Flying ice cube effect¶
A molecular-dynamics artifact in which offending velocity rescaling shifts kinetic energy toward lower-frequency motion and away from the expected partition among modes.
Core Idea¶
The flying ice cube effect is an error in some molecular-dynamics simulations: a velocity-rescaling thermostat shifts kinetic energy toward lower-frequency motion instead of preserving the appropriate energy partition among the simulated molecule's active modes. The overall temperature can still look controlled. The error may change properties calculated from the trajectory.[^ref-8d8585786a8a]
Harvey and colleagues gave the name to a severe case in which the energy concentrates in a molecule's overall translation or rotation while its internal motion is depleted.[^ref-ec81f39bf736] Braun and colleagues use the name for velocity-rescaling-caused equipartition failure more generally, including cases without visible flight.[^ref-8d8585786a8a]
Scope of Application¶
The pattern belongs to molecular-dynamics runs whose velocity-rescaling operation actually biases energy among the modes. Simple rescaling and Berendsen caused it in conditions Braun and colleagues studied; the CSVR method did not in their tested cases.[^ref-8d8585786a8a] A thermostat's name alone does not prove that a particular run has the effect. Nor does an equipartition error caused by a different numerical fault fall under this narrower source-defined mechanism.
Clarity¶
Separate a correct total temperature from a correct modal energy distribution. A thermostat can control an aggregate while translation, rotation, and vibration have the wrong shares. Do not diagnose solely from visible center-of-mass motion: Braun and colleagues found an abnormal partition in a simulated ethane system even when overall center-of-mass momentum was fixed to zero.[^ref-8d8585786a8a]
Manages Complexity¶
Braun and colleagues distinguish velocity-rescaling-caused modal bias from equipartition failures with other numerical causes.[^ref-8d8585786a8a] Three questions organize the diagnosis: Which velocity-rescaling rule was used? How was kinetic energy distributed among relevant modes? Does an appropriate control change that distribution? The questions isolate one simulation artifact amid many possible model, algorithm, and measurement problems. An abnormal partition alone does not establish the flying ice cube effect; its link to rescaling matters.
Abstract Reasoning¶
When a thermostatted simulation gives an unexpected structural or transport result, examine mode-resolved kinetic energy as well as the total temperature. Compare the result under a thermostat with a different, justified sampling rule, while keeping the physical system and other relevant conditions explicit. If the modal allocation and the derived result change together, investigate whether the original inference depended on a rescaling artifact. This is a diagnostic inference, not proof that the replacement run gives an absolutely correct physical value.
Knowledge Transfer¶
The same question applies to a simple molecule and to a confined guest/host simulation. Braun and colleagues computed a translational/vibrational partition for a one-dimensional ethane model; in their benzene/MOF-5 comparison, they attributed a false diffusion contrast to excess guest translational energy under Berendsen without reporting a separate full MOF vibrational-energy breakdown.[^ref-8d8585786a8a] Transfer the diagnostic question, not ethane's numerical partition or a claim that all thermostat settings fail.
Example¶
One-dimensional ethane. In Braun and colleagues' simulated model, simple rescaling or Berendsen concentrates energy in translation and depletes vibration under tested conditions. CSVR and Langevin controls do not show that same partition.[^ref-8d8585786a8a] The simulated molecule is the carrier; repeated rescaling is the operation; the biased kinetic-energy distribution is the effect.
Benzene in MOF-5. The authors reproduce an apparent flexible-versus-rigid diffusion difference under Berendsen. Nosé–Hoover and CSVR comparisons do not retain it; they attribute the spurious difference to excess guest translation under Berendsen.[^ref-8d8585786a8a] This is a different simulated carrier and downstream observable. Its loading conditions prevent the comparison from certifying an absolute diffusion coefficient.[^ref-8d8585786a8a]
Relationships to Other Abstractions¶
Current abstraction Flying ice cube effect Domain-specific
Parents (2) — more general patterns this builds on
-
Flying ice cube effect is a kind of Failure Prime
The actual rescaling-caused violation is a criterion-relative failure of modal-energy allocation in the simulation run.
-
Flying ice cube effect presupposes Molecular Dynamics Domain-specific
The rescaling-induced artifact requires a trajectory-producing molecular-dynamics method, but is not itself a simulation method.
Hierarchy paths (2) — routes to 2 parentless roots
- Flying ice cube effect → Failure
- Flying ice cube effect → Molecular Dynamics → State and State Transition → Phase Space
Neighborhood in Abstraction Space¶
Flying ice cube effect sits in a sparse region of the domain-specific corpus (62nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Thermodynamics & Dissipative Systems (21 abstractions)
Nearest neighbors
- Maxwell's Demon — 0.85
- Langevin Dynamics — 0.85
- Brownian Dynamics — 0.85
- Lattice Boltzmann Methods — 0.85
- Entropy of Activation — 0.84
Computed from structural-signature embeddings · 2026-10-17
Not to Be Confused With¶
- Any equipartition failure: other MD numerical faults can cause one without this velocity-rescaling route.[^ref-8d8585786a8a]
- Any Berendsen run: the effect is conditional, not guaranteed merely by the thermostat label.[^ref-8d8585786a8a]
- Visible flight: the severe image helps name the effect, but an actual rescaling-caused modal bias need not move the whole simulated system across the box.[ref-ec81f39bf736][ref-8d8585786a8a]
- An isokinetic Berendsen ensemble: Braun and colleagues say Berendsen does not correspond to a known ensemble; simple rescaling's isokinetic aim should not be transferred to it.[^ref-8d8585786a8a]
References¶
[^ref-ec81f39bf736]: Stephen C. Harvey, Robert K.-Z. Tan, and Thomas E. Cheatham III, “The Flying Ice Cube: Velocity Rescaling in Molecular Dynamics Leads to Violation of Energy Equipartition,” Journal of Computational Chemistry 19 (1998), 726–740, publisher abstract, https://doi.org/10.1002/(SICI)1096-987X(199805)19:7%3C726::AID-JCC4%3E3.0.CO;2-S. Only the publisher abstract was directly inspected for the claims cited here. [^ref-8d8585786a8a]: Efrem Braun, S. Mohamad Moosavi, and Berend Smit, “Anomalous Effects of Velocity Rescaling Algorithms: The Flying Ice Cube Effect Revisited,” Journal of Chemical Theory and Computation 14 (2018), 5262–5272, https://doi.org/10.1021/acs.jctc.8b00446. Full author manuscript at https://infoscience.epfl.ch/server/api/core/bitstreams/1a10bf1e-96d7-4fc7-9bde-67b0d1f695ca/content, especially Introduction; §§3.1, 3.3–3.5; Figures 2, 3, 11; Appendix C. The MOF-5 phase-envelope and absolute-diffusion caveats are in §3.5.