Chapter Four · failure evidence
What Molecular Dynamics Simulation got wrong, from 53 dissertations
Molecular dynamics simulations frequently encounter limitations across force field accuracy, simulation timescales, and structural stability under varied physical conditions. Across these investigations, researchers find that simulated models suffer from artificial boundary effects, numerical convergence failures, and inaccurate reproduction of experimental binding or transport behaviors. These records come from PhD theses at 16 institutions, 2021 to 2026. Each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Force field parameterizations and fixed-charge models fail to capture polarization and transport properties
Simulations using general or fixed-charge force fields produced overly collapsed disordered proteins, unphysical order peaks, and inaccurate solvation dynamics or transport coefficients. Neglecting electronic polarization and dipole spillover led to severe errors including artifactual contact ion pairing, underestimated capacitance, and inverted adsorption trends.
Tried and failed
molecular dynamics with standard force fields applied to intrinsically disordered proteins. Reason: produced overly collapsed and excessively helical conformations inconsistent with SAXS and NMR data
Tried and failed
molecular dynamics simulation of interfacial solvation dynamics applied to micellar interface with small molecule additives. Reason: simulations predicted accelerated solvation dynamics contradicting experimental spectroscopic measurements
Lighting up the stage : ultrafast dynamics of the reverse micelle interface · UT Austin
Tried and failed
isotropic and reduced-arm coarse-grained molecular dynamics applied to concentrated monoclonal antibody solution structure. Reason: produced unphysical intermediate range order peaks not seen in experimental structure factors
Protein-protein interactions and rheological properties of monoclonal antibodies at high concentration · UT Austin
Tried and failed
fixed-charge molecular dynamics with expanded explicit solvation applied to protein pKa prediction. Outcome: worse than baseline. Reason: Fixed-charge parameters over-stabilized hydrogen bonds and charged states as explicit solvent shell size increased
Quantifying protein electrostatics through polarizable force field molecular simulations · UT Austin
Tried and failed
all-atom molecular dynamics with standard force fields applied to organic solvent diffusion coefficient estimation. Outcome: worse than baseline. Reason: force field parameterisation underestimated molecular mobility and transport properties relative to experimental values
Molecular Investigations on the Behaviours of Surface-Active Materials · YorkSpace
Tried and failed
classical constant-potential molecular dynamics simulation applied to electrode-electrolyte interfacial capacitance modeling. Reason: omits electronic surface dipole spillover, underestimating capacitance by a factor of two or more
Computational Electrosynthesis: The Role of the Electrical Double Layer in Anodic Organic Transformations · Georgia Tech
Tried and failed
PCFF+ force field classical molecular dynamics applied to electrolyte solvent-salt coordination behavior. Outcome: did not generalise. Reason: failed to reproduce experimental association and dissociation behavior across most solvent-salt systems
Improving Understanding of Lithium-Oxygen Batteries Using Atomistic Simulations · MIT
Tried and failed
single-chain molecular dynamics simulation applied to polyelectrolyte adsorption in electrolyte solutions. Outcome: did not generalise. Reason: single-chain model and cation hydration overestimation inverted the experimental divalent vs monovalent adsorption trends
Tried and failed
applying dielectric boundary corrections to pressure applied to electrolyte solution molecular dynamics simulations. Outcome: did not generalise. Reason: the correction failed to fix qualitative errors in osmotic coefficient predictions from local parameter sets
On the development of SAFT-γ force fields for ionic surfactants and theories of micellization · Imperial
Tried and failed
nonpolarizable molecular dynamics force fields applied to charged electrode-electrolyte interfaces. Reason: neglect of electronic polarization produced artifactual contact ion pairing at the charged interface
Tried and failed
classical molecular dynamics with general force fields applied to water diffusivity in polymers. Outcome: no signal. Reason: limited force field transferability across diverse chemical structures and interactions
Polymer informatics advancements to accelerate the design of sustainable packaging materials · Georgia Tech
Tried and failed
molecular dynamics with general force field parameters applied to small molecule torsional conformational sampling. Reason: force field parameters over-stabilized a single dihedral state, failing to sample multi-state conformations in solution
Biomolecules' conformational changes studied by simulations and enhanced sampling · Georgia Tech
Computational scaling bottlenecks and timescale limitations prevent full atomistic simulation of large systems
Atomistic and ab initio molecular dynamics simulations were frequently rejected or failed because slow phase transformations, crystallization induction, and micellization exceeded feasible timescales. High computational training costs, memory bottlenecks from storing reference configurations, and intensive virial calculations rendered large-scale simulations intractable.
