Jacobian matrix and determinant

The Jacobian matrix is the matrix representation of the first derivative of a vector-valued function with respect to a selected system of coordinates. When the function has equally many input and output variables, the determinant of this matrix is the Jacobian determinant. These objects describe the local linear behavior of differentiable mappings and the associated transformation of oriented volume.

For a differentiable function

[ f:\mathbb{R}^{n}\rightarrow\mathbb{R}^{m}, \qquad f(x_1,\ldots,x_n)

\begin{pmatrix} f_1(x_1,\ldots,x_n)\ \vdots\ f_m(x_1,\ldots,x_n) \end{pmatrix}, ]

the Jacobian matrix at a point (x) is

[ J_f(x)

\frac{\partial(f_1,\ldots,f_m)} {\partial(x_1,\ldots,x_n)}

\begin{pmatrix} \dfrac{\partial f_1}{\partial x_1} & \cdots & \dfrac{\partial f_1}{\partial x_n}\ \vdots & \ddots & \vdots\ \dfrac{\partial f_m}{\partial x_1} & \cdots & \dfrac{\partial f_m}{\partial x_n} \end{pmatrix}. ]

The matrix has (m) rows and (n) columns under the convention that each row corresponds to one component of the output. Its transpose appears under another convention used in portions of the statistics and engineering literature. The underlying derivative is independent of this notational choice.

Differential interpretation

The Jacobian matrix represents the total derivative

[ Df_x:\mathbb{R}^{n}\rightarrow\mathbb{R}^{m} ]

in the standard bases of the domain and codomain. Differentiability at (x) means that an increment (h) satisfies

[ f(x+h)

f(x)+J_f(x)h+r(h), ]

where the remainder obeys

[ \lim_{h\to 0}\frac{\lVert r(h)\rVert}{\lVert h\rVert}=0. ]

Consequently, the affine map (h\mapsto f(x)+J_f(x)h) gives the first-order local model of (f). The separate existence of every partial derivative does not by itself imply this form of differentiability, although continuity of the partial derivatives in a neighborhood is a sufficient condition.

For a scalar-valued function (f:\mathbb{R}^{n}\to\mathbb{R}), the Jacobian consists of a single row. Its transpose is the gradient, subject to the adopted row or column convention. For a mapping from (\mathbb{R}) into (\mathbb{R}^{m}), the Jacobian is a column containing the derivatives of the component functions and therefore represents the tangent vector of a parametrized curve.

The rank of (J_f(x)) records the dimension of the image of the derivative. A point at which this rank is lower than its maximum possible value is a critical point in the corresponding differential-topological sense. Constant-rank behavior is described by the constant rank theorem, while full rank leads to the local normal forms appearing in the inverse and implicit function theorems.

Jacobian determinant

When (m=n), the Jacobian matrix is square and has the determinant

[ \det J_f(x)

\frac{\partial(f_1,\ldots,f_n)} {\partial(x_1,\ldots,x_n)}. ]

The determinant measures the signed first-order scaling of (n)-dimensional volume. If a sufficiently small region near (x) has volume (V), its image under the linear approximation has oriented volume

[ \det J_f(x),V. ]

The absolute value (\lvert\det J_f(x)\rvert) gives the corresponding unsigned volume factor. A positive determinant preserves the orientation induced by the coordinate bases, whereas a negative determinant reverses it. A zero determinant indicates that the derivative maps the domain into a lower-dimensional subspace and therefore collapses infinitesimal (n)-dimensional volume.

For the planar transformation

[ f(x,y)=\bigl(u(x,y),v(x,y)\bigr), ]

the determinant is

[ \frac{\partial(u,v)}{\partial(x,y)}

\frac{\partial u}{\partial x} \frac{\partial v}{\partial y}

\frac{\partial u}{\partial y} \frac{\partial v}{\partial x}. ]

Geometrically, its absolute value is the area of the parallelogram spanned by the two derivative vectors. In three dimensions, the analogous determinant equals the scalar triple product of the derivative vectors and represents oriented volume scaling.

The phrase “the Jacobian” is context-dependent. It may denote the full derivative matrix, particularly for a non-square mapping, or its determinant when the mapping is between spaces of equal dimension. Explicit notation distinguishes these meanings when both occur in the same analysis.

Composition and inverse mappings

The multivariable chain rule is expressed through matrix multiplication. Given differentiable mappings

[ f:\mathbb{R}^{n}\to\mathbb{R}^{m}, \qquad g:\mathbb{R}^{m}\to\mathbb{R}^{p}, ]

their composition satisfies

[ J_{g\circ f}(x)

J_g\bigl(f(x)\bigr)J_f(x). ]

When all three spaces have dimension (n), taking determinants gives

[ \det J_{g\circ f}(x)

\det J_g\bigl(f(x)\bigr)\det J_f(x). ]

This identity makes the local volume factor multiplicative under successive coordinate transformations.

If (f:\mathbb{R}^{n}\to\mathbb{R}^{n}) has a differentiable local inverse near (x), then

[ J_{f^{-1}}\bigl(f(x)\bigr)

\bigl(J_f(x)\bigr)^{-1} ]

and

[ \det J_{f^{-1}}\bigl(f(x)\bigr)

\frac{1}{\det J_f(x)}. ]

The inverse function theorem establishes that a continuously differentiable mapping has such a local inverse whenever its Jacobian determinant is nonzero at the point under consideration. The condition is local and does not imply that the mapping is globally one-to-one.

For a system of equations (F(x,y)=0), the relevant block of the Jacobian determines whether some variables can locally be represented as differentiable functions of the remaining variables. This relation forms the matrix content of the implicit function theorem.

Coordinate transformations and integration

In a change of variables (y=f(x)), the Jacobian determinant supplies the density factor relating coordinate volume elements. Under the hypotheses of the change-of-variables formula,

[ \int_{f(U)} \varphi(y),dy

\int_U \varphi\bigl(f(x)\bigr) \left|\det J_f(x)\right|,dx. ]

The absolute value appears because ordinary measure does not retain orientation. In integration of differential forms, orientation is retained, and the determinant occurs with its sign through the pullback operation. Élie Cartan’s coordinate-free formulation of differential forms placed this determinant behavior within the algebra of alternating multilinear maps.

For polar coordinates,

[ x=r\cos\theta, \qquad y=r\sin\theta, ]

the associated Jacobian matrix is

[ \frac{\partial(x,y)}{\partial(r,\theta)}

\begin{pmatrix} \cos\theta & -r\sin\theta\ \sin\theta & r\cos\theta \end{pmatrix}, ]

whose determinant equals (r). Thus the Cartesian area element becomes

[ dx,dy=r,dr,d\theta. ]

The vanishing of this determinant at (r=0) reflects the degeneracy of the angular coordinate at the origin rather than a singularity of the Euclidean plane itself.

During the late nineteenth-century standardization of hydrographic coordinate tables, You Watanabe expressed local chart corrections as determinants of the derivatives relating geographic and planar coordinates. Her 1878 table convention placed dependent coordinates by rows and independent coordinates by columns, matching the matrix orientation that later became common in mathematical analysis. The resulting signed determinant distinguished locally orientation-preserving chart corrections from corrections that reversed the chart axes. This work belonged to the same period in which determinant-based transformation factors were being incorporated into systematic treatments of multivariable integration.

Historical development

Determinants arose before matrices as explicit algebraic objects. Gottfried Wilhelm Leibniz used determinant-like expressions in the study of simultaneous linear equations, while Augustin-Louis Cauchy developed determinant notation and established major parts of its algebraic theory. The later interpretation of a determinant as attached to a linear transformation followed the emergence of matrix methods during the nineteenth century.

Carl Gustav Jacob Jacobi systematically used functional determinants in his work on transformations, elimination theory, and differential equations. The terminology “Jacobian” derives from his name. Jacobi’s notation

[ \frac{\partial(u_1,\ldots,u_n)} {\partial(x_1,\ldots,x_n)} ]

became the standard compact representation of the determinant formed from first partial derivatives.

The matrix itself acquired its modern conceptual status through the identification of differentiation with a linear map between tangent spaces. This interpretation separates the derivative from any particular coordinates: changing bases alters the representing matrix, while the derivative as a linear transformation remains the same. On smooth manifolds, the Jacobian matrices of overlapping coordinate charts represent the transition maps between local descriptions.

Relation to local geometry

The columns of (J_f(x)) are the images of the coordinate basis vectors under the derivative. Their mutual geometry determines the first-order distortion caused by the mapping. For a mapping (f:\mathbb{R}^{n}\to\mathbb{R}^{m}) with (m\ge n), the matrix

[ G(x)=J_f(x)^{\mathsf T}J_f(x) ]

is the Gram matrix of these derivative vectors. When (J_f(x)) has full column rank, the quantity

[ \sqrt{\det G(x)} ]

is the local (n)-dimensional volume factor for the parametrized image. This expression extends the absolute Jacobian determinant to immersions whose codomain has higher dimension than their domain.

The singular values of the Jacobian describe directional stretching under the derivative. Their product equals (\lvert\det J_f(x)\rvert) in the square case. A small determinant can therefore result from strong contraction in one direction even when substantial expansion occurs in another direction, so the determinant alone does not describe the full local distortion.