Long-Time Energy Conservation of Numerical Methods for Oscillatory Differential Equations
read more
Citations
Discrete mechanics and variational integrators
Geometric numerical integration illustrated by the Störmer-Verlet method
Asynchronous Variational Integrators
Variational time integrators
A stochastic immersed boundary method for fluid-structure dynamics at microscopic length scales
References
Numerical integration of ordinary differential equations based on trigonometric polynomials
On the Hamiltonian Interpolation of Near-to-the-Identity Symplectic Mappings with Application to Symplectic Integration Algorithms
Long-Time-Step Methods for Oscillatory Differential Equations
A Gautschi-type method for oscillatory second-order differential equations
The life-span of backward error analysis for numerical integrators
Related Papers (5)
Geometric Numerical Integration: Structure-Preserving Algorithms for Ordinary Differential Equations
A Gautschi-type method for oscillatory second-order differential equations
Frequently Asked Questions (6)
Q2. What is the formula for the second component of (2.2)?
Using the nonresonanceassumption (5.10), the condition x̂(0) = x0 = (x01, x02) becomes x01 = y1(0) +O ( hφ(hω)z2(0) ) ,x02 = z2(0) + z2(0) +O ( h2ψ(hω)/s21 ) +O ( hψ(hω)φ(hω)z2(0) ) . (5.13)The formula for the first component of (2.2), x11−x01 = hẋ01+ 12h2g1(Φx0), together with x̂1(h)− x̂1(0) = hẏ1(0) +O(h2) +O(hφ(hω)z2(0)) implies thatẋ01 = ẏ1(0) +O(h) +O ( φ(hω)z2(0) ) .(5.14)For the second component the authors have x12−coshω x02 = h sinchω ẋ02+O(h2ψ(hω)) from (2.2), and x̂2(h) − coshω x̂2(0) = (1 − coshω)y2(0) + O(h2ψ(hω)) + i sinhω ( z2(0) −z2(0) ) +O ( hψ(hω)φ(hω)z2(0) ) , which after division by h sinchω yieldsẋ02 = iω ( z2(0)− z2(0) ) + O(hψ(hω)/sinchω)+
Q3. what is the defining relation for y and zk?
The initial value for z2 satisfies z2(0) = O(ω −1), and it follows from (2.10) that hψ(hω)/s2 = O(ω −1), so that ż2 = O(ω−1z2) by (5.12).
Q4. What is the corresponding function for the Fermi–Pasta–Ulam problem?
Z0](0) +O(h2) +O(ω−1) = O(nhN+1) +O(h2) +O(ω−1),which gives the desired bound for the deviation of the total energy along the numerical solution.
Q5. what is the recurrence relation for y and zk?
Inserting this ansatz and its derivatives into(5.8) and comparing like powers of √ h yields recurrence relations for the functions fkjl, g k jl.
Q6. what is the proof of the theorem 4.1?
(5.7) Inserting the ansatz (5.1), expanding the right-hand side of (5.7) into a Taylor series around Φy(t), and comparing the coefficients of eikωt yield for the functions y(t) and zk(t)L(hD)y = h2Ψ ( g(Φy) + ∑ s(α)=0 1 m! g(m)(Φy)(Φz)α ) ,L(hD + ikωh)zk = h2Ψ ∑s(α)=k1m! g(m)(Φy)(Φz)α.(5.8)Here, α = (α1, . . . , αm) is a multi-index as in the proof of Theorem 4.1, s(α) =∑m j=1 αj , and (Φz) α is an abbreviation for the m-tupel (Φzα1 , . . . ,Φzαm).