Multivariable Calculus 25 | Implicit Function Theorem [dark version]

TL;DR

Split variables into x∈R^K and y∈R^M so the equation F(x,y)=0 becomes M equations in K+M unknowns.

Briefing Cornell Notes

Briefing

The implicit function theorem turns “contour lines” of a system of equations into honest-to-goodness local graphs—provided a specific Jacobian block is invertible. In practical terms, if a C1 vector-valued function F(x,y) = 0 defines M equations in N = K + M variables, then near a solution (x0,y0) the solution set can be written uniquely as y = G(x) on some neighborhood. That matters because it converts messy implicit constraints into a form where calculus (derivatives, linearization, local analysis) can be applied directly.

The setup starts with an open domain U ⊂ RN and a C1 map F: U → RM. At a chosen point u0 = (x0,y0) with F(u0) = 0, the variables are split into x ∈ RK and y ∈ RM. Geometrically, the equation F(x,y)=0 defines a generalized contour set: in two dimensions it would look like a curve where F vanishes. But not every such set is a graph. The theorem’s key condition rules out “turning back” behavior by requiring that the dependence on the y-variables is locally non-degenerate.

Concretely, the Jacobian of F at u0 is an M×N matrix, naturally partitioned into two blocks: D_yF(u0), an M×M matrix of partial derivatives with respect to the y-variables, and D_xF(u0), an M×K matrix of partial derivatives with respect to the x-variables. The crucial hypothesis is that D_yF(u0) is invertible—equivalently, det(D_yF(u0)) ≠ 0. This is the higher-dimensional analogue of the familiar single-variable condition “∂f/∂y ≠ 0” when there is only one y-variable.

Under that invertibility assumption, there exist open neighborhoods V1 ⊂ RK around x0 and V2 ⊂ RM around y0, along with a C1 function G: V1 → V2, such that every point (x,y) in V1×V2 satisfying F(x,y)=0 must have y = G(x). In other words, within the chosen neighborhoods, the zero set of F is exactly the graph of G—no extra branches of the contour line sneak into the local rectangle.

The theorem also provides a derivative formula for G. Differentiating the identity F(x, G(x)) = 0 yields the Jacobian of G in terms of the Jacobian blocks of F: D G(x) = −(D_yF(x, G(x)))^{-1} · D_xF(x, G(x)). Evaluated at x0, this gives the local linear approximation of the implicit solution y = G(x). The result is more than existence: it justifies treating implicit equations as locally solvable for the y-variables, enabling computation and further analysis in later examples.

Cornell Notes

The implicit function theorem addresses systems of equations F(x,y)=0 where x∈R^K and y∈R^M, with F:U→R^M of class C1. At a point (x0,y0) satisfying F(x0,y0)=0, the theorem requires that the y-Jacobian block D_yF(x0,y0) (an M×M matrix) is invertible (det ≠ 0). When this holds, there are neighborhoods V1 around x0 and V2 around y0 and a C1 function G:V1→V2 such that, inside V1×V2, the entire solution set is exactly the graph y=G(x). It also gives a derivative formula: D G(x)=−(D_yF)^{-1}D_xF evaluated along (x,G(x)). This matters because it converts implicit constraints into a differentiable local graph.

Why does invertibility of D_yF(x0,y0) matter for turning F(x,y)=0 into y=G(x)?

Invertibility of the M×M matrix D_yF(x0,y0) prevents the zero set from “folding” in a way that would stop it from being representable as a graph over x. In the 2D intuition, the curve F(x,y)=0 has a tangent direction determined by the gradient; requiring the y-derivative not to vanish ensures the curve crosses vertical lines only once locally. In higher dimensions, D_yF being invertible plays the same role: it guarantees local uniqueness of y for each nearby x, so the contour set inside a small rectangle becomes exactly y=G(x).

How are the Jacobian blocks D_xF and D_yF defined when F is vector-valued?

With F:U⊂R^{K+M}→R^M, the Jacobian at a point is an M×(K+M) matrix. Splitting variables into x=(x1,…,xK) and y=(y1,…,yM) partitions the Jacobian into two blocks: D_xF is the M×K matrix of partial derivatives ∂F/∂xj, and D_yF is the M×M matrix of partial derivatives ∂F/∂yi. The theorem’s condition focuses on D_yF because it measures how strongly the equations respond to changes in the y-variables.

What does the theorem guarantee about the solution set inside V1×V2?

Once V1⊂R^K and V2⊂R^M are chosen small enough, every pair (x,y) in V1×V2 satisfying F(x,y)=0 must satisfy y=G(x). That means there are no other nearby branches of the contour line inside the local product neighborhood; the zero set coincides with the graph of G over V1.

How does the derivative formula for G arise from F(x,G(x))=0?

Because F(x,G(x)) is identically zero for x in V1, differentiating with respect to x gives a matrix equation involving D_xF and D_yF. Solving for D G(x) yields D G(x)=−(D_yF(x,G(x)))^{-1}·D_xF(x,G(x)). The inverse exists precisely because D_yF is invertible at the base point and remains invertible in a neighborhood (continuity).

What is the special case when M=1, and how does it relate to the usual single-equation implicit function theorem?

When M=1, y is a single variable and D_yF becomes a 1×1 matrix, i.e., just the scalar partial derivative ∂F/∂y. The invertibility condition reduces to ∂F/∂y ≠ 0 at (x0,y0). The general theorem then reproduces the familiar criterion for writing y as a differentiable function of x near a solution.

Review Questions

Given F:U⊂R^{K+M}→R^M and a point (x0,y0) with F(x0,y0)=0, which Jacobian block must be invertible to solve locally for y as a function of x?
State the local form of the solution set guaranteed by the implicit function theorem and describe how it is represented using a function G.
Write the formula for D G(x) in terms of D_yF and D_xF, and explain what must be true for the inverse (D_yF)^{-1} to exist.

Key Points

1
Split variables into x∈R^K and y∈R^M so the equation F(x,y)=0 becomes M equations in K+M unknowns.
2
At a solution u0=(x0,y0), require the y-Jacobian block D_yF(u0) (an M×M matrix) to be invertible (det ≠ 0).
3
Under that condition, there exist neighborhoods V1 around x0 and V2 around y0 where the solution set inside V1×V2 is exactly the graph y=G(x).
4
The function G is C1, giving a differentiable local representation of the implicit solutions.
5
The derivative of G satisfies D G(x)=−(D_yF(x,G(x)))^{-1}·D_xF(x,G(x)).
6
Invertibility at the base point extends to nearby points because D_yF depends continuously on (x,y).

Highlights

The theorem’s entire solvability hinges on the invertibility of the y-derivative block D_yF, not the full Jacobian.

Within small neighborhoods, the zero set of F(x,y)=0 becomes precisely the graph y=G(x), with no extra branches.

The Jacobian of the implicit function is computed by D G(x)=−(D_yF)^{-1}D_xF evaluated along the solution curve.

For M=1, the condition collapses to the familiar requirement ∂F/∂y ≠ 0.

Topics

Implicit Function Theorem
Jacobian Block Invertibility
Local Graph Representation
Vector-Valued Equations
Derivative Formula