I am currently studying non-equilibrium thermodynamics in the linear regime, and I am wondering why we define the constitutive equations (i.e: the relationship between flux and force) in a basis that is not the eigenbasis of the phenomenologic coefficient matrix. That is to say, if we have two fluxes such that:
$$\begin{cases}J_1=L_{11}X_1+L_{12}X_2 \\ J_2=L_{21}X_1+L_{22}X_2\end{cases}$$
If the system satisfies Onsager's principle: $L_{12}=L_{21}$, which is often the case, at least in my exercises, then I would find it more natural to work in the basis of eigenvectors of the matrix $(L_{ij})$, which I would know to be symmetric and therefore have 2 eigenvalues, and their corresponding eigenvectors would be orthogonal. These new forces and new fluxes would then be independent from one another, just like one decouples two oscillators into their normal modes without giving it much thought, because it makes the problem simpler.
If we have $\lambda_1,\lambda_2$ eigenvalues of $L$, then there exist two thermodynamic forces such that: $L\tilde X_1=\lambda_1\tilde X_1$, $L\tilde X_2=\lambda_2\tilde X_2$, so the constitutive equations are now:
$$\begin{cases}\tilde J_1=\lambda_1\tilde X_1 \\ \tilde J_2=\lambda_2\tilde X_2 \end{cases}$$
If we were interested in knowing the original fluxes, we could then undo the change of variables to work again in the original basis and that would be all. Why don't we do this and prefer to work with the coupled equations?