Research
Publications¶
Things I worked on over the years. Full citation details on Google Scholar.
2025¶
Shaping a collaborative, sustainable, accessible, and reproducible future for computational modeling
Summary
Presents o2S2PARC, an open-source cloud platform built under the NIH SPARC program for running computational models collaboratively. The platform handles everything from electromagnetic-electrophysiology simulations of neural interfaces to personalized brain stimulation planning, AI training, and multi-objective optimization. The main pitch: make computational modeling FAIR (findable, accessible, interoperable, reusable) so researchers can actually reproduce each other's work instead of rewriting pipelines from scratch.2022¶
Summary
We look at what happens to a Gaussian wave packet in a harmonic oscillator when you add a disordered perturbation to the potential. In the clean harmonic case, the coordinate-space and momentum-space Shannon entropies are strictly periodic and have analytical expressions. Their sum gives an entropy tied to the position-momentum uncertainty. Disorder breaks that periodicity — the wave packet disperses, the quantum probability density starts resembling a classical oscillator distribution, and the entropy grows. But here's the interesting part: at later times, depending on how strong the disorder is and on the initial conditions, quantum revivals show up. The wave packet reassembles, the entropy drops back toward its starting value, and the information that seemed lost is recovered. This revival behavior is purely quantum mechanical — there is no classical counterpart.Non-standard computational approaches applied to molecular systems — PhD Thesis
Summary
This thesis covers several threads of computational chemistry work tied together by the need for methods that go beyond routine DFT. Four studies deal with biradical-type molecules — short-lived species whose open-shell electronic structure makes geometry prediction tricky. Getting the geometry right turns out to be essential for matching computed absorption spectra to experiment, and multi-reference methods are often the only way to get there. A separate investigation looks at how geometry relaxation affects exciton properties in perylene bisimide dimers used in organic semiconductors; the common approximation of ignoring geometry turns out to break down in specific regimes where it materially changes exciton localization and energetics. The TAD-tyrosine bioconjugation mechanism is also examined, with multi-reference and DFT calculations pointing to a base-mediated electrophilic aromatic substitution as the operative pathway. The final and largest part tackles the k-nearest-neighbor entropy estimator. This method should, in principle, give accurate entropy values even for flexible biomolecules with many structural minima, but in practice it converges painfully slowly with the number of molecular dynamics snapshots. A detailed investigation traces the problem to uneven variance across vibrational modes and proposes a rescaling fix that restores fast convergence.2021¶
Femtosecond dynamics of diphenylpropynylidene in ethanol and dichloromethane
Summary
Diphenylpropynylidene (C₆H₅–C₃–C₆H₅) is a carbon chain biradical with a triplet ground state, confirmed here by EPR at 4 K. We generated it from a diazo precursor with UV light and then tracked what happens on femtosecond timescales using transient absorption spectroscopy. In deaerated dichloromethane, the dynamics are relatively clean — you can pull out time constants for vibrational cooling and electronic relaxation. Add oxygen and things get more complicated: the biradical reacts with O₂ through a pathway we modeled computationally. In ethanol, the spectra with and without oxygen look nearly identical, which turns out to be because the reaction products have similar chromophores.2020¶
Methylbismuth: an organometallic bismuthinidene biradical
Summary
First spectroscopic characterization of methylbismuth (BiCH₃), a bismuthinidene biradical generated by pyrolysis of trimethylbismuth. We nailed down a Bi–C bond dissociation energy of 210 ± 7 kJ/mol — well above what people had previously assumed. Despite that, homolytic cleavage of Me–BiMe₂ bonds turns out to be feasible at surprisingly moderate temperatures (60–120 °C) in solution, which means both BiMe₂ and BiMe are accessible as reactive intermediates. This opens up organobismuth chemistry that was previously thought to require much harsher conditions.Geometry relaxation-mediated localization and delocalization of excitons in organic semiconductors
Summary
