Fact-checked by Grok 2 weeks ago

Leapfrog integration

Leapfrog integration is a second-order for integrating ordinary differential in the form \ddot{x} = f(x), commonly applied to simulate the of particles under conservative forces in . It operates by alternately updating positions at integer time steps and velocities at time steps in a staggered "leapfrog" fashion, which inherently enforces time-reversibility and requires only one force evaluation per full step. As a , it preserves the geometric structure of , leading to bounded energy errors and superior long-term stability compared to non-symplectic methods of similar order, such as the classical Runge-Kutta integrator. The method, also known as the velocity Verlet or Störmer-Verlet algorithm, was introduced to simulations by Loup Verlet in 1967 for computer simulations of classical fluids using Lennard-Jones potentials, where it enabled efficient computation of thermodynamic properties for systems of up to 864 interacting particles. The core update rules derive from a expansion of the , eliminating explicit dependence in the basic position-Verlet form while the leapfrog variant explicitly tracks velocities for clarity in implementation. Its equivalence to the Verlet scheme ensures second-order local accuracy (O(\Delta t^2)), with global errors that grow linearly rather than exponentially due to symplecticity. This makes it particularly effective for separable systems, where the splits into kinetic and terms, allowing drift-kick-drift compositions. Leapfrog integration has become a cornerstone in and chemistry, powering simulations in software like and LAMMPS for studying biomolecular conformations and material properties. In , it facilitates N-body simulations of gravitational systems, such as galactic dynamics or planetary orbits, due to its low computational cost and conservation of angular momentum. Extensions, including higher-order variants like the Yoshida integrator, build on its foundation to achieve greater accuracy while retaining symplecticity, and it is also employed in for , where its reversibility aids efficient sampling of posterior distributions. Despite its strengths, the method can exhibit instabilities with non-conservative forces or stiff potentials, prompting hybrid approaches in advanced applications like .

Fundamentals

Overview and Motivation

Leapfrog integration is a second-order, explicit, time-reversible designed for solving second-order ordinary differential equations of the form \ddot{x} = A(x), where x represents and A(x) encapsulates forces dependent on , as commonly encountered in for Hamiltonian systems. This integrator is particularly valued in simulations requiring long-term stability, such as and N-body problems, due to its ability to maintain qualitative features of the exact without introducing artificial dissipation. The primary motivation for integration arises in conservative systems, where traditional explicit methods like the forward Euler scheme lead to rapid drift and numerical over extended simulation times. By employing a staggered update scheme—alternating half-steps for velocities and full steps for positions— avoids these issues, achieving near-conservation of and enabling reliable simulations over astronomically long periods relative to the time step. This structure ensures , meaning the integration can be run backward to recover the initial state exactly (up to rounding errors), which is essential for reversible physical processes. A basic outline of the step, for a time step h, proceeds as follows in :
v_{n+1/2} ← v_n + (h/2) * A(x_n)
x_{n+1} ← x_n + h * v_{n+1/2}
v_{n+1} ← v_{n+1/2} + (h/2) * A(x_{n+1})
This formulation originated from practical needs in particle simulations, such as computing trajectories of charged particles in and modeling interactions, where preserving phase-space volume is crucial to uphold and statistical equilibrium.

Historical Background

Although the method has earlier origins, dating back to 1791 when it was used by Delambre for orbital computations and rediscovered several times thereafter, the leapfrog integration method traces its modern recognition to the early , when and astrophysicist Carl Størmer developed a numerical scheme to compute the trajectories of charged particles in the . Størmer developed this approach around 1907, applying it to explain auroral phenomena and later to model paths, addressing the lack of analytical solutions for the governing second-order differential equations. His work, known as the Størmer method, laid the groundwork for what would later be recognized as a key technique in for Hamiltonian systems. The method experienced a significant revival in 1967 through the efforts of French physicist Loup Verlet, who adapted it for simulations of classical fluids interacting via Lennard-Jones potentials. Verlet integrated the for systems of up to 864 particles using early computers, demonstrating the algorithm's efficiency and accuracy in preserving long-term stability for . This application renamed the technique as , with the variant—characterized by staggered updates of positions and velocities—gaining prominence for its computational simplicity and reduced error accumulation in time-stepping. In the and , leapfrog integration saw widespread adoption in , particularly for N-body problems simulating gravitational interactions among stars and particles. Pioneering simulations by researchers like Sverre Aarseth utilized the method to handle direct summations in , enabling the study of cluster evolution and galactic structures on early digital computers. This period marked the transition from manual calculations to automated numerical experiments, solidifying leapfrog's role in astrophysical modeling. A pivotal advancement occurred with Ronald Ruth's 1983 development of canonical integration techniques, which explicitly connected to the broader framework of integrators, emphasizing preservation of phase-space volume in systems. Building on this, Étienne Forest and Ronald Ruth further advanced the field in 1990 by constructing explicit fourth-order methods, linking the basic scheme to geometric integration theory and enhancing its applicability to long-term simulations. In 1990, Haruo Yoshida introduced a systematic approach to generating higher-order extensions of these integrators, optimizing coefficients for separable s and influencing subsequent developments in numerical methods for dynamical systems.

