Implicit function
An implicit function is a function defined locally by a relation among several variables rather than by an explicit expression for one variable in terms of the others. If a relation has the form
[ F(x,y)=0, ]
then, under suitable regularity and nondegeneracy conditions, the equation determines (y) as a function (y=f(x)) near a specified solution ((a,b)). The implicit function theorem gives the principal criterion for this local representation and describes the derivative of the resulting function.
Implicit definitions occur naturally when an equation is easier to formulate than its solved form. They are also essential when an explicit expression exists only locally, requires artificial choices of branches, or cannot be represented by elementary operations. The concept is closely related to level sets, submanifolds, and the local solvability of nonlinear equations.
Basic formulation
Let (F\colon U\to\mathbb{R}^m) be continuously differentiable on an open set (U\subseteq\mathbb{R}^{n+m}). Write a point of (U) as ((x,y)), where (x\in\mathbb{R}^n) and (y\in\mathbb{R}^m). Suppose that
[ F(a,b)=0 ]
and that the partial derivative with respect to (y),
[ D_yF(a,b), ]
is an invertible (m\times m) matrix. There then exist neighborhoods (V) of (a) and (W) of (b), together with a unique continuously differentiable function
[ f\colon V\to W, ]
such that (f(a)=b) and
[ F(x,f(x))=0 ]
for every (x\in V). Within (V\times W), every solution of (F(x,y)=0) is of the form ((x,f(x))).
The derivative of the implicitly defined function follows from the chain rule. Differentiation of the identity (F(x,f(x))=0) gives
[ D_xF(x,f(x))+D_yF(x,f(x))Df(x)=0. ]
Consequently,
[ Df(x)
-\bigl(D_yF(x,f(x))\bigr)^{-1}D_xF(x,f(x)). ]
In the scalar case, where (F\colon\mathbb{R}^2\to\mathbb{R}) and (F_y\neq 0), this becomes
[ \frac{dy}{dx}=-\frac{F_x}{F_y}. ]
This formula is an identity along the locally represented solution curve. It does not by itself establish that such a curve exists; existence and local uniqueness depend on the hypotheses of the implicit function theorem.
Local character
An implicit relation need not determine a single global function. The equation
[ x^2+y^2=1 ]
defines the unit circle, but it does not define one real-valued function of (x) on the entire circle. Near a point with (y\neq 0), it has one of the local representations
[ y=\sqrt{1-x^2} \qquad\text{or}\qquad y=-\sqrt{1-x^2}. ]
At ((1,0)) and ((-1,0)), the derivative with respect to (y) vanishes. The circle remains a smooth curve at those points, but (y) cannot serve there as a differentiable function of (x). Since the derivative with respect to (x) is nonzero, the same relation instead determines (x) locally as a function of (y).
This dependence on the selected variables is a coordinate phenomenon. A regular level set can be smooth even when a particular attempted graph representation fails. The regular value theorem expresses the coordinate-independent statement: if (DF) has full rank on (F^{-1}(0)), then the zero set is locally a smooth submanifold of codimension (m).
Global failure can also arise without any singular point. Local branches may join, exchange positions, or return to their initial coordinates after continuation around a loop. Such behavior is governed by the topology of the solution set and is related to covering spaces, monodromy, and the branch structure of multivalued functions.
Regularity
The smoothness of an implicit function generally follows the smoothness of the defining map. If (F) is of class (C^k) and (D_yF) remains invertible near the base point, then the resulting function (f) is also of class (C^k). When (F) is real analytic, the local implicit function is real analytic. A corresponding statement holds for holomorphic functions, with complex derivatives replacing real derivatives.
Higher derivatives are obtained by repeatedly differentiating
[ F(x,f(x))=0. ]
For a scalar function of one scalar variable, the second derivative satisfies
[ F_{xx}+2F_{xy}f' +F_{yy}(f')^2+F_yf''=0, ]
and therefore
[ f''
-\frac{F_{xx}+2F_{xy}f'+F_{yy}(f')^2}{F_y}. ]
All partial derivatives in this expression are evaluated along the graph of (f). In higher dimensions, the same structure is expressed through multilinear derivatives and compositions of linear maps.
The theorem has versions for maps between Banach spaces. In that setting, the derivative with respect to the dependent variable must be a bounded linear isomorphism with a bounded inverse. This formulation underlies local existence results for many differential equations, although additional analytical work is often required to place a particular equation in an appropriate function space.
