Here we introduce the SPMS method. Though this method is described in Miura and Kida[9] and Kida and Miura[8], we summarize how to implement this method in order to explain what we are doing to extract vortex structures in turbulence and how to visualize them. The essence of this method is explained by the following four processes. First, we find planes on which the pressure takes sectional-minima because we expect that swirling motions of fluids tend to reduce the pressure inside swirling motions. (It should be noted here that existence of sectional-pressure-minimum is neither the necessary condition nor the sufficient condition for swirling motions to exist there.) This procedure is achieved by considering the hessian of the pressure, . Suppose that the three real eigenvalues of the pressure hessian is given as . By using these three eigenvalues, the pressure can be written as
up to the second order where the coordinate axis of s are given by the three eigenvectors of the pressure hessian. Then, find grid points such that because the pressure takes a local minimum on a plane perpendicular to the third eigenvector associated with the eigenvalue . Thus we expect that a line l which is parallel to the third eigenvector and drawn on is a central axis of the vortex. Hereafter, we concentrate on points which are foot-points of vertical lines dropped from grid points to corresponding lines l.
Next, choose s (simply called ``candidates'' of skeletons) which are located in the neighborhood of s so that candidates far from s may be contaminated by various numerical errors and give zig-zag- or inappropriately connected skeletons. In other words, it is sufficient to adopt s neighboring to s to construct skeletons. This condition is expressed by an inequality .
Then we impose the swirl condition on the planes of the pressure-minimum so that points of the sectional-minimum without swirling motions are removed from the candidates. Project a flow field on a plane perpendicular to the third eigenvector and linearize the velocity field around . Then we have
on the frame moving with the velocity at . Then a condition for the flow to be elliptic or spiral on this plane is given by the negative of the discriminant of matrix A:
This discriminant is expressed by quantities of the original -coordinate system as
where angles , which give a rotational transform from the -coordinate system to -coordinate system, are determined by utilizing the third eigenvalue of the pressure hessian. By imposing on candidates of skeletons, we have less number of candidates of skeletons.
At last, we have a set of candidates of skeletons which are relatively less contaminated by numerical errors and accompanied with swirling motions. By connecting s, we have skeletons which represent central axes of vortices.
Once we have obtained skeletons of swirling vortices, we can define vortex cores without ambiguity. These cores are obtained by looking for the negative discriminant around the skeletons where are given by the third eigenvectors on the skeletons. We here put short comment on visualization of vortex cores. The vortex cores can be expressed in the three-dimensional graphics easily by using polygons. First, we find outer boundary of a cross-section of the vortex core. If we have two neighboring cross-sections, we put polygons (in many cases they are triangles) to cover the region between these two cross-sections. Similar representation of skeletons and vortex cores has been proposed by Banks and Singer[10], although there are some ambiguity in their definition of skeletons and cores.
In the next section, we apply the SPMS method to extract vortical structures of turbulence and analyze them.