Core Algorithm

Basic Formulation

The leapfrog integrator, also known as the Störmer-Verlet or velocity Verlet method in certain contexts, is a second-order explicit numerical scheme for solving the second-order \ddot{x} = A(x), where x represents and A(x) is the acceleration derived from forces. It employs a staggered temporal grid, with positions evaluated at integer time steps t_n = n \Delta t and velocities at half-integer steps t_{n+1/2} = (n + 1/2) \Delta t, which facilitates fully explicit updates without requiring the solution of implicit equations. The core update rules alternate between advancing the using the current half-step and then advancing the over a full time step \Delta t using the at the new . After initialization, the standard cycle from time t_n to t_{n+1} given x_n and half-step v_{n+1/2} (note the indexing after init) proceeds by first updating and then . Specifically: x_{n+1} = x_n + \Delta t \cdot v_{n+1/2} v_{n+3/2} = v_{n+1/2} + \Delta t \cdot A(x_{n+1}) These steps complete one full cycle from time t_n to t_{n+1}. To initialize the integration from initial conditions x_0 and v_0 at t_0 = 0, first compute the initial a_0 = A(x_0), then the first half-step velocity v_{1/2} = v_0 + \frac{\Delta t}{2} a_0, after which the standard update cycle proceeds. A pseudocode outline for multi-step integration over N steps is as follows:
Set x = x_0
a = A(x)
v_half = v_0 + (Δt / 2) * a
For n = 0 to N-1:
    x = x + Δt * v_half         // Update position to next integer step
    a = A(x)                    // Compute acceleration at new position
    v_half = v_half + Δt * a    // Update velocity to next half-step
    // Optionally, compute full-step velocity v_n = (v_{n-1/2} + v_{n+1/2}) / 2 for output
End
This structure ensures one force evaluation per step, making it computationally efficient for systems like particle simulations. The method exhibits a local truncation error of O(\Delta t^3) per step and a global error of O(\Delta t^2) over fixed integration intervals, consistent with its second-order accuracy.

Derivation from Hamilton's Equations

The leapfrog integrator originates from the numerical solution of Hamilton's equations for a separable Hamiltonian H(q, p) = T(p) + V(q), where T depends only on the momenta p and V only on the coordinates q. These equations are given by \dot{q} = \frac{\partial H}{\partial p} = \nabla_p T(p), \quad \dot{p} = -\frac{\partial H}{\partial q} = -\nabla_q V(q). The separability allows the dynamics to be decomposed into solvable subflows corresponding to the kinetic and potential energies. Operator splitting exploits this structure by approximating the exact operator \exp(\Delta t \mathcal{L}), where \mathcal{L} is the associated with the , as a of exact flows for the split parts. Define the kinetic flow \phi_T^{\Delta t}, which solves \dot{q} = \nabla_p T(p), \dot{p} = 0 (so momenta are constant and positions update linearly), and the \phi_V^{\Delta t}, which solves \dot{q} = 0, \dot{p} = -\nabla_q V(q) (positions constant, momenta updated by forces). For T(p) = \frac{1}{2} p^T M^{-1} p (as in standard systems), the kinetic flow is simply q \leftarrow q + \Delta t M^{-1} p. The leapfrog method arises as a symmetric of these flows, specifically the Strang splitting \phi_{\text{leap}}^{\Delta t} = \phi_T^{\Delta t/2} \circ \phi_V^{\Delta t} \circ \phi_T^{\Delta t/2}, which approximates the full evolution up to local error O((\Delta t)^3) due to the non-commutativity of the operators (i.e., [ \mathcal{L}_T, \mathcal{L}_V ] \neq 0). This second-order accuracy stems from the symmetric arrangement, which cancels leading-order error terms in the Baker-Campbell-Hausdorff formula. The symmetric structure imparts time-reversibility to the integrator: applying the method with time step -\Delta t yields the inverse transformation, \phi_{\text{leap}}^{-\Delta t} = (\phi_{\text{leap}}^{\Delta t})^{-1}. This property follows directly from the palindromic composition of flows and ensures qualitative in long-term simulations. The formulation is mathematically equivalent to the velocity Verlet algorithm, which interleaves half-steps differently but produces identical updates when velocities are identified with momenta (up to scaling by mass). In velocity Verlet, positions and velocities are advanced in a staggered manner, mirroring the kick-drift-kick sequence of the kick-drift-kick variant of .

Key Properties

Symplectic and Geometric Preservation

Leapfrog integration, also known as the Störmer-Verlet method, is a for systems, meaning it generates a discrete map that preserves the symplectic structure of . In the context of , a symplectic map preserves the symplectic form \omega = \mathrm{d}q \wedge \mathrm{d}p, where q and p are the generalized and coordinates, respectively. This preservation ensures that the numerical flow mimics the geometric properties of the exact flow, maintaining the area in and preventing artificial dissipation or growth of trajectories over long times. The symplecticity of the method arises from its formulation as a of exact flows for separable Hamiltonians of the form H(q,p) = T(p) + V(q). Specifically, the update consists of three substeps: an exact momentum-independent translation in position (drift step), an exact velocity-dependent update via (kick step), and another drift step. Each of these substeps corresponds to the exact flow of a solvable subsystem, which is itself , and the of symplectic maps yields a symplectic map overall. This property holds for the full step, as verified through the preservation of the wedge product form in the discrete transformation. In the broader framework of geometric integration, maintains key invariants of , such as bounded oscillations rather than secular drift, and approximates by preserving phase-space volume up to higher-order terms. For integrable systems, the energy error remains bounded over exponentially long times, in contrast to non-symplectic methods like explicit Runge-Kutta integrators, which exhibit linear drift in due to the accumulation of non-symplectic errors. This bounded error behavior stems from the method's ability to shadow the exact closely in a geometric sense. A deeper justification for this long-term stability lies in the concept of the shadow Hamiltonian: backward error analysis shows that leapfrog exactly conserves a perturbed Hamiltonian \tilde{H}, which is close to the original H (with perturbation of order h^2, where h is the time step). This shadow Hamiltonian governs the numerical solution exactly, ensuring that the discrete trajectory lies on its level sets, thereby explaining the qualitative fidelity of the integration even when local errors are present.

Stability and Error Analysis

The local truncation error of the leapfrog integrator is derived by expanding the exact solution of the using around the current time step and comparing it to the numerical update. For a separable H(q, p) = T(p) + V(q), the and updates satisfy the exact solution up to terms of \Delta t^2, with the leading term arising from the third-order derivatives, yielding a local of O(\Delta t^3). This second-order accuracy in the local sense stems from the symmetric structure of the method, which cancels lower-order error contributions. Over multiple steps, these local errors accumulate, resulting in a global error of O(\Delta t^2) when integrating over a fixed time interval t = N \Delta t as N \to \infty and \Delta t \to 0. For nearly integrable systems with Diophantine frequencies, backward error analysis further bounds the long-term global error by O(t \Delta t^2), ensuring linear growth rather than exponential divergence. Regarding stability, the leapfrog method demonstrates unconditional stability for linear forces, such as in the harmonic oscillator, where the numerical solution remains bounded for all times provided the timestep satisfies \Delta t < 2 / \omega (where \omega is the angular frequency). In contrast, for nonlinear forces, stability becomes conditional, requiring timestep restrictions to prevent numerical blow-up. A distinctive feature in periodic or quasi-periodic systems is the presence of long-term resonance errors, where small phase mismatches accumulate but are mitigated by the symplectic nature of the integrator, confining errors to bounded oscillations rather than unbounded drift. This contrasts with non-symplectic methods, which exhibit secular growth. The method's symplecticity also limits energy fluctuations to an amplitude of O(\Delta t^2), with near-conservation over exponentially long times t \leq e^{c / \Delta t} for analytic Hamiltonians, avoiding the linear or worse drift seen in non-symplectic integrators like explicit Euler.

