Shock Dynamics in Layered Periodic Media

Authors

David I. Ketcheson

Randall J. LeVeque

Shock Dynamics in Layered Periodic Media

Communications in Mathematical Sciences, 10(3):859--874 (2012)

Abstract

Solutions of constant-coefficient nonlinear hyperbolic PDEs generically develop shocks, even if the initial data is smooth. Solutions of hyperbolic PDEs with variable coefficients can behave very differently. We investigate formation and stability of shock waves in a one-dimensional periodic layered medium by computational study of time-reversibility and entropy evolution. We find that periodic layered media tend to inhibit shock formation. For small initial conditions and large impedance variation, no shock formation is detected even after times much greater than the time of shock formation in a homogeneous medium. Furthermore, weak shocks are observed to be dynamically unstable in the sense that they do not lead to significant long-term entropy decay. We propose a characteristic condition for admissibility of shocks in heterogeneous media that generalizes the classical Lax entropy condition and accurately predicts the formation or absence of shocks in these media.

Code and data

Reproducibility repository: https://github.com/ketch/layeredmediashocks

Animations

Below are some animations of some of the ideas discussed in the paper. These are solutions of the one-dimensional nonlinear elasticity equations \[ \epsilon_t - u_x = 0, \qquad \rho(x)\, u_t - \sigma(\epsilon, x)_x = 0, \] where \(\epsilon\) is strain, \(u\) is velocity, \(\rho\) is density, and the stress is \(\sigma(\epsilon, x) = \exp(K(x)\epsilon) - 1\) with bulk modulus \(K\). The layered media alternate between material A, with \(\rho_A = K_A = 1\), and material B, with \(\rho_B = K_B\). The impedance of a material is \(Z = \sqrt{\rho K}\), so the impedance ratio is \(Z_B/Z_A = K_B\).

In each animation the upper curve is stress. The lower curve is either velocity or the total entropy \(\int \eta\, dx\) plotted against time, where \(\eta = \tfrac{1}{2}\rho u^2 + \int_0^\epsilon \sigma(s, x)\, ds\).

Time reversal in a homogeneous medium

In the time-reversal test, we solve forward to some time, negate the velocity, and solve forward again for the same length of time. If no shocks have formed, this returns the initial condition. In these two movies the medium is homogeneous, the initial stress is a Gaussian hump, and the initial velocity is zero. After the reversal, the black line is the reversed solution and the blue dots are the forward solution at the matching time, for comparison. If the solution is reversible, the two stresses agree and the two velocities have opposite signs.

In the first movie the velocity is reversed before shocks form, and the initial condition is recovered:

In the second movie the velocity is reversed after shocks have formed. The final solution differs from the initial condition:

Time reversal in a layered medium

Here the initial pulse is the one used in the stegoton problem of LeVeque and Yong (2003), and the boundaries are periodic, so waves that leave on the right re-enter on the left. As above, after the reversal the black line is the reversed solution and the blue dots are the forward solution.

With impedance ratio \(Z_B/Z_A = 4\), the pulse breaks up into a train of solitary waves (stegotons), and the reversed solution returns to the initial pulse:

With impedance ratio \(Z_B/Z_A = 2\), the waves have steep fronts with oscillations behind them, and the reversed solution does not return to the initial pulse:

Entropy

The total entropy is conserved by smooth solutions and decreases when shocks form. In these movies the lower curve shows the total entropy versus time.

Homogeneous medium, with the same Gaussian initial stress as above. The entropy stays constant until shocks form and then decreases:

Layered medium with \(Z_B/Z_A = 2\) and the stegoton-problem initial pulse: