Skip to content
Symi Docs
Swift
On this page

Partial differential equations

A partial differential equation relates an unknown function of several variables to its partial derivatives. Everything on this page starts from a PartialDifferentialEquationProblem: a validated signature that fixes the equation, the dependent function, the ordered independent variables, and any initial or boundary data. Constructing the problem is a separate step from solving it, so that a malformed equation is rejected once, at construction, rather than reappearing as a strange answer from a solver.

The unknown and its derivatives are written with undefined functions.

The records these entries return — problems, classifications, principal parts, transformations, symmetry algebras, solution families, and verification reports — are documented together in Partial differential equation results.

Constructing a problem#

partial_differential_equation#

Explicit context

func partialDifferentialEquation(
    equation: UniffiExpression,
    dependentFunction: UniffiUndefinedFunction,
    independentVariables: [String],
    initialConditions: [UniffiExpression],
    boundaryConditions: [UniffiExpression],
) -> UniffiPartialDifferentialEquationProblem

Construct a validated scalar PDE with an explicit dependent function and an explicit independent-variable order. The order is part of the signature: it fixes which coordinate the classifier and every transformation call the first axis, so the same equation given two orders is two problems. Conditions are residuals read as equal to zero.

Errors: an equation that does not involve the declared dependent function, a repeated or undeclared independent variable, or a condition outside the declared coordinates is an invalid-argument error naming the offending parameter.

discover_partial_differential_equation#

Explicit context

func discoverPartialDifferentialEquation(
    equation: UniffiExpression,
    initialConditions: [UniffiExpression],
    boundaryConditions: [UniffiExpression],
) -> UniffiPartialDifferentialEquationProblem

Recover the same signature from one governing residual alone, by reading the dependent function and the independent variables out of the derivatives that appear in it. It declines rather than guessing when the residual does not determine a unique signature — more than one undefined function, or a coordinate order the residual leaves ambiguous. Prefer partial_differential_equation when the signature is known; use this entry when the equation is the only thing in hand.

partial_differential_equation_arbitrary_function#

Explicit context

func partialDifferentialEquationArbitraryFunction(
    preferredName: String,
    arity: UInt64,
) -> UniffiPartialDifferentialEquationArbitraryFunction

Allocate a collision-free arbitrary function of the stated arity, for use as a free element of a candidate solution family. The allocated name is derived from preferred_name but is guaranteed not to collide with a function already present in the context, so a family's free elements cannot be captured by a name the problem already uses.

Classifying#

classify_partial_differential_equation#

Explicit context

func classifyPartialDifferentialEquation(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationClassification

Exact structural analysis of a validated scalar PDE: its order, its linearity, its principal part, and the coefficient data each of those is read from. The analysis is structural and exact — a coefficient whose sign or reality cannot be proved is reported as undetermined rather than assumed.

classify_second_order_partial_differential_equation#

Explicit context

func classifySecondOrderPartialDifferentialEquation(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationSecondOrderClassification

The elliptic/parabolic/hyperbolic type of a two-variable linear second-order PDE over real coordinates, taken from the discriminant of the principal part's symmetric matrix. The type is reported per case, with a verdict, because a discriminant with a symbolic sign is of one type on one region and another type elsewhere; a type is claimed only where its sign condition is proved.

Changing coordinates#

change_partial_differential_equation_variables#

Explicit context

func changePartialDifferentialEquationVariables(
    problem: UniffiPartialDifferentialEquationProblem,
    targetVariables: [String],
    forwardCoordinates: [UniffiExpression]?,
    inverseCoordinates: [UniffiExpression]?,
    targetDependentFunction: UniffiUndefinedFunction?,
) -> UniffiPartialDifferentialEquationTransformationResult

Transform a two-variable PDE through an exact user-supplied change of variables. The forward and inverse coordinate maps are both required for the chain rule to be applied exactly; the transformation is carried out symbolically, never numerically, and the result carries the transformed problem together with the maps used.

canonicalize_partial_differential_equation#

Explicit context

func canonicalizePartialDifferentialEquation(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationCanonicalizationResult

Construct and independently verify exact canonical coordinates for a supported PDE — the coordinates in which a second-order equation takes its canonical principal form for its type. The coordinates are derived and then checked by transforming the original equation into them, so a canonicalization is reported only when its own verification passes.

verify_partial_differential_equation_coordinate_transformation#

Explicit context

func verifyPartialDifferentialEquationCoordinateTransformation(
    problem: UniffiPartialDifferentialEquationProblem,
    transformation: UniffiPartialDifferentialEquationTransformationResult,
) -> UniffiPartialDifferentialEquationTransformationVerificationReport

Independently check a coordinate transformation without producing one: the coordinate maps are confirmed mutually inverse and the transformed equation is confirmed to be the exact pullback of the original. No transformation runs, so a supplied transformation is reported as given rather than repaired.

Solving#

solve_partial_differential_equation#

Explicit context

func solvePartialDifferentialEquation(
    problem: UniffiPartialDifferentialEquationProblem,
    method: UniffiPartialDifferentialEquationMethod,
) -> UniffiPartialDifferentialEquationSolveResult

The sound PDE dispatcher under a stable method name. automatic lets the dispatcher choose among the implemented routes; a named method runs only that route and declines when it does not apply, so a named request is never silently answered by a different method.

Use solve_partial_differential_equation_request when the scope, geometry, regularity, or solution class is part of the question.

solve_partial_differential_equation_request#

Explicit context

func solvePartialDifferentialEquationRequest(
    problem: UniffiPartialDifferentialEquationProblem,
    scope: UniffiDifferentialEquationSolutionScope,
    geometricDomain: UniffiPartialDifferentialEquationGeometricDomain,
    regularity: UniffiPartialDifferentialEquationRegularity,
    solutionClass: UniffiPartialDifferentialEquationSolutionClass,
    method: UniffiPartialDifferentialEquationMethod,
) -> UniffiPartialDifferentialEquationSolveResult

Solve a PDE request in which the geometry, regularity, and solution class are part of the question. geometric_domain is whole_space, disk (radial then angular variable, origin interior), or real_line_evolution (whole-line spatial variable then positive evolution variable). regularity is unrestricted or origin_regular_periodic (disk only). solution_class is unrestricted or polynomial_growth.

A general request is answered only by a family whose completeness theorem is proved over the requested class. Otherwise the result declines with complete_family_not_established or completeness_holds_only_on_a_narrower_solution_class — the latter is the verdict for the unrestricted heat request, whose initial-data representation is complete only among solutions of polynomial growth. Geometry or regularity no implemented method supports declines with declared_geometry_not_supported or declared_regularity_not_supported.

solve_partial_differential_equation_with_transform_axis_specifications#

Explicit context

func solvePartialDifferentialEquationWithTransformAxisSpecifications(
    problem: UniffiPartialDifferentialEquationProblem,
    method: UniffiPartialDifferentialEquationMethod,
    transformAxisSpecifications: [String],
) -> UniffiPartialDifferentialEquationSolveResult

Solve through a transform route with the axis geometry stated explicitly and in order: one specification per axis, naming the axis kind the transform is to treat that coordinate as. A transform route needs to know which coordinate is the whole line and which is the positive evolution direction; this entry makes that part of the request rather than something a solver infers.

separate_partial_differential_equation#

Explicit context

func separatePartialDifferentialEquation(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationSeparationResult

Separate a supported PDE into exact one-variable ODEs. The result carries the separated equations, the separation constants introduced, and the strategy that produced them; each separated equation is an ordinary differential equation that the ODE routes can close.

Point symmetries#

derive_partial_differential_equation_point_symmetries#

Explicit context

func derivePartialDifferentialEquationPointSymmetries(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationPointSymmetryAlgebra

Derive the point-symmetry algebra of a scalar first- or second-order PDE from its prolonged determining equations. Each basis generator is verified by an independent infinitesimal-criterion check. For a linear homogeneous equation the additive solution symmetry is reported separately (admits_additive_solution_symmetry), because it is constrained to solutions rather than being an arbitrary function. Problems carrying initial or boundary data decline, as do infinite-dimensional algebras the bounded reduction cannot close. The result carries decline, parameters, basis, general_generator, determining_equations, and outstanding_obligations; discrete symmetries are never claimed.

partial_differential_equation_point_symmetry_generator#

Explicit context

func partialDifferentialEquationPointSymmetryGenerator(
    problem: UniffiPartialDifferentialEquationProblem,
    coordinateInfinitesimals: [UniffiExpression],
    dependentInfinitesimal: UniffiExpression,
) -> UniffiPartialDifferentialEquationPointSymmetryGenerator

A candidate generator, one coordinate infinitesimal per independent variable and a dependent infinitesimal written in the unknown evaluated at them. It is a candidate, not a certificate: every consumer verifies it is a symmetry.

construct_partial_differential_equation_point_symmetry_action#

Explicit context

func constructPartialDifferentialEquationPointSymmetryAction(problem: UniffiPartialDifferentialEquationProblem) -> UniffiPartialDifferentialEquationPointSymmetryAction

Integrate every basis generator into a one-parameter group, verify each group by its own intertwining identity, and certify their composition as the connected local action u = a + A f(X). The result carries parameters, transformed_coordinates, amplitude, multiplier, intertwining, invertibility, outstanding_obligations, and decline or point_symmetry_decline. applied_to_seed(seed) returns A f(X) and applied_to_seed_with_additive_solution(seed, additive_solution) returns a + A f(X); the transformed coordinates must lie in the seed's chart.

solve_partial_differential_equation_by_similarity_reduction#

Explicit context

func solvePartialDifferentialEquationBySimilarityReduction(
    problem: UniffiPartialDifferentialEquationProblem,
    generator: UniffiPartialDifferentialEquationPointSymmetryGenerator,
) -> UniffiPartialDifferentialEquationSimilaritySolution

Reduce a two-variable PDE by a fiber-linear point symmetry to an ODE in the invariant, solve that ODE as a complete family, and lift the family back. The result carries invariant, similarity_variable, weight, reduced_equation, multiplier, lifted_family, lifted_constants, lifted_family_verification, outstanding_obligations, and decline. The lifted family is a family of invariant solutions and always carries the obligation that it is not the general solution.

Candidate solutions and verification#

A candidate family is stated, then checked. The two are deliberately separate operations: a verified residual proves membership in the solution set and never coverage of it, so a family that is verified is not thereby complete.

explicit_partial_differential_equation_solution#

Explicit context

func explicitPartialDifferentialEquationSolution(solution: UniffiExpression) -> UniffiPartialDifferentialEquationSolutionFamily

State a candidate family in explicit form — the unknown written as an expression in the coordinates — together with the arbitrary functions it treats as free and any proof obligations it carries. Stating a family asserts nothing; verify_partial_differential_equation_solution is what establishes it.

implicit_partial_differential_equation_solution#

Explicit context

func implicitPartialDifferentialEquationSolution(residual: UniffiExpression) -> UniffiPartialDifferentialEquationSolutionFamily

State the same kind of candidate family as a residual read as equal to zero, for a family that cannot be solved for the unknown in closed form.

verify_partial_differential_equation_solution#

Explicit context

func verifyPartialDifferentialEquationSolution(
    problem: UniffiPartialDifferentialEquationProblem,
    solution: UniffiPartialDifferentialEquationSolutionFamily,
) -> UniffiPartialDifferentialEquationVerificationReport

Independently verify a candidate family against a problem by exact residual reasoning: substitute the family into the equation and every declared condition, and reduce the residuals symbolically. The report carries a verdict, the residuals that were not proved zero, and any outstanding obligations, so an inconclusive verification is distinguishable from a refutation.

Symi documentation · Swift