Advanced Variants

Higher-Order Composition Methods

Higher-order leapfrog integrators are constructed by composing multiple instances of the basic second-order leapfrog step, denoted as \phi_h, to achieve improved accuracy while preserving symplecticity. A seminal example is the Forest-Ruth method, introduced in , which employs a three-stage composition \phi_h = \phi_{\theta h/2} \circ \phi_{(1-2\theta)h} \circ \phi_{\theta h/2}, where \theta = 1/(2 - 2^{1/3}) \approx 0.6756. This arrangement yields a fourth-order integrator, with the leading error term of order five, requiring three evaluations of the basic step per full time step h. In general composition theory for symplectic integrators, even-order methods are derived from symmetric products of basic steps, ensuring time-reversibility and even maximal order, whereas odd-order methods arise from asymmetric compositions that break this to cancel odd-powered error terms. To reduce the local to fourth order, the coefficients in the must satisfy conditions that eliminate the third-order error term, typically involving the solution of nonlinear equations derived from the Baker-Campbell-Hausdorff formula or B-series expansions. A key conceptual advancement enables the systematic generation of coefficients for arbitrary even orders using generating functions, which parameterize recursive s such as the triple-jump to build higher-order integrators from lower-order ones without resolving increasingly nonlinear systems at each level. This approach facilitates the of families of methods with controlled error growth over long integrations. The computational cost of these higher-order methods scales with the number of basic evaluations per step; for instance, achieving fourth order typically demands at least three such evaluations, compared to one for the standard second-order , thereby increasing overhead but offering superior long-term for stiff systems.

Yoshida Integrators

Haruo introduced a systematic in for constructing higher-order symplectic integrators through compositions of second-order steps, ensuring explicitness, , and preservation of symplectic structure for separable Hamiltonian systems of the form H = T(\mathbf{p}) + V(\mathbf{q}). This approach leverages generating functions and the Baker-Campbell-Hausdorff formula to derive coefficients that achieve desired orders while minimizing the number of force evaluations. Particularly suited for long-term simulations where phase errors accumulate, Yoshida's integrators were developed with astronomical N-body problems in mind, offering reduced energy drift and improved accuracy over lower-order methods. A key example is the fourth-order integrator, formed as a triple product of leapfrog operators: \mathcal{W}(w_0 h) \circ \mathcal{W}(w_1 h) \circ \mathcal{W}(w_2 h), where \mathcal{W}(\xi h) denotes a leapfrog step of size \xi h, but symmetrized to yield an explicit scheme with five substeps and three force evaluations. The coefficients are w_0 = w_4 = \frac{0.5}{2 - 2^{1/3}}, w_1 = w_3 = \frac{1 - 2^{1/3}}{2 - 2^{1/3}}, and w_2 = \frac{-2^{1/3}}{2 - 2^{1/3}}, ensuring the local truncation error is \mathcal{O}(h^5). Although w_1 and w_2 are negative, indicating backward time steps in intermediate substeps, the overall scheme remains stable for sufficiently small h and avoids the oscillatory instabilities common in non-symplectic methods, as the negative steps are compensated by the symmetric structure. For sixth-order accuracy, extends the composition to a seven-step symmetric product, solving a of algebraic equations derived from the triple-product to eliminate lower-order error terms up to \mathcal{O}(h^7). This requires 14 force evaluations but provides significantly lower phase errors in periodic orbits, crucial for N-body simulations of . The coefficients, obtained numerically by solving the nonlinear equations, include values such as w_0 \approx 0.5129189, w_1 \approx -0.05072096, w_2 \approx 0.3917522, w_3 \approx -1.438206, with w_4 = w_0, w_5 = w_1, w_6 = w_2, though exact forms are of a . Yoshida's method enables recursive construction of for arbitrary even orders greater than two, by composing a fourth-order integrator with additional steps and solving for coefficients that raise the order while keeping the scheme explicit. This recursive nature allows efficient implementation in codes for astronomical simulations, where higher orders balance computational cost against long-term accuracy without introducing artificial dissipation.

