Why Do You Use the Absolute Value of the Jacobian Determinant?

When you change variables in a double or triple integral, the Jacobian determinant may be positive or negative. The sign depends on whether the coordinate transformation preserves or reverses orientation. But ordinary area and volume are never negative, so the change-of-variables formula uses the absolute value of the determinant.

For a transformation $(x,y)=T(u,v)$, the standard formula is

$$\iint_{T(D)} f(x,y)\,dx\,dy = \iint_D f(T(u,v))\left|\det DT(u,v)\right|\,du\,dv.$$

Here, $DT$ is the derivative matrix of the transformation, and $|\det DT|$ is the local area-scaling factor.

What the determinant tells you

Near a point, a differentiable transformation behaves approximately like its derivative matrix, a linear map. A linear map sends a tiny square or cube to a parallelogram or parallelepiped. The magnitude of its determinant gives the factor by which area or volume changes.

The sign carries different information: it records orientation. In two dimensions, a negative determinant indicates that the map flips orientation, as a reflection does. In three dimensions, the sign likewise distinguishes orientation-preserving from orientation-reversing maps. Neither sign makes the size of the transformed region negative.

For example, a small region of area $dA$ is transformed to a region with area approximately

$$dA_{xy}=\left|\det DT(u,v)\right|\,du\,dv.$$

That is why the absolute value appears in an ordinary integral. A signed determinant is useful in some settings involving orientation, but it is not the area or volume element used in the usual change-of-variables formula.

A worked example with a negative determinant

Consider the transformation

$$x=u,\qquad y=-2v,$$

from the unit square $D=[0,1]\times[0,1]$. Its derivative matrix and determinant are

$$DT= \begin{pmatrix} 1&0\\ 0&-2 \end{pmatrix}, \qquad \det DT=-2.$$

The map reflects the square across the $x$-axis and stretches it vertically by a factor of $2$. Its image is the rectangle $R=[0,1]\times[-2,0]$, whose area is $2$. Accordingly, the area element transforms by $|{-2}|=2$.

To calculate the area of $R$ by integration, set $f(x,y)=1$:

$$\begin{aligned} \iint_R 1\,dx\,dy &=\iint_D 1\cdot|\det DT|\,du\,dv\\ &=\int_0^1\int_0^1 2\,dv\,du\\ &=2. \end{aligned}$$

If you used $-2$ instead, you would get $-2$, which is not the rectangle’s area. The negative sign correctly describes the reflection, but area ignores orientation.

Which determinant goes in the formula?

Use the determinant of the map from the variables you are integrating over to the variables in the original integral. If you write $x=x(u,v)$ and $y=y(u,v)$, use

$$\left|\frac{\partial(x,y)}{\partial(u,v)}\right|.$$

If instead you solve for $u$ and $v$ as functions of $x$ and $y$, you have the inverse map. Its determinant is the reciprocal (where the map is invertible), so be consistent about which differential you are replacing. In particular,

$$dx\,dy=\left|\frac{\partial(x,y)}{\partial(u,v)}\right|du\,dv.$$

For polar coordinates, $x=r\cos\theta$ and $y=r\sin\theta$, the determinant is $r$. With the usual convention $r\geq0$, it is already nonnegative, so the area element is $dx\,dy=r\,dr\,d\theta$. If signed values of $r$ are allowed, the absolute value must be considered, and the coordinates may cover points more than once.

Conditions and common mistakes

The usual textbook change-of-variables theorem assumes that the transformation is continuously differentiable and one-to-one on the region (typically on an open set containing its interior), with suitable region and boundary conditions, and that its Jacobian determinant is nonzero in the relevant interior. If the map covers part of the image multiple times, integrating over the parameter region counts that part multiple times; you must restrict the region or account for the multiplicity.

Common mistakes include:

• Dropping the absolute value: This can produce a negative area or volume when the map reverses orientation.
• Using the inverse determinant by accident: Check which variables the original and transformed differentials contain.
• Forgetting the Jacobian factor: The integrand changes to $f(T(u,v))$, and the area or volume element also changes by $|\det DT|$.
• Ignoring multiple coverage: A locally valid Jacobian factor does not make a many-to-one parameterization one-to-one.

The practical rule is simple: compute the determinant for the chosen direction of substitution, then use its magnitude as the area or volume multiplier.