The Implicit Function Theorem (IFT) is a fundamental result in multivariable calculus that determines when a relation defined by a [set](/page/Set) of equations can be locally solved to express some variables as [functions](/page/Function) of others. It guarantees that if a system of non-linear equations is "smooth enough" and has a non-singular linearization with respect to a subset of variables, the level set locally looks like the graph of a function.

## Motivation

Consider the unit circle defined by $x^2 + y^2 - 1 = 0$. We often want to solve for $y$ in terms of $x$, yielding $y = \sqrt{1-x^2}$. This is possible essentially everywhere except at the points $(1, 0)$ and $(-1, 0)$, where the tangent is vertical. The IFT generalizes this to high-dimensional systems: we can solve for "dependent" variables in terms of "independent" variables precisely when the derivative restricted to the dependent variables is invertible.

[quotetheorem:52]

## Secondary Definitions

[definition:RegularPoint] A point $p \in \mathbb{R}^k$ is called a **regular point** of a map $F: \mathbb{R}^k \to \mathbb{R}^m$ (where $k > m$) if the total derivative $D F(p)$ has maximal rank (rank $m$). The IFT implies that near any regular point of the zero level set, the set looks like the graph of a function (possibly after permuting coordinates). [/definition]

## Examples

[example:SimpleCurve] For the circle $F(x, y) = x^2 + y^2 - 1 = 0$ at $(0, 1)$, we check invertibility w.r.t $y$. $D_y F(x, y) = \frac{\partial F}{\partial y} = 2y$. At $(0, 1)$, $2(1) \neq 0$, so we can solve for $y(x)$ locally. At $(1, 0)$, $D_y F = 0$, so the theorem gives no guarantee (and indeed, $y$ is not a function of $x$ there). [/example]

[example:NonLinearSystem] **Complex Case (System):** Consider the system solving for $(u, v)$ in terms of $(x, y)$: \begin{align*} F_1(x, y, u, v) &= u^3 + x v - y = 0 \\ F_2(x, y, u, v) &= v^3 + y u - x = 0 \end{align*} Let $F = (F_1, F_2)$. We check solvability near a solution, say $p_0 = (x, y, u, v) = (1, 1, 1, 0)$. The Jacobian block corresponding to the last two variables $(u, v)$ is: \begin{align*} D_{(u,v)} F = \begin{bmatrix} \frac{\partial F_1}{\partial u} & \frac{\partial F_1}{\partial v} \\ \frac{\partial F_2}{\partial u} & \frac{\partial F_2}{\partial v} \end{bmatrix} = \begin{bmatrix} 3u^2 & x \\ y & 3v^2 \end{bmatrix} \end{align*} Evaluating at $p_0=(1,1,1,0)$: \begin{align*} \det \begin{bmatrix} 3(1)^2 & 1 \\ 1 & 3(0)^2 \end{bmatrix} = \det \begin{bmatrix} 3 & 1 \\ 1 & 0 \end{bmatrix} = -1 \neq 0 \end{align*} Since the determinant is non-zero, there exist unique functions $u(x, y)$ and $v(x, y)$ defined in a neighborhood of $(1, 1)$ satisfying the system. [/example]

## Key Results

[theorem:ImplicitDifferentiation] If $g: V \to W$ is the implicit function defined by $F(x, g(x)) = 0$, its derivative can be computed without solving for $g$ explicitly. Differentiating the identity $F(x, g(x)) = 0$ with respect to $x$ via the Chain Rule yields: \begin{align*} D_x F(x, g(x)) + D_y F(x, g(x)) \circ D g(x) = 0 \end{align*} Rearranging terms (since $D_y F$ is invertible): \begin{align*} D g(x) = - [D_y F(x, g(x))]^{-1} \circ D_x F(x, g(x)) \end{align*} [/theorem]

[proof] This follows directly from applying the Chain Rule to the zero map $x \mapsto F(x, g(x))$ and isolating $Dg(x)$. [/proof]

[theorem:Equivalence] **Equivalence to [Inverse Function Theorem](/page/Inverse%20Function%20Theorem):** The Implicit Function Theorem and the Inverse Function Theorem are mathematically equivalent; each can be proven as a corollary of the other. * **IFT $\implies$ InvFT:** Given $y = f(x)$, define $F(x, y) = y - f(x) = 0$. Solving for $x$ in terms of $y$ using IFT yields the inverse map $x = f^{-1}(y)$. * **InvFT $\implies$ IFT:** Given $F(x, y) = 0$, construct a map $H(x, y) = (x, F(x, y))$. If $D_y F$ is invertible, $H$ is a local diffeomorphism. Its inverse $H^{-1}(u, v)$ has the form $(u, g(u, v))$. Setting $v=0$ recovers the implicit function $g(x)$. [/theorem]

