Intelligent tutoring system (ITS) provides learners with step-by-step problem-solving support through scaffolding. Most ITSs have been developed in the USA and incorporate American instructional strategies. How do non-American students perceive and use ITS with different native problem-solving strategies? The present study compares Stoich Tutor, an ITS with a high level of scaffolding, with ORCCA, an ITS with dynamic scaffolds that can support a range of problem-solving strategies. We conducted a think-aloud study with university students in the USA (N = 10) and Germany (N = 11), where students worked with either Stoich Tutor and ORCCA before solving stoichiometry problems on paper. Two human coders derived a coding scheme to investigate the strategies American and German students employ during problem solving on paper without instructional support. We derive a taxonomy of three stoichiometry problem-solving strategies. Next to the American factor labeling method, this taxonomy includes a strategy based on equation transformations and a previously undocumented strategy using abstract symbols to isolate a target variable and then pluck in given values and compute the solution. German students exclusively used the latter strategy, which was not explicitly supported by any of the two tutoring systems. Further, students who did not use the factor-label method for paper-based problem solving, most of whom were German, initially had difficulty setting appropriate goals and working with fractions in the Stoich Tutor. While German students preferred ORCCA based on short interviews, they more often successfully solved problems in Stoich Tutor. Therefore, Stoich Tutor, although misaligned with German instruction, could still support German students’ learning. Still, revisions to ITS based on local instructional cultures could make them potentially more effective and aligned with curricular goals.

Neural network potentials (NNPs) have achieved impressive accuracy on local, short-range chemical environments, yet their ability to capture long-range electronic effects remains unclear. We investigate this question using a controlled testbed for electron delocalization: 1D tight-binding chains with 2–10 atoms. We generated synthetic datasets by drawing nearest-neighbor coupling values from a narrow normal distribution and determining the resulting ground-state energies using Hamiltonian diagonalization. We benchmark two architectures: (1) a “full chain” feed-forward MLP that ingests the entire coupling vector and directly predicts the total energy, and (2) a Behler–Parrinello–style “atom environment” model that sums per-atom energies from local windows (sizes 1–3). Across chain lengths, the full-chain model consistently attains lower error and cleaner predicted-vs-true alignments, though accuracy degrades with system size for all models. For 10-atom chains, mean errors are 0.0279 (full-chain) versus 0.0613, 0.0524, and 0.0417 for window sizes 1, 2, and 3, respectively. Loss curves show stable optimization with limited overfitting on small and mid-sized systems. These results indicate that even simple global-feature MLPs can learn delocalization effects that challenge purely local descriptors. Key limitations include the idealized 1D tight-binding ground truth, nearest-neighbor couplings only, and the absence of symmetry/equivariance constraints, which limit transferability to realistic 3D chemistry. Nonetheless, the findings motivate combining global receptive fields (or longer range descriptors) with physics-aware architectures to improve generalization in delocalized electronic systems.

Dechlorination is one of the main processes for the natural degradation of polychlorinated biphenyls (PCBs) in an anaerobic environment. However, PCB dechlorination pathways and products vary with PCB congeners, types of functional dechlorinating bacteria, and environmental conditions. The present study develops a novel model for determining dechlorination pathways and fluxes by tracking redox potential variability, transforming the complex dechlorination process into a stepwise sequence. The redox potential is calculated via the Gibbs free energy of formation, PCB concentrations in reactants and products, and environmental conditions. Thus, the continuous change in the PCB congener composition can be tracked during dechlorination processes. The new model is assessed against four measurements from several published studies on PCB dechlorination. The simulation errors in all four measurements are calculated between 2.67 and 35.1% under minimum (n = 0) and maximum (n = 34) numbers of co-eluters, respectively. The dechlorination fluxes for para-dechlorination pathways dominate PCB dechlorination in all measurements. Furthermore, the model also considers multiple-step dechlorination pathways containing intermediate PCB congeners absent in both the reactants and the products. The present study indicates that redox potential might be an appropriate indicator for predicting PCB dechlorination pathways and fluxes even without prior knowledge of the functional dechlorinating bacteria.

Research spanning nearly a century has found that math plays an important role in the learning of chemistry. Here, we use a large dataset of student interactions with online courseware to investigate the details of this link between math and chemistry. The activities in the courseware are labeled against a list of knowledge components (KCs) covered by the content, and student interactions are tracked over a full semester of general chemistry at a range of institutions. Logistic regression is used to model student performance as a function of the number of opportunities a student has taken to engage with a particular KC. This regression analysis generates estimates of both the initial knowledge and the learning rate for each student and each KC. Consistent with results from other domains, the initial knowledge varies substantially across students, but the learning rate is nearly the same for all students. The role of math is investigated by labeling each KC with the level of math involved. The overwhelming result from regressions based on these labels is that only the initial knowledge varies strongly across students and across the level of math involved in a particular topic. The student learning rate is nearly independent of both the level of math involved in a KC and the prior mathematical preparation of an individual student. The observation that the primary challenge for students lies in initial knowledge, rather than learning rate, may have implications for course and curriculum design.

