1 Definition and basic setup

Second variation is the second-order change of a functional under an admissible perturbation of its argument. In the calculus of variations, it is used to study whether an extremal, meaning a critical point of the functional, is locally minimizing, locally maximizing, or neither. The concept parallels the second derivative in ordinary calculus, but it applies to function spaces and often involves integrals, boundary conditions, and constraint sets.

1.1 Functionals and admissible variations

A functional assigns a number to a function, curve, or surface rather than to a finite-dimensional vector. Common examples include integral functionals whose values depend on a function and its derivatives. An admissible variation is a perturbation that preserves the conditions imposed on the problem, such as fixed endpoints or prescribed boundary values. The class of admissible variations determines which directions are allowed when testing optimality.

1.2 One-parameter variation families

A standard way to study variation is to introduce a one-parameter family of trial functions depending smoothly on a parameter. The original extremal corresponds to a particular parameter value, while nearby values represent perturbed configurations. This formulation makes it possible to differentiate the functional with respect to the parameter and isolate first- and second-order terms.

1.3 From first to second variation

The first variation measures the linear response of the functional to a perturbation. At an extremal, this first-order term vanishes for all admissible variations, expressing a necessary condition for criticality. The second variation is the next nonzero term in the expansion and determines the leading behavior of the functional near the extremal when the first variation is zero.

1.4 Notation and conventions

The second variation is often denoted by a symbol such as δ²J or by a quadratic form evaluated on a variation field. In many settings, the perturbation is written as a base function plus a small parameter times a variation function. Conventions differ across texts, especially regarding whether the second variation includes a factor of one half or whether boundary terms are absorbed into the definition.

2 Derivation in the calculus of variations

The derivation of second variation formulas begins with a functional expressed as an integral involving a dependent variable and its derivatives. After differentiating once to obtain the Euler–Lagrange equation, one differentiates again with respect to the variation parameter. The resulting expression typically contains terms quadratic in the variation and its derivatives, together with possible boundary contributions.

2.1 The Euler–Lagrange condition recap

The Euler–Lagrange condition gives the necessary equation satisfied by an extremal of a smooth integral functional. It arises from requiring the first variation to vanish for all admissible perturbations. Once this condition is satisfied, the second variation provides refined information about local behavior near the critical function.

2.2 Computing the second variation formula

To compute the second variation, one expands the functional to second order in the perturbation parameter. The expansion usually produces terms involving the second partial derivatives of the integrand with respect to the function and its derivatives. After integration by parts, the expression is often rewritten in a symmetric form that highlights the associated quadratic form.

2.3 Boundary terms and transversality

When endpoints are not fixed or when admissible variations permit movement along a boundary, additional terms appear in the variation formulas. These are often called transversality terms and encode compatibility conditions at the boundary. Proper treatment of such terms is essential, since they can alter both the second variation and the admissible class of perturbations.

2.4 Regularity assumptions

A meaningful second variation usually requires sufficient smoothness of the functional and the admissible functions. Typical derivations assume differentiability of the integrand with respect to its arguments and enough regularity to justify exchanging differentiation and integration. In more advanced settings, weaker hypotheses can be used, but the formulas may then be interpreted in a distributional or weak sense.

3 Interpretation and classification of extremals

The sign and definiteness of the second variation help classify a critical point. If the quadratic form is positive on all nontrivial admissible variations, the extremal is locally minimizing. If it is negative, the extremal is locally maximizing. If the quadratic form changes sign, the point is a saddle.

3.1 Local minimum vs. local maximum

A positive second variation indicates that nearby admissible perturbations increase the functional to second order. This is the variational analogue of a positive second derivative in one-variable calculus. Conversely, a negative second variation suggests that nearby admissible perturbations decrease the functional, which is associated with local maximality.

3.2 Saddle points and indefiniteness

Many extremals are neither minima nor maxima. In such cases, the second variation may take both positive and negative values depending on the chosen variation. This indefinite behavior signals a saddle point, where the functional rises in some directions and falls in others.

3.3 Positive definiteness criteria

Positive definiteness means that the second variation is strictly positive for every nonzero admissible variation. This is stronger than nonnegativity and usually gives a robust form of local minimality. In practice, verifying positive definiteness may require auxiliary conditions, such as positivity of coefficients or spectral bounds on the associated operator.

3.4 Relation to stability intuition

The language of stability often parallels the classification of extremals. A configuration with positive second variation is typically viewed as stable because small perturbations increase the associated energy or cost. If the second variation is indefinite or negative in some directions, the system may be unstable or only conditionally stable.

4 Sufficient conditions for optimality

The second variation alone is not always enough to guarantee optimality, but it is central in sufficient conditions. Classical results combine positivity of the second variation with structural assumptions on the integrand and boundary conditions. These criteria sharpen the conclusion from mere criticality to local optimality.