Specialized Extensions

Leapfrog integration has been adapted for relativistic charged-particle dynamics through Boris-like pushers that incorporate the Lorentz force while preserving key invariants such as the magnetic moment. In the work of He et al. (2016), an explicit high-order symplectic integrator is developed for non-relativistic and relativistic charged particles in electromagnetic fields, extending the standard leapfrog scheme to maintain volume preservation and symplecticity, which ensures long-term stability in simulations of particle accelerators and plasma physics. This adaptation approximates the exact solution of the Lorentz equations by composing velocity and position updates in a staggered manner, similar to the classical Boris pusher, but with higher-order accuracy to handle relativistic effects without energy drift. For multi-scale systems where particles experience widely varying forces, asynchronous methods allow individual particles or groups to use different timesteps, improving efficiency over global timestepping. Multiple time stepping approaches in , such as those using hierarchical grids, separate fast and slow interactions, enabling smaller timesteps only for stiff components like bonded interactions while using larger steps for non-bonded forces, thus reducing computational cost by up to an in heterogeneous systems. This approach preserves the second-order accuracy and time-reversibility of the standard while avoiding resonance instabilities common in synchronous multiple timestep schemes. In oscillatory systems prone to high-frequency instabilities, filtered leapfrog integrators dampen spurious modes to enhance stability without sacrificing the method's symplectic properties. Filtered versions of the leapfrog scheme for second-order differential equations incorporate a on velocity updates to suppress high-frequency oscillations, which is particularly useful in simulations of molecular vibrations or wave equations where unfiltered leapfrog exhibits nonlinear . The filtering parameter is chosen to minimally affect low-frequency , allowing stable with timesteps larger than those permitted by the unfiltered method while maintaining global error bounds consistent with second-order . A specialized extension for constant-temperature (MD) simulations is the middle- , which applies and random forces at the half-step to improve and accuracy in Nosé-Hoover ed systems. This scheme, detailed in Sun et al. (2021), inserts the thermostat coupling midway through the step, achieving higher-order consistency between the and compared to end-placed variants, resulting in reduced fluctuations and better sampling efficiency in tests of Lennard-Jones fluids and biomolecular systems. Numerical s show that the middle- version maintains stable dynamics even at larger timesteps, with errors in thermodynamic properties below 0.1% over long trajectories. Hybrid variants combine explicit steps with implicit treatments for stiff components, enabling robust integration of systems with disparate timescales. In the multirate leapfrog-type methods proposed by Carle and Hochbruck (2022), stiff oscillatory terms are handled implicitly via a modified rule within the leapfrog framework, while non-stiff parts remain explicit, allowing timestep independence for the stiff sector and demonstrating superior stability for equations like the Klein-Gordon model with error growth bounded by O(h^2) independent of stiffness parameter. This hybrid approach is particularly effective for stiff problems in and , where pure explicit leapfrog would require prohibitively small timesteps.

Applications and Comparisons

Use in Physical Simulations

Leapfrog integration, also known as the velocity Verlet algorithm, serves as the standard method for advancing (MD) simulations of biomolecular systems, such as protein trajectories in solvent environments. Its adoption in widely used software packages like stems from its excellent long-term , which minimizes artificial drift in total energy over extended simulation times, enabling reliable predictions of conformational dynamics. This property is crucial for studying processes like binding or enzymatic , where maintaining physical realism is essential. The leapfrog integrator traces its origins to the seminal work of Loup Verlet in 1967, who introduced a position-based variant for simulating classical fluids, laying the foundation for atomic-level MD simulations that were later extended to biomolecular contexts starting in the late 1970s. The first molecular dynamics simulation of a protein was reported in 1977 by McCammon, Gelin, and Karplus, simulating the bovine pancreatic trypsin inhibitor. For instance, a 1998 simulation of the villin headpiece subdomain used the Verlet-leapfrog algorithm to observe early stages of folding, including the formation of secondary structures such as helices, demonstrating its utility in exploring protein folding pathways and intermediate states. In astrophysical N-body gravitational simulations, leapfrog integration is employed in codes to model the evolution of star clusters and planetary systems over cosmological timescales spanning billions of years. Its nature ensures bounded errors, preserving orbital stability and preventing artificial ejections or collapses in hierarchical systems, which is vital for accurately reproducing observed galactic dynamics. Practical implementations of in MD often incorporate (PBC) to mimic bulk environments without surface artifacts, where particle positions are wrapped around the simulation box to enforce minimum image conventions. Efficiency is further enhanced by neighbor lists, which precompute pairs of interacting atoms within a cutoff , reducing the computational cost from O(N²) to nearly linear scaling for large systems like biomolecular complexes. A illustrative case study is the simulation of a , where the method maintains bounded total energy with minimal oscillation around the exact value, even over thousands of periods, due to its time-reversibility. In contrast, the explicit exhibits rapid energy divergence, with the oscillatory amplitude growing unbounded, highlighting leapfrog's superior for conservative systems. This advantage extends to physical simulations, where such prevents unphysical artifacts over long runs.