Quantum chemistry provides chemists with invaluable information, but the high computational cost limits the size and type of systems that can be studied. Machine learning (ML) has emerged as a means to dramatically lower the cost while maintaining high accuracy. However, ML models often sacrifice interpretability by using components such as the artificial neural networks of deep learning that function as black boxes. These components impart the flexibility needed to learn from large volumes of data but make it difficult to gain insight into the physical or chemical basis for the predictions. Here, we demonstrate that semiempirical quantum chemical (SEQC) models can learn from large volumes of data without sacrificing interpretability. The SEQC model is that of density-functional-based tight binding (DFTB) with fixed atomic orbital energies and interactions that are one-dimensional functions of the interatomic distance. This model is trained to ab initio data in a manner that is analogous to that used to train deep learning models. Using benchmarks that reflect the accuracy of the training data, we show that the resulting model maintains a physically reasonable functional form while achieving an accuracy, relative to coupled cluster energies with a complete basis set extrapolation (CCSD(T)*/CBS), that is comparable to that of density functional theory (DFT). This suggests that trained SEQC models can achieve a low computational cost and high accuracy without sacrificing interpretability. Use of a physically motivated model form also substantially reduces the amount of ab initio data needed to train the model compared to that required for deep learning models.

The Next Generation Science Standards encourage students to develop the reasoning and practice skills of scientists. Rather than promoting rote memorization, teachers engage students with real-world phenomena and prompt students to think critically. ChemVLab+ is a series of freely available, online modules that require students to reason like chemists by integrating core concepts with investigations in authentic contexts. Each activity reflects four key design principles as well as a novel framework for developing chemical reasoning specific to the practice of designing and interpreting experiments. The current paper describes the development and testing of the ChemVLab+ modules. A within-teacher randomized control trial study (treatment = 15 class sections with 371 students; control 14 class sections with 332 students) found that students using the ChemVLab+ modules outperformed students completing a set of active control modules that offered a more traditional instructional approach and problem-solving opportunities. ChemVLab+ students scored higher than their active control peers on 7 of 8 module quizzes and performed particularly well on open-ended response items that were focused on chemical reasoning for experiments. Analyses of demographic effects also revealed a positive interaction between BIPOC students and the ChemVlab+ intervention.

A new class of conjugated macrocycle, the cyclo[4]thiophene[4]furan hexyl ester (C4TE4FE), is reported. This cycle consists of alternating alpha-linked thiophene-3-ester and furan-3-ester repeat units, and was prepared in a single step using Suzuki-Miyaura cross-coupling of a 2-(thiophen-2-yl)furan monomer. The ester side groups help promote a syn conformation of the heterocycles, which enables formation of the macrocycle. Cyclic voltammetry studies revealed that C4TE4FE could undergo multiple oxidations, so treatment with SbCl5 resulted in formation of the [C4TE4FE](2+) dication. Computational work, paired with H-1 NMR spectroscopy of the dication, revealed that the cycle becomes globally aromatic upon 2e(-) oxidation, as the annulene pathway along the outer ring becomes Huckel aromatic. The change in ring current for the cycle upon oxidation was clear from H-1 NMR spectroscopy, as the protons of the thiophene and furan rings shifted downfield by nearly 6 ppm. This work highlights the potential of sequence control in furan-based macrocycles to tune electronic properties.

This report showcases a new type of online homework system that provides students with a free-form interface and dynamic feedback. The ORCCA Tutor (Open-Response Chemistry Cognitive Assistance Tutor) is a production rules-based online tutoring system utilizing the Cognitive Tutoring Authoring Tools (CTAT) developed by Carnegie Mellon University. In this report, we discuss the interface design and the production rules that allow for a multitude of chemistry calculations to be solved in a wide variety of ways by students. We discuss improvements of the software, already implemented or planned for the future, based on think aloud/interviews with students.

Quantum chemistry provides chemists with invaluable information, but the high computational cost limits the size and type of systems that can be studied. Machine learning (ML) has emerged as a means to dramatically lower cost while maintaining high accuracy. However, ML models often sacrifice interpretability by using components, such as the artificial neural networks of deep learning, that function as black boxes. These components impart the flexibility needed to learn from large volumes of data but make it difficult to gain insight into the physical or chemical basis for the predictions. Here, we demonstrate that semiempirical quantum chemical (SEQC) models can learn from large volumes of data without sacrificing interpretability. The SEQC model is that of Density Functional based Tight Binding (DFTB) with fixed atomic orbital energies and interactions that are one-dimensional functions of interatomic distance. This model is trained to ab initio data in a manner that is analogous to that used to train deep learning models. Using benchmarks that reflect the accuracy of the training data, we show that the resulting model maintains a physically reasonable functional form while achieving an accuracy, relative to coupled cluster energies with a complete basis set extrapolation (CCSD(T)*/CBS), that is comparable to that of density functional theory (DFT). This suggests that trained SEQC models can achieve low computational cost and high accuracy without sacrificing interpretability. Use of a physically-motivated model form also substantially reduces the amount of ab initio data needed to train the model compared to that required for deep learning models.