4.1 Legendre-type conditions

Legendre-type conditions require the integrand to have suitable convexity with respect to derivative variables. In simple cases, this amounts to positivity of certain second partial derivatives. Such conditions help ensure that the quadratic part of the second variation has the right sign and therefore supports minimality.

4.2 Jacobi fields and their role

Jacobi fields are solutions of the linearized equations associated with the second variation. They describe infinitesimal deformations of an extremal that preserve the Euler–Lagrange structure to first order. The presence or absence of nontrivial Jacobi fields often reflects whether the extremal is isolated or whether the second variation degenerates.

4.3 Conjugate points

Conjugate points mark locations along an extremal where the second variation can lose positivity. They are defined through the existence of nontrivial Jacobi fields satisfying specified boundary conditions. Their occurrence is a classical indicator that a minimizing property may fail beyond a certain interval.

4.4 Strong vs. weak sufficiency

Weak sufficiency refers to minimality with respect to small perturbations measured in a relatively weak topology, while strong sufficiency concerns a stronger notion of closeness, such as uniform proximity. The second variation frequently supports weak optimality results more directly than strong ones. Additional arguments are often needed to pass from local quadratic positivity to a stronger minimizing statement.

5 Necessary conditions and degeneracies

A second variation analysis also reveals when critical points are degenerate or when standard tests are inconclusive. If the quadratic form vanishes in some nontrivial direction, then higher-order terms may determine the behavior. Such cases are important because they can conceal flat directions, symmetries, or families of nearby extremals.

5.1 When the first variation vanishes

The vanishing of the first variation is the hallmark of a critical point. It means that all admissible first-order perturbations leave the functional unchanged to linear order. Only after this cancellation does the second variation become the leading tool for distinguishing local behavior.

5.2 Degenerate critical points

A critical point is degenerate when the second variation has a nontrivial null space. In this situation, the quadratic test does not fully separate minima, maxima, and saddles. Degeneracy often signals the need for a more refined analysis involving symmetry, constraints, or higher-order terms.

5.3 Higher-order variations beyond the second

If the second variation vanishes or is inconclusive, one may examine third or higher variations. These higher-order terms can reveal whether the functional changes on a slower scale along special directions. Although less commonly used, they are important in singular or highly symmetric problems.

5.4 Symmetry and null directions

Symmetries of the functional can produce directions along which the second variation vanishes. For example, invariance under translations or reparametrizations may yield neutral modes. These null directions do not necessarily indicate instability; instead, they often reflect a continuum of equivalent critical configurations.

6 Constrained problems

Many variational problems include constraints that restrict the allowable functions. In such cases, the second variation must be computed only along perturbations compatible with the constraints or after introducing multipliers. The constrained setting often changes both the formula and the interpretation of second-order terms.

6.1 Isoperimetric constraints

Isoperimetric problems fix an auxiliary quantity, such as area, volume, or mass, while optimizing another functional. The admissible variations must preserve the constraint to first order, which modifies the space of test functions. The second variation then measures curvature only within this restricted set of perturbations.

6.2 Lagrange multipliers in second variation

Lagrange multipliers incorporate constraints into an augmented functional. After introducing the multiplier, one computes the second variation of the augmented problem rather than the original one alone. This method produces a unified condition, but the multiplier must be handled consistently when interpreting the final quadratic form.

6.3 Reduced admissible variations

An alternative approach is to eliminate the constraints by restricting to a smaller class of variations. The second variation is then evaluated on that reduced tangent space. This viewpoint is often convenient for deriving clean positivity conditions because the constraints are built into the choice of allowable perturbations.

6.4 Multiple constraints and feasibility

With several constraints, feasibility becomes more delicate because each perturbation must respect all conditions simultaneously. The admissible space may shrink considerably, and the second variation must be examined on the intersection of the corresponding tangent spaces. In some cases, constraint interactions can create degeneracies even when the unconstrained problem is nondegenerate.

7 Examples and worked computations

Concrete examples make the abstract notion of second variation more transparent. In simple integral functionals, the computation reduces to an explicit quadratic form in the variation and its derivative. In geometric or boundary-value settings, the same principle applies, but the formulas often reflect the underlying structure of the problem.

7.1 Simple integral functionals

For a basic functional depending on a function and its derivative, the second variation can often be written as an integral of a quadratic expression. The coefficients are determined by the second derivatives of the integrand. These examples show directly how positivity of the second variation is linked to convexity.

7.2 Variation of geodesics

Geodesics provide a geometric example in which the second variation measures how the length or energy changes under nearby curves. The resulting quadratic form involves curvature terms and the derivative of the variation field along the curve. This setting illustrates how geometry influences stability and the appearance of conjugate points.

7.3 Boundary value problems

In boundary value problems, the admissible perturbations must satisfy prescribed conditions at the endpoints or on the boundary. The second variation then tests whether the given solution minimizes an associated energy among all compatible alternatives. Boundary conditions can strongly affect the sign of the quadratic form.

7.4 Comparison of competing extremals

When multiple extremals solve the same variational problem, the second variation can help distinguish them. One solution may be locally minimizing while another is unstable or only a saddle. Such comparisons are often made by evaluating the quadratic form on representative perturbations or by analyzing the spectrum of the associated operator.

8 Functional-analytic viewpoint

In a more abstract setting, the second variation is viewed as a bilinear form on a function space. This perspective is especially useful in infinite-dimensional analysis, where issues of continuity, compactness, and spectral theory become important. It connects the calculus of variations with operator theory and partial differential equations.

8.1 Second variation as a bilinear form

The second variation is often represented by a symmetric bilinear form obtained by polarization of the quadratic expression. This bilinear form acts on pairs of admissible perturbations and encodes the curvature of the functional at the critical point. Symmetry is a key feature, reflecting the mixed partial derivatives of the underlying integrand.

8.2 Quadratic forms and coercivity

A coercive quadratic form dominates an appropriate norm on the space of variations. Coercivity implies strong control over perturbations and is one of the clearest routes to proving local minimality. It is stronger than mere positivity and is especially valuable in infinite-dimensional problems where compactness may fail.

8.3 Spectral interpretations

When the second variation corresponds to a self-adjoint operator, its sign can be studied through the operator’s spectrum. Positive eigenvalues contribute to stability, while negative eigenvalues indicate directions of decrease. This spectral viewpoint is widely used because it converts a variational question into an eigenvalue problem.

8.4 Weak formulations and Sobolev spaces

Many modern treatments place admissible functions in Sobolev spaces rather than spaces of smooth functions. The second variation is then interpreted weakly, allowing derivatives to exist in an integral sense. This framework is important for proving existence and stability results under minimal regularity assumptions.

9 Connections to other theories

Second variation is closely related to several broader mathematical and scientific ideas. It connects the local study of functionals to numerical methods, differential equations, and optimization. The same quadratic approximation principle appears across many disciplines, though the interpretation varies by context.

Newton’s method relies on a quadratic approximation built from first and second derivatives. The second variation plays a similar role by capturing the leading curvature of a functional near a critical point. In both settings, the second-order term guides local behavior more accurately than the linear term alone.

9.2 Stability in differential equations

Stability analysis for differential equations often begins by linearizing around an equilibrium or special solution. The resulting quadratic forms are closely related to second variations of associated energy functionals. This connection explains why variational methods are so effective in studying stability and bifurcation.

9.3 Variational methods in physics

In physics, many equilibrium states are characterized by stationarity of an action or energy functional. The second variation then indicates whether the state is energetically favorable under small perturbations. This principle appears broadly in classical mechanics, field theory, elasticity, and related areas.

9.4 Second-derivative tests in optimization theory

Finite-dimensional optimization uses second-derivative tests to classify critical points. The second variation is the infinite-dimensional analogue of that idea. Although the setting is more complex, the underlying logic is the same: the sign of the quadratic term near a critical point reveals local shape.

10 Practical considerations and common pitfalls

Applying second variation theory requires careful attention to admissibility, boundary conditions, and the function space under study. A correct formula can still be misused if the variation class is chosen too broadly or too narrowly. Numerical approximations may also obscure the true sign of the second variation.

10.1 Choosing admissible variations correctly

The variation must satisfy the same constraints as the problem, at least to the required order. If the admissible class is incorrect, the computed second variation may test the wrong directions and lead to false conclusions. This is one of the most common sources of error in applications.

10.2 Handling endpoints and constraints

Fixed endpoints, natural boundary conditions, and side constraints all influence the variation formulas. Neglecting boundary contributions can change the classification of an extremal. Careful bookkeeping is especially important when the endpoints themselves are allowed to move.

10.3 Checking sign definiteness rigorously

A quadratic form may appear positive in sample tests yet fail to be positive on the full admissible space. Rigorous verification often requires inequalities, comparison arguments, or spectral analysis. In practice, definiteness should be established on the entire variation space, not just on a few trial functions.

10.4 Numerical approximations and discretization effects

Discrete approximations of variational problems can alter the apparent sign structure of the second variation. A mesh or finite-dimensional model may introduce spurious stability or hide genuine degeneracy. For reliable conclusions, numerical results should be interpreted alongside the underlying continuous theory.