In this section we apply a double-scale expansion method to our problem. Let us consider an initial-value problem (IVP) for a quasilinear hyperbolic system:
$$\begin{aligned} \begin{array}{l} \displaystyle \frac{\partial \varvec{w}^\epsilon }{\partial t}+\varvec{A}({\varvec{w}}^\epsilon )\,\frac{\partial \varvec{w}^\epsilon }{\partial x}=\mathbf {0}, \qquad \varvec{w}^{\epsilon }(0,x)= \epsilon \,\varvec{w}_{1}(0,x), \end{array} \end{aligned}$$
(23)
where
\(\epsilon \) is a small parameter,
\( {\varvec{w}}^\epsilon = [{ \varvec{v}}^\epsilon , { \varvec{m}}^\epsilon ]^\mathrm{{T}}\), and matrix
\(\varvec{A}\) is as in (
18). We assume that the hyperbolic system (
23) is not strictly hyperbolic at the constant state
\({\mathbf {0}}\), and that
\(\lambda _1({\mathbf {0}}) = \lambda _2({\mathbf {0}}) \equiv \lambda \) is a double eigenvalue of matrix
\({{\mathbf {A}}}(0)\) with corresponding left
\({{\mathbf {l}}}_k \) and right
\({{\mathbf {r}}}_k\) eigenvectors, respectively,
\(k = 1,2\), and
\(\varvec{l_i} \cdot \varvec{r_j} = \delta _{ij}\), where
\(\delta _{ij}\) is Kronecker’s delta.
We assume a Taylor expansion around the zero constant state:
$$\begin{aligned} \varvec{A}({\varvec{w}}^\epsilon ) = \varvec{A}(\varvec{0} + \epsilon \tilde{\varvec{w}}) = \varvec{A}(\varvec{0}) + \epsilon \mathcal {B} \; \tilde{\varvec{w}} + \displaystyle \frac{1}{2}\epsilon ^2 \mathcal {C} \tilde{\varvec{w}}\tilde{\varvec{w}} + \mathcal {O} (\epsilon ^3), \end{aligned}$$
(24)
where
$$\begin{aligned}&\mathcal {B} \;\tilde{\varvec{w}} \equiv \nabla _{\varvec{u}}\left( \varvec{A}(\varvec{w}) \tilde{\varvec{w}}\right) \!\!\big |_{{\varvec{w}=\varvec{0}} \big .},\end{aligned}$$
(25)
$$\begin{aligned}&\mathcal {C} \tilde{\varvec{w}}\tilde{\varvec{w}} \equiv \nabla _{\varvec{u}} ( \nabla _{\varvec{u} }\left( \varvec{A}(\varvec{w}) \tilde{\varvec{w}} ) \tilde{\varvec{w}}\right) \!\!\big |_{{\varvec{w}= \varvec{0}} \big .}. \end{aligned}$$
(26)
We seek a solution to IVP (
23) around a
\({\varvec{0}} \) constant state in the form
$$\begin{aligned} \varvec{w}^{\epsilon }(t,x) = \epsilon \varvec{w}_1(\tau ,\theta ) + \epsilon ^2 \varvec{w}_2(\tau ,\theta ) + \mathcal {O}(\epsilon ^3). \end{aligned}$$
(27)
We introduce two new variables: a slow time
\( \tau = \epsilon t\) and a characteristic
\( \theta = x - \lambda t\). We treat them as new independent variables and apply the method of double-scale asymptotics. Let us first calculate derivatives with respect to the old variables
\(t\) and
\(x\) and express them in terms of the derivatives with respect to the new variables
\(\tau \) and
\( \theta \). We have
$$\begin{aligned} \displaystyle \frac{\partial \varvec{w}^\epsilon }{\partial t}&= \epsilon \frac{\partial {\varvec{w}_1}}{{\partial {\tau }}}\frac{\partial {\tau }}{\partial {t}} + \epsilon \frac{\partial {\varvec{w}_1}}{{\partial \theta }} \frac{\partial {\theta }}{\partial {t}} + \epsilon ^2 \frac{\partial {\varvec{w}_2}}{{\partial \tau }} \frac{\partial {\tau }}{\partial {t}} + \epsilon ^2 \frac{\partial {\varvec{w}_2}}{{\partial \theta }} \frac{\partial {\theta }}{\partial {t}} + \mathcal {O}(\epsilon ^3) \nonumber \\&= \epsilon (- \lambda ) \frac{\partial {\varvec{w}_1}}{\partial {\theta }} + \epsilon ^2 \left( \frac{\partial {\varvec{w}_1}}{{\partial {\tau }}} - \lambda \frac{\partial {\varvec{w}_2}}{{\partial \theta }} \right) + \mathcal {O}(\epsilon ^3),\end{aligned}$$
(28)
$$\begin{aligned} \displaystyle \frac{\partial \varvec{w}^\epsilon }{\partial x}&= \epsilon \frac{\partial {\varvec{w}_1}}{{\partial {\theta }}}\frac{\partial {\theta }}{\partial {x}} + \epsilon ^2 \frac{\partial {\varvec{w}_2}}{{\partial \theta }} \frac{\partial {\theta }}{\partial {x}} + \mathcal {O}(\epsilon ^3) = \epsilon \frac{\partial {\varvec{w}_1}}{{\partial \theta }} + \epsilon ^2 \frac{\partial {\varvec{w}_2}}{\partial {\theta }} + \mathcal {O}(\epsilon ^3). \end{aligned}$$