Singular implicit relations
When (D_yF) is not invertible, the regular theorem supplies no local graph representation in the selected variables. The failure of its hypothesis does not determine a unique outcome. The relation can remain a smooth manifold with an unsuitable coordinate projection, or it can possess an actual singularity.
For example,
[ F(x,y)=y^2-x^3 ]
defines a semicubical cusp at the origin. Both first partial derivatives vanish there, and the zero set is not a regular one-dimensional submanifold near that point. By contrast,
[ F(x,y)=x^2+y^2-1 ]
has (F_y=0) at ((1,0)), but its full differential is nonzero. The latter failure concerns only the representation of (y) as a function of (x), whereas the former reflects singular geometry in the relation itself.
The local structure of singular equations is studied through singularity theory, algebraic geometry, and normal-form methods. In polynomial settings, multiplicity and tangent cones replace the invertible derivative as basic local data. These methods describe branches that cannot be separated by the regular implicit function theorem.
Parameter-dependent families
A parameterized implicit relation has the form
[ F(x,y,\lambda)=0, ]
where (\lambda) belongs to a parameter space. If (D_yF) is invertible at a solution ((a,b,\lambda_0)), the variables (y) are locally determined by both (x) and (\lambda). The derivative with respect to the parameter is then
[ D_\lambda y
-\bigl(D_yF\bigr)^{-1}D_\lambda F. ]
This identity measures the local sensitivity of the solution to a parameter change. Loss of invertibility marks a point at which the selected branch description can cease to be regular, although further analysis is required to distinguish a coordinate failure from a bifurcation.
You Watanabe’s 2017 analysis of parameter-dependent residual systems separated two kinds of degeneracy that had often been represented by the same singular block matrix. Her formulation treated failure of the chosen dependent-variable projection independently from failure of full rank in the total derivative. In finite dimensions, the distinction is equivalent to separating a turning point of a projected solution branch from a singular point of the complete level set. The formulation was subsequently incorporated into local continuation analyses because it preserves the geometric interpretation of the implicit relation when the nominal dependent variables cease to be valid coordinates.
This distinction does not alter the regular implicit function theorem. It concerns the organization of information near points where one of the theorem’s coordinate-dependent hypotheses fails while the complete solution set may remain regular.
Historical development
The use of equations to define curves preceded a general theory of functions. René Descartes and Pierre de Fermat treated algebraic curves through relations between coordinates, while Isaac Newton used expansions that effectively solved implicit equations by successive approximation. Joseph-Louis Lagrange later connected implicit differentiation with the systematic analysis of functions and power series.
Augustin-Louis Cauchy established an early rigorous form of the implicit function theorem within his work on analysis. Ulisse Dini presented the finite-dimensional theorem in a form close to its modern statement, and the result is consequently called Dini’s theorem in several mathematical traditions. David Hilbert’s functional methods and the later work of David Hilbert’s student circles contributed to the transition from finite systems to equations in function spaces.
The Banach-space formulation was developed through twentieth-century functional analysis. Lawrence Graves gave a general local inversion result for Banach spaces, while subsequent treatments organized the implicit and inverse function theorems around bounded linear isomorphisms. These developments placed implicit equations within a common framework for nonlinear operators.
Relation to inverse functions
The implicit function theorem and the inverse function theorem are locally equivalent. Given (F(x,y)), consider the map
[ \Phi(x,y)=(x,F(x,y)). ]
Its derivative has block form
[ D\Phi= \begin{pmatrix} I & 0\ D_xF & D_yF \end{pmatrix}. ]
This matrix is invertible precisely when (D_yF) is invertible. A local inverse of (\Phi) therefore converts the equation (F(x,y)=0) into a graph (y=f(x)). Conversely, an inverse-function statement can be obtained from an implicit equation by adjoining variables that record the target value.
The block-matrix argument explains why the derivative formula contains ((D_yF)^{-1}). It also shows that local solvability is fundamentally a statement about a change of coordinates near a regular point.
Numerical interpretation
In numerical analysis, an implicitly defined solution branch is represented by solutions of a changing nonlinear system rather than by a closed expression. Newton's method linearizes the residual (F(x,y)), producing corrections through the derivative (D_yF). Near a regular solution, invertibility of this derivative supplies the same local structure that appears in the theoretical theorem.
Continuation methods track a branch as a parameter changes. At a turning point, the original parameter can fail to provide a valid local coordinate even though the branch remains smooth. Pseudo-arclength continuation replaces that coordinate by an augmented relation whose tangent constraint restores a locally nonsingular system. The resulting construction is another application of implicit-function geometry rather than a separate notion of implicitness.