A preprint comparing two computer-generated glass models reports that the version made from stiff three-atom molecules had higher elastic moduli, a larger low-frequency proxy for the reach of quasilocalized vibrations, and stronger relative coupling between nonlinear plastic modes and volume-changing deformation than a matched atomistic glass. The pattern is an association within the simulations, not evidence that the constraints cause the same behavior in real materials.
Two glass models, built to be comparable
The molecular system used three-atom molecules with stiff intramolecular FENE bonds. Its comparator was a 50:50 binary atomistic glass using the same intermolecular Lennard-Jones potential. The main systems contained 1,372 molecules or 1,372 atoms, respectively; the quasilocalized-vibration comparison used 10,000 relatively small glasses of each type.
The molecular model was stiffer
On basic elastic measures, the molecular model’s shear modulus was 12.9 versus 10.3 for the atomistic model, and its bulk modulus was 66.3 versus 48.9. The reported ratios were about 1.25 and 1.35, respectively. Poisson’s ratio was nearly unchanged at 0.41 versus 0.40, while the study’s mechanical-disorder index χ, calculated from shear-modulus fluctuations, was 5.1 versus 4.0.
Simulated waves pointed in the same direction. The molecular model’s shear-wave speed was 3.60 versus 3.23, and its longitudinal-wave speed was 9.07 versus 7.90. The fitted dispersion lengths were 12.1 versus 10.3 for shear waves and 15.0 versus 13.5 for longitudinal waves; the shear-length ratio was about 1.17.
Vibrations split into distinct bands
The molecular glass showed a three-part vibrational density of states, or VDoS—the distribution of vibration frequencies. It included low-frequency translation and rotation without deformation, followed by intramolecular-deformation bands near 40 and 55. The lower band had about twice the spectral weight of the higher one. This analysis used 100 independent realizations per model.
At low frequencies, the study’s QLV spatial proxy Nₑ—an estimate of how widely a quasilocalized vibration extends—leveled off near 120 for the molecular model and 70 for the atomistic model. Its minimum values were about 5 and 2, respectively, giving a plateau ratio of about 1.7.
A separate estimate based on the stiffness of the response to a force dipole put the characteristic QLV frequency roughly 10% lower in the molecular model. The corresponding values were 2.85 versus 3.15. The molecular estimate for QLV core length was 7.92 versus 6.45, while the QLV density parameter N was 1.1 versus 1.85, about 40% lower; the VDoS prefactor Aᵍ was 6×10⁻³ in both models.
The clearest contrast appeared in plastic modes
The analysis then examined nonlinear plastic modes, or NPMs, simulated patterns used to study plastic deformation. It selected one low-energy NPM per glass sample, producing 10,000 NPMs per model. The low-frequency spatial proxy leveled off near 110 for molecular modes and 60 for atomistic modes.
The sharper difference appeared in Rds, the study’s dilation-to-shear measure. The molecular distribution peaked near 0.17, while the atomistic distribution peaked at 0; mean Rds was about 0.209 versus 0.076, a factor of about 2.7. Because Rds is a mode-coupling metric, this does not by itself show that either model would undergo more volume-changing plastic deformation.
Not every mode property differed. The molecular and atomistic planarity-index distributions largely overlapped, and the study found no correlation between Rds and planarity index. The authors concluded that molecular constraints did not substantially affect NPM planarity.
A correction narrows the gap
The authors interpret the apparent mechanical-disorder differences through effective degrees of freedom—the independent ways a system can move. They frame the rigid molecular system as having 2N effective degrees of freedom, versus 3N for the atomistic model. Their proposed correction brought several estimates closer: χ was about 4.15 versus 4.0, QLV density was 1.65 versus 1.85, and QLV spatial extent was 80 versus 70. These are author-proposed corrections rather than independently validated estimates.
What the simulations leave unresolved
The evidence is limited to two matched athermal computer-glass models. The stiff few-atom molecular setup does not test polymeric or conformationally rich glasses, and the study does not directly validate behavior in experimental materials.
The reported comparisons did not include confidence intervals, error bars or inferential tests for most outcomes. The study identifies longer and more flexible molecules, different intramolecular-to-intermolecular bond-length ratios and direct deformation simulations as open tests of whether these patterns extend beyond the models examined here.
Paper data and sources
Original title: Effect of molecular constraints on vibrational and quasilocalized excitations in glasses
Authors: Keane Ramdin, Edan Lerner
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text