Tried and failed
on-the-fly machine-learned force field molecular dynamics applied to long timescale atomistic simulations. Outcome: infeasible cost. Reason: memory scaling bottlenecks from storing and processing large reference configurations during simulation
Rapid computational screening of materials for energy storage applications · Imperial
Tried and failed
ab initio molecular dynamics simulations applied to structural phase transformations during desolvation. Outcome: too slow. Reason: transformation dynamics exceeded achievable simulation timescales
Considered and rejected
Considered and rejected: Rejected SchNet for running extensive QM/MM molecular dynamics due to high computational training/inference cost and GPU memory limitations when scaling the explicit solvent cutoff.
Considered and rejected
Considered and rejected: Rejected full atomistic molecular dynamics simulations of polymer-wrapped nanocrystals and depletants due to intractable computational cost, adopting coarse-grained bead-spring models instead.
Understanding effects of disorder on the plasmonic response of nanoparticle assemblies · UT Austin
Considered and rejected
Considered and rejected: Rejected Molecular Dynamics (MD) for calculating collision-induced temperature changes because virial theorem approximations are computationally intensive and unreliable for negligible effects.
Thermodynamics of Neurotransmission · Publikationssystem UB Tuebingen
Considered and rejected
Considered and rejected: Rejected multi-particle probe rheology in molecular dynamics simulations due to prohibitive computational costs and hydrodynamic interactions among multiple probes in accessible box sizes.
Viscoelastic properties and dynamics of cross-linked polymer network systems · Texas Tech
Considered and rejected
Considered and rejected: All-atom molecular dynamics was rejected because computational limitations prevent reaching microsecond timescales required to observe micellization.
Synthesis and self-assembly of hydrophobically modified polybetaines · Texas Tech
Considered and rejected
Considered and rejected: Finite Volume Method (FVM), Finite Element Method (FEM), and Molecular Dynamics (MD) rejected for microfluidic systems due to intensive meshing/poor cell quality in curved channels, continuum assumption failure, FSI computational demand, and prohibitive molecular computational costs
Tried and failed
molecular dynamics simulation of heterogeneous crystallization applied to polymer chain nucleation on crystalline substrates. Outcome: too slow. Reason: induction times exceeded simulation limits, leading to incomplete nucleation events within feasible compute windows
Understanding Heterogeneous Nucleation Mechanisms in Polyolefins · MIT
Considered and rejected
Considered and rejected: All-atom molecular dynamics simulations were rejected for full parameter exploration because they are too computationally expensive compared to coarse-grained MADna simulations
Investigating Poloidal Bias in DNA Minicircles and Overhang influence on DNA Dehybridization · Georgia Tech
Docked ligands and biomolecules exhibit pose instability and fail to maintain target binding modes
Refinement runs resulted in docked ligands drifting from binding sites, peptides rapidly dissociating within nanoseconds, or substrates failing to penetrate due to rigid salt bridges. Simulations also failed to reproduce intact assembly binding modes, variant classifications, and membrane insertion or lipid-dependent binding differences.
Tried and failed
molecular dynamics refinement of docked ligand poses applied to protein-ligand complex binding pose validation. Outcome: unstable. Reason: docked ligand poses drifted away from the binding site during simulation
Structure-based discovery of lipoteichoic acid synthase inhibitors. · Imperial
Tried and failed
molecular dynamics simulation of ligand binding applied to intact multi-protein complex. Outcome: did not generalise. Reason: isolated domain binding mode was not reproduced in the intact multi-subunit protein assembly
Atomistic simulations of cardiac troponin with phosphorylation andmutation effects · Imperial
Tried and failed
linear interaction energy for binding affinity estimation applied to protein-ligand complex molecular dynamics simulations. Outcome: no signal. Reason: ligand stabilized at an off-target site outside the catalytic pocket during simulation
Tracing the Evolution of Substrate Specificity in Low-Molecular-Weight Protein Tyrosine Phosphatases and Arsenate reductases · Georgia Tech
Tried and failed
molecular dynamics simulation of docked peptide poses applied to flexible extracellular protein loops. Outcome: unstable. Reason: replicas failed to converge to a single pose and the peptide dissociated within nanoseconds
The Hybrid Barrel Mechanism of Outer Membrane Protein Folding by the Bam Complex and Its Inhibition · Georgia Tech
Tried and failed
unbiased molecular dynamics and hierarchical clustering applied to protein kinase deletion variants. Outcome: no signal. Reason: simulations failed to correctly classify variants and yielded counterintuitive hydrogen-bonding and exchange properties
Tried and failed
multi-replica conformational population coarse-grained molecular dynamics applied to lipid-dependent protein membrane binding. Outcome: did not generalise. Reason: increasing ensemble size failed to reproduce experimental differences in membrane binding across different lipid compositions
Tried and failed
coarse-grain molecular dynamics simulations applied to pore-forming protein prepore membrane insertion. Outcome: no signal. Reason: simulations failed to induce membrane thinning, pore formation, or significant protein-lipid interactions
Tried and failed
geometry-based tunnel prediction for transport pathways applied to transmembrane protein substrate permeation. Reason: Rigid salt bridge prevented channel opening during steered molecular dynamics simulations
Exploring molecular mechanisms with simulations and data analyses · Iowa State
Elevated temperatures and high steering forces cause unphysical structural instability and integration breakdowns
Applying aggressive pulling rates or high force constants during steered simulations triggered structural instabilities and integration failures. High temperatures and incompatible machine learning corrections led to unphysical dimer formation, premature crystal melting, chemical bond dissociation, and unphysical interlayer sliding.