In organic optoelectronic materials, excitons can get trapped through geometry relaxation — monomers shift relative to each other (inter-monomer) or distort internally (intra-monomer). Most computational studies ignore the intra-monomer part because it's expensive to calculate. Using perylene-bisimide dimers and diindenoperylene tetramers as test cases, we showed that intra-monomer distortions actually matter a lot for the shape of the potential energy surface. The effect depends strongly on which electronic state you're looking at — first vs. second excited state behave quite differently. Delocalization of the exciton over more monomers also reduces the magnitude of these geometry relaxation effects.2019¶
Summary
We measured the photoelectron spectra of three triplet carbene carbon chains — pentadiynylidene (HC₅H) and its methyl-substituted variants — using VUV synchrotron radiation at the Swiss Light Source. The unsubstituted species has an adiabatic ionization energy of 8.36 ± 0.03 eV. We also found a surprise: a cyclic isomer (3-(didehydrovinylidene)cyclopropene) with a singlet ground state, identified at 8.60 ± 0.03 eV. This cyclic isomer is interesting computationally because standard coupled-cluster theory badly overestimates its ionization energy — you need multireference methods (CASPT2) to get it right.Summary
Tolane (diphenylacetylene) was studied in a supersonic jet with picosecond time-resolved photoionization and photoelectron imaging. Near the S₁ origin the excited state lives for nanoseconds, but at higher excitation energies the lifetimes drop to tens of picoseconds. We observed a two-step relaxation into a long-lived electronic state and, comparing with quantum chemical calculations, identified it as the trans-bent biradicalic ¹Aᵤ state. This state had been predicted computationally but never directly observed before.2018¶
Using data analysis to evaluate and compare chemical syntheses
Summary
We built ChemPager, a freely available tool for systematically evaluating chemical syntheses. It processes and visualizes data that chemists already collect — yields, purities, batch records — and turns them into actionable metrics like process mass intensity, volume-time output, and production costs. Beyond the standard indicators, we introduced scoring functions for harder-to-pin-down qualities: process robustness, design quality, and safety. The robustness score, for instance, compares lab results against plant outcomes, tracks batch-to-batch variation, flags unexpected events during manufacturing, and penalizes unusually strict process requirements. Everything sits in a hierarchical data layout built on Excel or Google Sheets for entry and Spotfire for visualization, so adopting it doesn't require new infrastructure. Keeping all project data in one place makes cross-project comparison straightforward and lets you link out to other data sources when needed.Mechanistical insights into the bioconjugation reaction of triazolinediones with tyrosine
Summary
PTAD-tyrosine bioconjugation is one of those reactions people use all the time for selective protein functionalization without fully understanding why it works. Four plausible mechanisms exist: electrophilic aromatic substitution (SEAr), an ene reaction, a stepwise radical path, and a higher-order concerted pericyclic route. We ran high-level quantum chemistry calculations on all four for the reaction of HTAD with p-cresol and p-cresolate in water. The SEAr and ene barriers are comparable but still too high to explain observed reaction rates — until you account for the cresolate anion. With the deprotonated phenol the SEAr barrier essentially vanishes, which explains both the fast conversion rates seen experimentally and the strong pH dependence reported in the literature.The ortho-benzyne cation is not planar
Summary
Earlier photoelectron spectra of ortho-benzyne had some suspicious bands that people attributed to side products, and there were lingering doubts about whether the cation is actually planar. We settled it: it's not. Using mass-selected threshold photoelectron spectroscopy with synchrotron radiation at the Swiss Light Source, combined with CASPT2(11,14) calculations, we showed the ortho-benzyne cation adopts a twisted C₂ geometry rather than the assumed planar C₂ᵥ structure. The adiabatic ionization energy comes out to 9.51 eV experimentally (9.56 eV computed). A Franck–Condon simulation covering three electronic states of the cation matches the measured spectrum well, which required going beyond single-reference methods.2016¶
Adiabatic and non-adiabatic electron-nuclear motion: Quantum and classical dynamics