(29)
Taking into account (
24), (
28), and (
29) we obtain
$$\begin{aligned} \displaystyle \frac{\partial \varvec{w}^\epsilon }{\partial t}+\varvec{A}({ \varvec{w}}^\epsilon )\,\frac{\partial \varvec{w}^\epsilon }{\partial x} = \epsilon \left( \varvec{A}( {\varvec{0}}) - \lambda \varvec{I} \right) \frac{\partial {\varvec{w}_1}}{{\partial \theta }} + \epsilon ^2\left( \frac{\partial {\varvec{w}_1}}{{\partial {\tau }}} + \mathcal {B} \varvec{w}_1 \frac{\partial {\varvec{w}_1}}{{\partial \theta }}+ (\varvec{A}({\varvec{0}}) - \lambda \varvec{I}) \frac{\partial {\varvec{w}_2}}{{\partial \theta }} \right) + \mathcal {O}(\epsilon ^3). \;\;\;\;\;\;\; \end{aligned}$$
(30)
Now, gathering together terms that appear with like powers of a small parameter
\(\epsilon \), we equate these consecutive terms to zero and obtain the following equations:
$$\begin{aligned} \!\!\bullet \quad {\mathcal {O}}(\epsilon )\,\, \hbox {terms vanish} \Leftrightarrow \displaystyle (\varvec{A}({\varvec{0}}) - \lambda _j \varvec{I})\,\frac{\partial {\varvec{w}_1}}{{\partial \theta }}= \mathbf{0}. \end{aligned}$$
(31)
Since we focus on shear waves, we take a solution of these equations in the form
$$\begin{aligned} \varvec{w}_1(\tau ,\theta ) = a_1(\tau ,\theta )\;\varvec{r}_1 + a_2(\tau ,\theta )\;\varvec{r}_2, \end{aligned}$$
(32)
with
\(a_j(\tau ,\theta )\) the unknown amplitudes of quasi-shear waves and
\(\varvec{r}_j\) the corresponding eigenvectors of matrix
\(\varvec{A}({\varvec{0}})\);
$$\begin{aligned} \!\!\bullet \quad {\mathcal {O}}(\epsilon ^{2})\,\, \hbox {terms vanish} \Leftrightarrow (\varvec{A}({\varvec{0}}) - \lambda \varvec{I})\, \displaystyle \frac{\partial {\varvec{w}_2}}{{\partial \theta }} = - \varvec{\mathcal {F}}, \end{aligned}$$
(33)
with
$$\begin{aligned} \displaystyle \varvec{\mathcal {F}} \equiv \, \mathcal {B}\, \varvec{w}_{1}\, \frac{\partial {\varvec{w}_1}}{{\partial \theta }}\, + \frac{\partial {\varvec{w}_1}}{{\partial \tau }}. \end{aligned}$$
(34)
The inhomogeneous algebraic equation (
33) has a nonzero solution provided the right-hand side of (
33) is orthogonal to the left eigenvectors of matrix
\( \varvec{A}({\varvec{0}})\). Let us take two such eigenvectors
\(\varvec{l}_1\) and
\(\varvec{l}_2 \), which correspond to a pair of shear waves propagating in the reference state with a common phase speed
\(\lambda \). From the solvability condition we have in particular
$$\begin{aligned} \varvec{l}_1 \cdot \varvec{\mathcal {F}}&= 0,\end{aligned}$$
(35)
$$\begin{aligned} \varvec{l}_2 \cdot \varvec{\mathcal {F}}&= 0. \end{aligned}$$
(36)
Let us introduce the following notation:
$$\begin{aligned} {} \Gamma ^j_{pq} = \varvec{l}_j \cdot \big ( \nabla _{\varvec{w}}\,\varvec{A}({\varvec{w}} ){\big |_{\varvec{w}=\varvec{0}} \big . }\big )\, \varvec{r}_p \, \varvec{r}_q \end{aligned}$$
(37)
for the interaction coefficient. Taking into account (
25), (
32), and (
34) we can write conditions (
35) and (
36) as follows:
$$\begin{aligned}&\!\!\!\displaystyle { \frac{\partial a_{1}}{\partial \tau } + \Gamma ^1_{11} a_{1}\frac{\partial a_{1}}{\partial \theta } + \Gamma ^1_{21} a_{2}\frac{\partial a_{1}}{\partial \theta } + \Gamma ^1_{12}a_{1}\frac{\partial a_{2}}{\partial \theta } + \Gamma ^1_{22}a_{2}\frac{\partial a_{2}}{\partial \theta } = 0,}\end{aligned}$$
(38)
$$\begin{aligned}&\!\!\!\displaystyle { \frac{\partial a_{2}}{\partial \tau } + \Gamma ^2_{11}a_{1}\frac{\partial a_{1}}{\partial \theta } + \Gamma ^2_{21}a_{2}\frac{\partial a_{1}}{\partial \theta } + \Gamma ^2_{12}a_{1}\frac{\partial a_{2}}{\partial \theta } + \Gamma ^2_{22}a_{2}\frac{\partial a_{2}}{\partial \theta } = 0.} \end{aligned}$$
(39)
We obtain the pair of coupled nonlinear partial differential equations that are the asymptotic evolution equations for the amplitudes of the pair of quasi-shear elastic waves in the vicinity of a double umbilic point.