large language models (LLMs) and large vision models (LVMs), are often fine-tuned for various downstream tasks. Popular fine-tuning techniques include Reinforcement Learning from Human Feedback (RLHF) and supervised fine-tuning (SFT). Fine-tuning approaches, including RLHF and SFT, involve modifying model parameters, which is computationally expensive and may lead to unintended degradation of model performance, such as overfitting to specific biases or reducing response diversity. \noindent This thesis addresses three key challenges in the robustness and efficiency of foundation models:\begin{itemize}
\item Learning interpretable disentangled representations (Chapter \ref{chap:pisco}): We propose a computationally cheap post-processing technique to separate style and content features in LVMs, leading to representations that improve out-of-distribution (OOD) generalization.
\item Efficiently aligning LLMs (Chapter \ref{chap:aligners}): We introduce a lightweight pipeline that generates synthetic data to train aligners and inspectors, enabling on-demand alignment of any LLM. Inspectors are lightweight BERT models used to determine when an output needs to be aligned, and aligners are small LLMs used to perform alignment with respect to a particular targeted human value. We use AlpacaEval and PairRM, two standard automatic LLM evaluators, to establish the competitive advantages of our approach against baseline LLMs and alternative alignment procedures. \item Mitigating style-induced prompt brittleness in LLMs (Chapter \ref{chap:mof}): We present a novel mixture of formats (MOF) prompting strategy that incorporates diverse stylistic variations in few-shot examples, enhancing robustness across different models and task domains. MOF uses simple modifications to few-shot prompting, so it avoids the cost of post-processing of the LLM outputs that is incurred by other approaches to mitigating style-induced prompt brittleness.\end{itemize}
Each of these approaches is designed as a post-processing technique, meaning they do not require modifying the original foundation model’s parameters. As a result, they are efficient and they provide performance improvements that are competitive with or exceed that of alternative solutions to these challenges.Ph
Artificial impulse responses for emulating decca trees
No full textMay2025School of ArchitectureABSTRACTThis research explores the generation of artificial impulse responses designed to emulate
the acoustical properties of the Decca Tree microphone array, a widely adopted method in
classical and film music recording known for its immersive and spatially cohesive sound characteristics.
A series of Dirac impulses was systematically created, incorporating calculated
time delays and sound pressure level adjustments to represent realistic microphone capture
scenarios. These impulses were then combined with various early reflection and diffuse tail
algorithms to produce authentic spatial audio simulations.
The thesis presents three listening tests conducted to evaluate the efficacy of these artificially
generated impulse responses in emulating the acoustic characteristics captured by traditional
Decca Tree setups. Results showed a clear listener preference for Cinematic Rooms
early reflections over Valhalla Rooms algorithmic reverb, particularly highlighting moderate
gain adjustments. Additionally, the research demonstrated listeners’ nuanced sensitivity to
distance-dependent processing, revealing that while detailed, custom-tailored reflections and
tails were generally favored, simpler uniform reflections remained competitively preferred by
attentive listeners. Overall, this study validates a methodological approach for convincingly
emulating the revered Decca Tree microphone technique in spatial audio production.M
Anticipatory control of steering through multiple waypoints
No full textMay2025School of Humanities, Arts, and Social SciencesMany skilled locomotor tasks require steering through complex environments at speed, moving smoothly from one waypoint to the next while avoiding obstacles. The existing body of literature examining the ability of humans to navigate complex environments leaves critical open questions about the control strategy that governs this behavior and, specifically, the role of information from multiple future waypoints. We performed two experiments using a simulated drone-flying task with different vehicle dynamics to empirically address these questions. When the drone dynamics simulated low agility, subjects made a large deviation in their trajectory, veering away from the immediate waypoint (WPN) in the opposite direction from the next future waypoint (WPN+1) before turning back. In the experiment with high agility, subjects steered more directly to WPN, turning toward WPN+1 just before passing WPN. Subjects adapted their trajectories differently under different drone dynamics, supporting an affordance-based control strategy. Within each experiment, we also manipulated the angle, distance, and orientation of WPN+1. Subjects’ speed and the shape of their trajectory were strongly affected by both angle and distance of WPN+1 in a manner modulated by agility. The results from this study provide important insights into the control strategy used to guide human steering, including the specific role of WPN+1, that can be used to inform development of a model of steering through multiple waypoints.M