## Problems

[problem] Let the surface $S$ in $\mathbb{R}^3$ be defined by the equation $F(x, y, z) = xz + \sin(yz) + x^2 - 1 = 0$. 1. Prove that we can solve for $z$ as a function of $(x, y)$ near the point $P = (1, \pi, 0)$. 2. Compute the partial derivative $\frac{\partial z}{\partial x}$ at $(1, \pi)$. [/problem]

[solution] **1. Existence:** Check the partial derivative with respect to $z$, denoted $F_z$: \begin{align*} F_z(x, y, z) = x + y\cos(yz) \end{align*} Evaluate at $P = (1, \pi, 0)$: \begin{align*} F_z(1, \pi, 0) = 1 + \pi\cos(0) = 1 + \pi \end{align*} Since $1 + \pi \neq 0$, the IFT guarantees $z = g(x, y)$ exists locally. **2. Derivative:** Using the implicit [differentiation](/page/Derivative) formula: \begin{align*} \frac{\partial z}{\partial x} = -\frac{F_x}{F_z} \end{align*} Calculate $F_x(x, y, z) = z + 2x$. At $P$, $F_x = 0 + 2(1) = 2$. \begin{align*} \frac{\partial z}{\partial x}(1, \pi) = -\frac{2}{1 + \pi} \end{align*} [/solution]

[problem] Consider the system of equations defining $(u, v)$ as functions of $(x, y)$: \begin{align*} xu + yvu^2 &= 2 \\ xu^3 + y^2v^4 &= 2 \end{align*} Find the Jacobian matrix of the implicit mapping $\phi: (x, y) \mapsto (u, v)$ at the point $(x, y, u, v) = (1, 1, 1, 1)$. [/problem]

[solution] Let $F(x, y, u, v) = \begin{pmatrix} xu + yvu^2 - 2 \\ xu^3 + y^2v^4 - 2 \end{pmatrix}$. First, calculate the matrices $D_x F$ (derivatives w.r.t $x, y$) and $D_u F$ (derivatives w.r.t $u, v$) at the point $(1,1,1,1)$. \begin{align*} D_x F &= \begin{bmatrix} u & vu^2 \\ u^3 & 2yv^4 \end{bmatrix} \bigg|_{(1,1,1,1)} = \begin{bmatrix} 1 & 1 \\ 1 & 2 \end{bmatrix} \\ D_u F &= \begin{bmatrix} x + 2yvu & yu^2 \\ 3xu^2 & 4y^2v^3 \end{bmatrix} \bigg|_{(1,1,1,1)} = \begin{bmatrix} 3 & 1 \\ 3 & 4 \end{bmatrix} \end{align*} We need $D \phi = - [D_u F]^{-1} [D_x F]$. First, invert $D_u F$: \begin{align*} \det(D_u F) &= 12 - 3 = 9 \\ [D_u F]^{-1} &= \frac{1}{9} \begin{bmatrix} 4 & -1 \\ -3 & 3 \end{bmatrix} \end{align*} Now multiply: \begin{align*} D \phi &= -\frac{1}{9} \begin{bmatrix} 4 & -1 \\ -3 & 3 \end{bmatrix} \begin{bmatrix} 1 & 1 \\ 1 & 2 \end{bmatrix} \\ &= -\frac{1}{9} \begin{bmatrix} 3 & 2 \\ 0 & 3 \end{bmatrix} = \begin{bmatrix} -1/3 & -2/9 \\ 0 & -1/3 \end{bmatrix} \end{align*} [/solution]

## Applications

1. **Level Sets and Manifolds:** The IFT is the primary tool used to verify that a level set $F^{-1}(c)$ forms a smooth manifold of dimension $n$. 2. **Econometrics:** In comparative statics, it allows economists to determine how optimal choices (endogenous variables) change in response to changes in parameters (exogenous variables), even without an explicit solution. 3. **[Numerical Analysis](/page/Numerical%20Analysis):** It underpins "Continuation Methods" (or Homotopy methods) for tracing solution curves of non-linear systems.