Tried and failed
molecular dynamics equilibration of trans-coordinated multivalent complexes applied to polyhistidine metal-ligand coordination complexes. Outcome: unstable. Reason: trans bidentate configuration was structurally unstable during equilibration, preventing steered molecular dynamics pulling
Mechanochemical Understanding of Metal-Coordinated Polymers Using Simulation and Experiment · MIT
Tried and failed
machine learning interatomic potential with repulsive correction applied to high-temperature molecular dynamics simulation. Outcome: unstable. Reason: potential formed unphysical short-range dimers; adding repulsive screening cutoffs degraded dynamical transport properties
Theory and simulations of ionic conductors · Imperial
Tried and failed
molecular dynamics with machine learning potential applied to high-temperature high-pressure crystal synthesis. Outcome: unstable. Reason: target crystal phases transition to liquid-like states at synthesis temperatures
Tried and failed
stochastic thermostat with machine-learned force corrections applied to multiple time step molecular dynamics. Outcome: unstable. Reason: ML force-correction model was not trained on data generated with the stochastic thermostat dynamics
Advancing first principle-based molecular dynamics of biological systems with machine learning · EPFL
Tried and failed
steered molecular dynamics with high force constants applied to lipid bilayer permeation simulations. Outcome: unstable. Reason: excessive pulling force constants combined with high pull rates caused severe structural instability and integration failure
Tried and failed
thermal annealing equilibration in ab initio molecular dynamics applied to electrolyte solvent and salt mixtures. Outcome: unstable. Reason: high-temperature ramp caused unphysical chemical bond dissociation of salt anions and inadequate phase space sampling
Improving Understanding of Lithium-Oxygen Batteries Using Atomistic Simulations · MIT
Tried and failed
ab initio molecular dynamics at elevated temperatures applied to layered crystalline solid electrolytes. Outcome: unstable. Reason: elevated temperatures induced unphysical interlayer sliding along the basal plane
Understanding ion transport in halide solid electrolytes with varying mobile species · Imperial
Boundary conditions and equilibration density errors generate artificial transport and surface artifacts
Simulations suffered from proteins adhering to periodic boundary edges, unphysical pulling trajectories, and rapid gradient relaxations that prevented sufficient statistical sampling. In transport systems, boundary density assumptions produced spurious profiles, unimpeded plug flow eliminated density gradients, and equilibration density errors biased diffusivity predictions.
Tried and failed
gradient relaxation molecular dynamics applied to diffusiophoresis transport coefficients. Outcome: data insufficient. Reason: transient relaxation of concentration gradients prevented collecting sufficient sampling statistics
Diffusiophoresis in complex and confined fluids · Cambridge
Tried and failed
molecular dynamics of finite planar lipid bilayer applied to protein-membrane binding affinity simulation. Reason: Proteins preferentially adhered to hydrophobic periodic boundary or edge artifacts rather than the planar bilayer surface.
Characterizing & modelling the bio-nano interface · Imperial
Tried and failed
distance-based steered molecular dynamics pulling applied to periodic membrane permeation simulations. Reason: periodic boundary conditions introduced directional artefacts and unphysical pulling trajectories
Tried and failed
calculating transport coefficients using bulk reservoir densities applied to boundary-driven molecular dynamics transport simulation. Reason: leads to spurious profiles and an unphysical minimum around reduced density 0.4
Study of transport of fluids under nano-confinement using non-equilibrium molecular simulation · Imperial
Tried and failed
boundary-driven non-equilibrium molecular dynamics applied to liquid transport in smooth nanopores. Outcome: no signal. Reason: unimpeded plug flow prevented the formation of a measurable internal density gradient
Study of transport of fluids under nano-confinement using non-equilibrium molecular simulation · Imperial
Tried and failed
multistep molecular dynamics equilibration for property prediction applied to gas diffusivity in dense polymers. Reason: equilibration protocol produced lower simulated densities, causing systematic overestimation of diffusivity in low-diffusivity regimes
Polymer informatics advancements to accelerate the design of sustainable packaging materials · Georgia Tech
Ab initio simulations and unseeded self-assembly suffer from convergence failures and inaccessible target phases
Ab initio and density functional theory simulations experienced severe numerical or electronic convergence instabilities in alloy models and thin multilayer heterostructures. Additionally, unseeded self-assembly formed disordered polycrystalline aggregates instead of ordered crystals, and melt-quench procedures failed to synthesize thermodynamically inaccessible phases.