Comparison with Other Integrators

The leapfrog integrator demonstrates superior long-term compared to the forward , where tends to increase without bound due to its non- nature and accuracy, leading to over extended simulations. In contrast, leapfrog, as a second-order symplectic method, causes to oscillate boundedly around the true value, preserving qualitative behavior in systems without artificial growth or decay. Similarly, against the Crank-Nicolson scheme, leapfrog avoids numerical damping in oscillatory problems while maintaining balance, though Crank-Nicolson offers unconditional for linear wave equations at the cost of potential instabilities in nonlinear cases without specific modifications. When benchmarked against the fourth-order Runge-Kutta (RK4) method, achieves comparable per-step accuracy for second-order error terms but excels in systems by avoiding artificial dissipation, as evidenced in periodic orbital simulations where RK4 gradually circularizes eccentric orbits due to secular energy drift. RK4 provides higher short-term precision with four force evaluations per step, yet its non-symplectic structure leads to poorer long-term in phase-space preserving applications, making preferable for simulations requiring over many periods. Both and the are integrators suitable for dynamics, preserving phase-space volume and linear invariants, but 's explicit formulation requires only one force evaluation per step, rendering it computationally cheaper for non-stiff problems compared to the 's need for iterative solvers. The offers broader for general systems but incurs higher overhead from solving nonlinear equations, whereas suffices for separable potentials without such costs. A key distinction in periodic orbits is phase error accumulation: exhibits linear growth in phase lag over time due to its symplectic preservation of bounded errors, while non- methods like Runge-Kutta show exponential error growth, amplifying deviations in long integrations. 's explicit nature introduces trade-offs for stiff equations, lacking A-stability and thus requiring smaller time steps than implicit methods like backward differentiation formulas (BDF) to avoid instability from high-frequency modes, though it remains efficient for moderately non-stiff problems.