In this report, the synthesis and properties of an ester-functionalized macrocyclic sexifuran (C6FE) are presented.

Effective diffusion of educational innovations is essential for evidence-based practices to have broad impacts on student learning. One of the barriers to such diffusion is the large inertia associated with changing one's teaching practices. Educational disruptions, such as COVID-19, may lower this barrier by making business-as-usual no longer feasible. Here, we reflect on the ways in which an online community, designed to promote diffusion of educational innovation, supported Advanced Placement (AP) Chemistry teachers in the rapid transition to online instruction necessitated by the pandemic. The lessons learned from supporting teachers as they responded to the pandemic may provide insights into how to better promote and support adoption of reformed teaching strategies.

A high-throughput optical screening method for the photocatalytic activity of a structurally diverse library of 1152 cationic iridium(III) complexes ([Ir(C boolean AND N)(2) (NAN)](+)), corresponding to all combinations of 48 cyclometalating (C boolean AND N) and 24 ancillary (N boolean AND N) ligands, was developed. This rapid assay utilizes the colorimetric changes of a high contrast indicator dye, coumarin 6, to monitor the photo-induced electron transfer from a sacrificial amine donor to the metal complex excited state. The resulting [Ir(C boolean AND N)(2) (N boolean AND N)](0) can then reduce an aryl bromide to form the highly reactive aryl radical intermediate. The rate of this reaction is dictated by the molecular structure of both coordinating ligands. Relative reaction rate constants determined via this method correlated closely with F-19 NMR measurements obtained using a fluorinated substrate. A simple model that expresses the rate constant as a product of a single "strength" parameter assigned to each of the 72 ligands can well account for the 1152 measured rate constants. The best performing complexes exhibit much higher reactivity than the benchmark photocatalysts commonly used in photoredox transformations. The catalysts were also successfully tested for their chemoselectivity. The developed screening methodology can enable generation of the large data sets needed to use modern data science to extract structure-activity relationships.

Nanovoids in organic semiconductors can serve as hosts for water inclusions that lead to trapping of electrons and holes.

This report describes the design and synthesis of a new class of polyfurans bearing ester side chains. The macromolecules can be synthesized using catalyst-transfer polycondensation, providing precise control over molecular weight and molecular weight distribution. Such obtained furan ester polymers are significantly more photostable than their alkyl analogues owing to the electron-withdrawing nature of the attached subunit. Most interestingly, they spontaneously fold into a compact π-stacked helix, yielding a complex multilayer cylindrical nanoparticle with a hollow, rigid, conjugated core composed of the polyfuran backbone and a soft, insulating outer layer formed by the ester side chains. The length of polymer side chains dictates the outer diameter of such nanoparticles, which for the hexyl ester groups used in the present study is equal to ∼2.3 nm. The inner cavity of the conjugated core is lined with oxygen atoms, which set its effective diameter to 0.4 nm. Furthermore, installation of bulkier, branched chiral ester side chains on the repeat unit yields structures that, upon change of solvent, can reversibly transition between an ordered chiral helical folded and disordered unfolded state.

Controlling the sequence of repeat units in a synthetic polymer has been a long-standing topic of interest in chemistry. As methods to regulate sequence become more sophisticated, it is critical to consider how controlling the arrangement of repeat units along the polymer backbone impacts properties. In this work, thiophene-selenophene copolymers (statistical and periodic) were compared to elucidate the impact of periodicity on electronic properties and structural organization in conjugated macromolecules. Polymers were synthesized using catalyst-transfer polycondensation (CTP) enabling control over molecular weight and dispersity. The study revealed that optical bandgaps and redox potentials of periodic and statistical copolymers varied with composition in a predictable manner, regardless of monomer ordering along the chain. While the bandgaps of the two types of copolymers were indistinguishable, X-ray scattering revealed differences in solid-state packing. Both types of copolymers exhibited well-defined morphologies, but larger pi-stacking distances and more orientational disorder were evident in the statistical systems. This indicates periodicity is an attribute that should be considered when synthesizing semiconducting materials.

We devise a novel technique to control the shape of polymer molecular weight distributions (MWDs) in atom transfer radical polymerization (ATRP). This technique makes use of recent advances in both simulation-based, model-free reinforcement learning (RL) and the numerical simulation of ATRP. A simulation of ATRP is built that allows an RL controller to add chemical reagents throughout the course of the reaction. The RL controller incorporates fully-connected and convolutional neural network architectures and bases its decision upon the current status of the ATRP reaction. The initial, untrained, controller leads to ending MWDs with large variability, allowing the RL algorithm to explore a large search space. When trained using an actor-critic algorithm, the RL controller is able to discover and optimize control policies that lead to a variety of target MWDs. The target MWDs include Gaussians of various width, and more diverse shapes such as bimodal distributions. The learned control policies are robust and transfer to similar but not identical ATRP reaction settings, even under the presence of simulated noise. We believe this work is a proof-of-concept for employing modern artificial intelligence techniques in the synthesis of new functional polymer materials.