Tried and failed
Ab initio molecular dynamics applied to special quasirandom structure alloy models. Outcome: did not converge. Reason: Electronic or structural instabilities caused convergence difficulties requiring truncated simulation times
ML-accelerated pipeline for understanding atomistic hardening of MPEAs and predicting hardness in additive manufacturing · Virginia Tech
Tried and failed
density functional theory molecular dynamics applied to ultrathin heterostructure junction interfaces. Outcome: did not converge. Reason: numerical convergence was unstable during simulation of the thin multilayer structure
New materials for magnetic tunnel junction devices for neuromorphic and probabilistic computing · UT Austin
Tried and failed
direct unseeded self-assembly in molecular dynamics applied to large-scale MOF crystal growth. Outcome: did not converge. Reason: larger systems formed disordered, polycrystalline aggregates instead of ordered single crystals
MOLECULAR AND NANOSCALE SELF-ASSEMBLY DESIGN WITH COARSE-GRAINED COMPUTATIONAL MODELS · Cornell
Tried and failed
ab initio melt and quench molecular dynamics applied to synthesizing metastable crystal polymorphs. Reason: phase remains thermodynamically inaccessible across temperatures and pressures despite dynamic stability
Theoretical investigation of III-V and metal oxide compounds · UT Austin
Left open by the authors
Problems the authors named and did not get to.
Left open
Model and simulate interactions between water molecules and functional hydrogel polymers, particularly large molecules, using molecular dynamics simulations. Blocker: Lack of specific modeling methodology, polymer configurations, or quantitative target metrics defined in the thesis
Left open
Diagnose constructed Madelung electrostatic potential profiles beyond the Debye mode, incorporating non-local dynamical quantities from molecular dynamics simulations. Blocker: Lacks specific diagnostic metrics, test systems, or benchmark datasets defining successful evaluation beyond the Debye mode.
Molecular Optimization for Classical and Quantum Condensed Phase Systems · MIT
Left open
Investigate solvation structure changes and cation coordination in P(DXL-Me0.20) at r = 0.16 to explain the local current fraction maximum. Blocker: Requires experimental synthesis/characterization equipment or advanced molecular dynamics validation without provided parameters.
RATIONAL DESIGN OF HIGH-PERFORMANCE AMORPHOUS POLYMER ELECTROLYTES · Cornell
Left open
Run extended molecular dynamics simulations to analyze long time scale heterogeneous dynamics and activated cage escaping in supercooled liquids. Blocker: Lacks specific simulation parameters, target liquid systems, and precise quantitative milestones
Structural and Dynamic Heterogeneities in Supercooled Liquids · Iowa State
Left open
Simulate subgranular cell sizes (0.5-1 µm) at lower strain rates across multiple crystallographic texture orientations using molecular dynamics. Blocker: Extreme computational cost of reaching experimental strain rates and micro-scale domain sizes in molecular dynamics.
Process-Structure-Property Relationship Study of Selective Laser Melting using Molecular Dynamics · Virginia Tech
Left open
Refine GraphCMC graph neural network models on high-loading configurations generated via ab initio molecular dynamics, training on adsorption energy instead of total energy. Blocker: Requires performing computationally intensive ab initio molecular dynamics (AIMD) simulations to generate training data.
Adsorption Isotherm Prediction of Diverse Adsorbates in Metal–Organic Frameworks using Machine Learning · Georgia Tech
Left open
Run molecular dynamics simulations of asymmetric lipid bilayers to evaluate the theoretical relationship κeff = κ0 - γ|ζ|. Blocker: None
Left open
Model lipid nanoparticle fusion energetics and endosomal escape bottlenecks using coarse-grained molecular dynamics simulations and potential of mean force calculations. Blocker: None
A multiscale computational framework for targeted delivery of Nanoparticles · Penn
Left open
Perform molecular dynamics simulations and membrane-mimicking characterization on the four-subunit Aβ(1-42) oligomer model to evaluate pore-forming insertion. Blocker: Requires unpublished structural coordinates from the thesis and experimental wet-lab facilities for membrane characterization.
Use solid-state NMR to study the molecular structures of disease-associated peptide aggregates · Georgia Tech
Left open
Validate the image-charge electrosorption isotherms using molecular dynamics simulations of electrolyte solutions at electrified interfaces. Blocker: None
Nonlocal correlation effects in water and aqueous electrolyte solutions at electrified interfaces · Imperial
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.