Stability of Equilibria in the Calculus of Variations


Isoperimetric Conjugate Points (cf. Manning, Rogers & Maddocks, 1998 )

Suppose that we have a solution to the Euler-Lagrange equilibrium equations for some calculus of variations functional. This computed equilibrium is a critical point of the functional, but not necessarily a stable local minimum. In the classic theory of the unconstrained calculus of variations, the Jacobi/Morse conjugate point theory is used to classify critical points as local minima or some other type. The number of conjugate points is called the index, and should give the number of downward-turning directions on the (infinite-dimensional) graph of the functional in the neighborhood of the critical point.

This conjugate point test can be generalized to isoperimetrically constrained problems. This generalization was done over a century ago for some particular cases by Bolza and others, e.g. for one or two variables and a few constraints. In Manning, Rogers & Maddocks, 1998, we have treated the general isoperimetric problem, framing the process in modern functional analytic language. With the appropriate inclusion of projection operators (the projection being onto the space orthogonal to the linearized constraints), the isoperimetric theory can be made to look almost exactly parallel to the unconstrained theory.

For example, the twisted ring is described by an isoperimetric calculus of variations problem. In the bifurcation diagram for the "perfect" (intrinsically straight with circular cross-section) rod (recreated below), the index is denoted by color: green = 0 (local minimum), red = 1 (saddle with one downward-turning direction), light blue = 2, orange = 3, purple = 4, and dark blue = 5.




Buckling into a soft wall (cf. Bulman and Manning, 2001 (preprint) )

Another extension to the conjugate point theory allows its application to cases where the second variation operator is more complicated than a second-order differential operator as in the classic theory. For example, in contact problems, the second variation operator is often an integrodifferential operator. One simple contact problem is the buckling of an elastic rod in the presence of an external field designed to mimic contact with a wall (hence, the field acts as a sort of "soft" wall).



By virtue of the general functional analytic approach taken to isoperimetric conjugate points in Manning, Rogers & Maddocks, 1998, the theory can be extended relatively easily to this soft wall case. All that is required are some basic properties of the integrodifferential second variation operator, which are readily verified. In addition, although in general one would expect the conjugate point equation to become an integrodifferential equation, in this case it can be transformed into a differential equation for easy numerical solution. We have used this test to determine the stability of various equilibria of the soft wall problem, for various values of the rod and wall parameters.



(This material is based upon work supported by the National Science Foundation under Grant No. DMS-9973258. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).)

Conjugate points for parameter-dependent problems (cf. Manning and Hoffman, 2001 and Hoffman, Manning, and Paffenroth, 2001 (preprint))

The above theory is designed primarily for numerical implementation, since, in most cases, it is impossible to solve the conjugate point equation in closed form. Sometimes, however, an analytic solution is possible, or at least a semi-analytic one (one which involves solving algebraic but not differential equations numerically). This is particularly the case when a parameter appears linearly in the second-variation operator, in which case we have shown in Hoffman, Manning, and Paffenroth, 2001 (preprint) that one can completely determine stability by locating values of the parameter where there are conjugate points at s=1. Since conjugate points at s=1 are often easier to determine than conjugate points at general s, this is a significant simplification. We have applied this technique to two problems:

* The stability of multi-covered circular configurations of an elastic rod with constant planar intrinsic curvature, as a function of the intrinsic curvature kappa and the rod stiffness parameters rho and gamma (rho is ratio of the out-of-plane bending stiffness to the in-plane bending stiffness, and gamma is the ratio of the twisting stiffness to the in-plane bending stiffness). The index can be determined in closed form; for example, for an intrinsically straight rod, the 1-covered circle is stable if and only if rho>=1 (a fact that was previously known), while for n>1, the n-covered circle is stable if and only if rho>1, gamma>1, and (n-1)2/n2 < (gamma-1)(rho-1)/(gamma rho). Below we show the dependence of the index on n, kappa, rho, and gamma. The subfigure columns are for n=1, 2, and 3, while the rows are for increasing values of kappa as labeled. In each subfigure, we color the gamma-rho plane by the index of the n-covered circle for that value of rho, gamma, n, and kappa.





* The stability of an intrinsically straight elastic strut under twisting and endloading, as a function of the twist, endloading, and rod anisotropy (we consider rods with elliptical cross-section). In this case we can determine the index semi-analytically, in that there are no differential equations to be solved, but an algebraic equation remains for determining the index. Below we show the dependence of the index on rho (as in previous example), gamma (as in previous example), alpha (twist angle), and lambda (endloading). The subfigure columns are for gamma = 1/2, 1, and 2, while the rows are for increasing values of rho as labeled. In each subfigure, we color the alpha-lambda plane by the index of the straight strut for that value of alpha, lambda, gamma, and rho, using the color scheme shown at the top.







For this problem, we have also computed families of buckled configurations numerically, and assigned stabilities by the numerical isoperimetric conjugate point test described above. These results include force-extension curves for lambda varying at fixed values of rho, gamma, and alpha; and also 2-dimensional bifurcation surfaces for alpha and lambda varying at fixed rho and gamma, as shown below:

See Hoffman, Manning, and Paffenroth, 2001 (preprint) for more examples.

(This material is based upon work supported by the National Science Foundation under Grant No. DMS-9973258. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).)

Return to my home page
rmanning@haverford.edu