The heterogeneous multiscale method
read more
Citations
Simulating Hamiltonian dynamics.
A review of predictive nonlinear theories for multiscale modeling of heterogeneous materials
Numerical methods for kinetic equations
A posteriori error estimates for the virtual element method
Continuum theory for dense gas-solid flow: A state-of-the-art review
References
Computer Simulation of Liquids
Theory of elasticity
Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations
Unified Approach for Molecular Dynamics and Density-Functional Theory
Embedded-atom method: Derivation and application to impurities, surfaces, and other defects in metals
Related Papers (5)
Frequently Asked Questions (14)
Q2. What constraints have to be implemented in order to guarantee that the algorithms give rise to the correct numerical?
Important constraints, such as conservation properties or upwinding, have to be implemented in order to guarantee that the algorithms give rise to the correct numerical solutions.
Q3. What is the way to solve the macroscopic model?
Since the macroscopic model (5.1) is in the form of conservation laws, it is natural to choose as the macroscale solver a finite volume method.
Q4. How can the authors control the bias introduced by modifying the cell problem?
The bias introduced by modifying the cell problem can be controlled by tuning the constant associated with the zeroth-order term.
Q5. How can the authors reduce the cost of the FE-HMM to log-linear complexity?
the authors note that by using spectral methods or high-order finite element methods for the micro solvers, it is possible to reduce the total cost of the FE-HMM to log-linear complexity in the macro degrees of freedom.
Q6. What boundary conditions are used to model the convective heat transfer with the surrounding air?
Convective heat transfer with the surrounding air is modelled by the Robin and Neumann boundaryconditions:n · (aε∇uε) + αuε = gR on ∂ΩR, (4.22) n · (aε∇uε) = gN on ∂ΩN , (4.23)where Ω is the domain of the object considered, and ∂ΩR and ∂ΩN are the surfaces of the three-dimensional object with Robin and Neumann boundary conditions, respectively.
Q7. What is the key observation that this re-interpretation permits?
The key observation that this re-interpretation permits is this: as long asΔt Mδt ⇔ ′ = α 1, (3.15)the limit theorem the authors used to justify the HMM scheme guarantees that the solution to (3.13) remains close to that of the original system (3.1), in the sense that x ′ ≈ x pathwise, and the stationary distribution of y ′ conditional on the current value of x ′ approximates that of y conditional on the current value of x .
Q8. What is the simplest example of the microscopic model that one can consider?
An example of the microscopic model that one can consider is that of molecular dynamics, that is, Newton’s equations of motion for the constituting atoms:ẋi = vi, miv̇i = ∑ j =i f ( xi(t)− xj(t) ) .
Q9. What can be done to speed up the computation time?
If the micro structure is regular enough, one can take advantage of fast micro solvers (based on spectral methods, for example) to speed up the computation time considerably.
Q10. What is the simplest way to estimate the effective force of a system of interest?
Motivated by the analytic averaging techniques (see, e.g., E (2011)), the authors hypothesize that the effective force of a system of interest can be defined byf̄(t) = lim δ−→0 [ lim −→0 1 δ ˆ t+δ t f (τ) dτ ]
Q11. What is the criterion for obtaining the optimal convergence rate?
Comparing the rate of convergence of the macro solver (Theorem 4.5) with the rate of convergence of the micro solver (see Theorem 4.7) gives a criterion to obtain the optimal (macro) convergence rate with minimal computational cost.
Q12. What is the common reason for the rate of reaction to be different?
Since the rate of reaction often depends exponentially on physical parameters such as temperature, it is quite common for these rates to be very different.
Q13. What is the proof of the uniqueness of a numerical solution of a problem?
The proof of the uniqueness of a numerical solution of problem (4.12) is quite involved, but can be derived without any structure assumption on the oscillation provided thats ∈ R → aε(·, s) ∈ (W 1,∞(Ω))d×d is of class C2 and |∂kuaε(·, s)|W 1,∞(Ω) ≤ Cε−1, k ≤ 2, (4.50)where C is independent of s and ε (see Abdulle and Vilmart (2011a, Theorem 3.3)).
Q14. What is the key to getting concrete error estimates?
(2.38) The key to getting concrete error estimates, and thereby giving guide-’in’ > ’to’lines to designing multiscale methods, lies in the estimation of e(HMM).