References

  1. [1]
    Computer "Experiments" on Classical Fluids. I. Thermodynamical ...
    Computer "Experiments" on Classical Fluids. I. Thermodynamical Properties of Lennard-Jones Molecules. Loup Verlet*.
  2. [2]
    Leapfrog integrator - Applied Mathematics Consulting
    Jul 13, 2020 · The leapfrog integrator is a numerical method for solving second-order differential equations, also known as Störmer-Verlet method, and is ...
  3. [3]
    Qualitative study of the symplectic Störmer–Verlet integrator
    Jun 8, 1995 · This paper analyzes the Störmer–Verlet method as applied to the simple harmonic model, whose generalization is an important model for molecular ...
  4. [4]
    [PDF] Lecture 04: Numerical Integration Methods (Continued)
    Apr 9, 2023 · Drift- and Kick-Operators. Separable Hamiltonian. The Leapfrog. The drift and kick operators are symplectic transformations of phase-space !
  5. [5]
    [PDF] arXiv:astro-ph/9710043v1 3 Oct 1997
    ABSTRACT. Leapfrog integration has been the method of choice in N-body simulations owing to its low computational cost for a symplectic integrator with ...
  6. [6]
  7. [7]
    Geometric Numerical Integration: Störmer–Verlet Method
    Jul 29, 2003 · This article illustrates concepts and results of geometric numerical integration on the important example of the Störmer–Verlet method.
  8. [8]
    Fourth-order symplectic integration - ScienceDirect.com
    This method preserves the property that the time evolution of such a system yields a canonical transformation from the initial conditions to the final state.Missing: technique | Show results with:technique
  9. [9]
    Construction of higher order symplectic integrators - ScienceDirect
    The simplest one is the 4th order integrator which agrees with one found by Forest and by Neri. For 6th and 8th orders, symplectic integrators with fewer steps ...
  10. [10]
    Accuracy and Optimal Time Steps of Stoermer-Leapfrog Integrators
    Jul 9, 1997 · View a PDF of the paper titled Common Molecular Dynamics Algorithms Revisited: Accuracy and Optimal Time Steps of Stoermer-Leapfrog ...
  11. [11]
    [PDF] A Canonical Integration Technique - JACoW
    The purpose of this note is to develop an explicit third order symplectic map (i.e. a third order integra- tion step that preserves exactly the canonical char-.
  12. [12]
    [PDF] arXiv:1411.3367v1 [math.NA] 3 Nov 2014
    Nov 3, 2014 · Abstract We present a method for explicit leapfrog integration of inseparable Hamiltonian systems by means of an extended phase space.Missing: Hamilton's | Show results with:Hamilton's
  13. [13]
    On the Nonlinear Stability of Symplectic Integrators
    McLachlan, R.I., Perlmutter, M. & Quispel, G.R.W. On the Nonlinear Stability of Symplectic Integrators. BIT Numerical Mathematics 44, 99–117 (2004). https ...
  14. [14]
    [PDF] Lecture 2: Symplectic integrators
    The statement follows from the fact that the Störmer–Verlet scheme is the composition of the two symplectic Euler methods (1) with step size h/2. Even order 2 ...Missing: leapfrog paper
  15. [15]
    [PDF] Multiple Grid Methods for Classical Molecular Dynamics
    For integration of molecular dynamics the use of different time steps for different interactions allows longer time steps for many of the interactions, and this ...
  16. [16]
    Error Analysis of Multirate Leapfrog-Type Methods for Second-Order ...
    In this paper we consider the numerical solution of second-order semilinear differential equations for which the stiffness is induced by only a few ...Missing: seminal | Show results with:seminal
  17. [17]
    Molecular Dynamics - GROMACS 2025.3 documentation
    This searching, usually called neighbor search (NS) or pair search, involves periodic boundary conditions and determining the image (see sec. ... 7 The Leap-Frog ...
  18. [18]
    [PDF] Energy Conservation as a Measure of Simulation Accuracy - bioRxiv
    Oct 24, 2016 · Abstract: Energy conservation is widely used as a measure of accuracy for molecular simulations. When reporting rates of energy drift, ...<|separator|>
  19. [19]
    The early stage of folding of villin headpiece subdomain observed in ...
    Molecular dynamics (MD) simulations with full atomic representation of both protein and solvent possess a unique advantage to study protein folding because of ...
  20. [20]
    [astro-ph/9710043] Time stepping N-body simulations - arXiv
    Oct 3, 1997 · Leapfrog integration has been the method of choice in N-body simulations owing to its low computational cost for a symplectic integrator with ...
  21. [21]
    [PDF] PHY411 Lecture notes Part 7 –Integrators
    When the half step at either end is combined, the method can be called the leapfrog method and only requires evaluating the forces once per step. It turns out ...<|control11|><|separator|>
  22. [22]
    [PDF] Physics 115/242 Leapfrog method and other “symplectic” algorithms ...
    In fact one can show that the results from an approximate symplectic integrator are equal to the exact dynamics of a “close by” Hamiltonian, H0(h) where ...
  23. [23]
    [PDF] On the Instability of Leap-Frog and Crank-Nicolson Approximations ...
    It turns out that the properties of the linearized equations are not at all sufficient for determining stability. The first example of pure nonlinear ...
  24. [24]
    [PDF] Lecture 03: Numerical Integration Methods - Duarte Lab @ UCSD
    Apr 7, 2023 · • Implicit midpoint method: • 2nd-order accurate. • Time symmetric and symplectic. • But still implicit. • Explicit midpoint method. S n+1. = S.
  25. [25]
    [PDF] Geometrical numerical integration methods for differential equations
    If F = H, then scheme preserves energy. • THEOREM: • Leapfrog method: preserves linear invariants and quadratic invariants of the form.
  26. [26]
    Cheap implicit symplectic integrators - ScienceDirect.com
    Hence for moderately accurate integration of such problems by, say, the leapfrog method the time step tends to be limited by stability restrictions rather than ...
  27. [27]
    Error Growth in the Numerical Integration of Periodic Orbits, with ...
    The authors develop and test variable step symplectic Runge–Kutta–Nyström algorithms for the integration of Hamiltonian systems of ordinary differential ...