Calculus › Section O · Multiple integrals › Topic O.8
Change of variables and Jacobians
The extra r in polar integrals, the substitution factor g′(t), the area of a tilted parallelogram — one determinant explains them all. Change of variables trades an awkward region for an easy one, at the price of a Jacobian.
When regions fight back
Some double integrals are awkward because of the region, not the integrand: the \(e^{x^3}\) example in the double-integrals topic forced a change of order, and regions bounded by tilted lines or curved arcs can block every order. The fix is to redesign the coordinates: choose new variables \(u\) and \(v\) so the awkward region \(R\) in the \(xy\)-plane is the image of a simple region \(S\) in the \(uv\)-plane — a rectangle, usually. If \(R\)'s boundary curves are the level curves of the new variables, each boundary becomes a straight grid line \(u = c\) or \(v = c\), and \(S\) is trivial to describe. The one-variable version of this move is substitution, \(\int f(g(t))g'(t)\,dt = \int f(u)\,du\), where the factor \(g'(t)\) compensates for the stretching. In two variables that factor becomes an entire determinant.
The Jacobian of a transformation
Given a transformation \(x = x(u, v)\), \(y = y(u, v)\), the Jacobian is the determinant of its partial derivatives:
It measures how the transformation scales area near each point — a local magnification factor, not a constant. Two properties to bank. The reciprocal rule: \(\frac{\partial(u, v)}{\partial(x, y)} = 1 / \frac{\partial(x, y)}{\partial(u, v)}\), signs included, so you may compute whichever direction is easier. And the theorem's conditions: the transformation must be smooth and one-to-one on the region's interior, with a Jacobian that does not vanish there; quirks confined to the boundary are harmless.
Why a determinant measures area
A small \(du\)-by-\(dv\) rectangle in the \(uv\)-plane maps to a near-parallelogram in the \(xy\)-plane whose sides are the image vectors \((x_u, y_u)\,du\) and \((x_v, y_v)\,dv\). The cross product from the vectors topic gives that parallelogram's area as \(|x_u y_v - x_v y_u|\,du\,dv\) — the determinant's absolute value. Summing \(f\) over all the little pieces turns this into the change of variables formula:
The absolute value matters: a Jacobian of \(-2\) magnifies area by 2 while flipping orientation, and area is not negative. The formula is exact for linear transformations; for smooth ones the little parallelograms straighten out as \(du\) and \(dv\) shrink, which is what the limit in the integral secretly performs.
Worked example: polar, again
Compute the Jacobian of polar coordinates \(x = r\cos\theta\), \(y = r\sin\theta\). The ingredients are \(x_r = \cos\theta\), \(x_\theta = -r\sin\theta\), \(y_r = \sin\theta\), \(y_\theta = r\cos\theta\), so
The mysterious extra \(r\) from the polar double-integrals topic was a Jacobian all along. The same holds upstairs in three variables: \(\frac{\partial(x, y, z)}{\partial(r, \theta, z)} = r\) and \(\frac{\partial(x, y, z)}{\partial(\rho, \phi, \theta)} = -\rho^2\sin\phi\), so the cylindrical and spherical volume elements \(r\,dz\,dr\,d\theta\) and \(\rho^2\sin\phi\,d\rho\,d\phi\,d\theta\) are absolute Jacobians — which is why each factor shrinks near an axis exactly as the coordinate slices do.
Worked example: a parallelogram's area
Find the area of the region \(R\) bounded by the lines \(x + y = 1\), \(x + y = 3\), \(x - y = -1\) and \(x - y = 1\). The boundary lines are level curves of \(u = x + y\) and \(v = x - y\), so those are the right new variables. Invert the relationship: \(x = \frac{u + v}{2}\), \(y = \frac{u - v}{2}\), and the Jacobian is \((\tfrac{1}{2})(-\tfrac{1}{2}) - (\tfrac{1}{2})(\tfrac{1}{2}) = -\tfrac{1}{2}\), so the magnification is \(\tfrac{1}{2}\). The region \(S\) is the rectangle \(1 \le u \le 3\), \(-1 \le v \le 1\), area 4, and \(\iint_R 1\,dA = \iint_S \tfrac{1}{2}\,du\,dv = \tfrac{1}{2} \cdot 4 = 2\). Check by vectors: \(R\) is the parallelogram with vertices \((0, 1)\), \((1, 0)\), \((1, 2)\), \((2, 1)\), whose edge vectors \(\langle 1, -1\rangle\) and \(\langle 1, 1\rangle\) have cross product magnitude \(|1 \cdot 1 - (-1) \cdot 1| = 2\). Both routes give 2.
Worked example: an integral over the parallelogram
Evaluate \(\iint_R (x + y)^2\,dA\) over the same parallelogram. Under the same substitution the integrand becomes \(u^2\), so
Confirm the arithmetic by shifting to the centre: \(u = t + 2\) turns the interval into \(-1 \le t \le 1\), and \(\int_{-1}^1 (t + 2)^2\,dt = \int_{-1}^1 (t^2 + 4t + 4)\,dt = \tfrac{2}{3} + 0 + 8 = \tfrac{26}{3}\). The odd middle term drops out by symmetry — a one-line audit of the values at the endpoints 1 and 3.
Three dimensions, and one
The formula generalises directly: \(\iiint_R g\,dV = \iiint_S g\,\big|\frac{\partial(x, y, z)}{\partial(u, v, w)}\big|\,du\,dv\,dw\), with the \(3 \times 3\) determinant computed by the same rule. Cylindrical and spherical coordinates are the two standard examples — their volume elements \(r\,dz\,dr\,d\theta\) and \(\rho^2\sin\phi\,d\rho\,d\phi\,d\theta\) are the absolute Jacobians, which retroactively names every factor the previous topic charged for roundness. And in one variable, substitution \(\int f(g(t))g'(t)\,dt = \int f(u)\,du\) is the same theorem with one dimension removed: \(g'(t)\) is the Jacobian of \(u = g(t)\). One idea, every dimension.
Common mistakes to avoid
First, dropping the absolute value or the factor entirely: without \(|J|\) the answer is scaled point by point, not by a single constant, so no later correction rescues it. Second, computing the reciprocal: the formula wants \(\frac{\partial(x, y)}{\partial(u, v)}\); if your substitution is naturally written \(u = u(x, y)\), invert it first or use the reciprocal rule. Third, misidentifying \(S\): transform \(R\)'s boundary curves first — each boundary of \(R\) must land on a boundary of \(S\), and a rectangle is confirmed only when all four sides check out. Fourth, tolerating a zero Jacobian inside \(S\): the transformation folds there, points overlap and the integral double-counts, so keep the interior where \(J \ne 0\). Fifth, forgetting that the integrand transforms too: \(f(x, y)\) becomes \(f(x(u, v), y(u, v))\) before any integration happens.
Key takeaways
A transformation \(x = x(u, v)\), \(y = y(u, v)\) maps a simple region \(S\) onto the awkward region \(R\), and the Jacobian \(\frac{\partial(x, y)}{\partial(u, v)} = x_u y_v - x_v y_u\) is the local area magnification — the two-variable descendant of the substitution factor \(g'(t)\). The change of variables formula \(\iint_R f\,dA = \iint_S f(x(u, v), y(u, v))\,|J|\,du\,dv\) trades a hard region for an easy one; choose \(u\) and \(v\) as the functions whose level curves trace \(R\)'s boundary. Polar's \(r\), cylindrical's \(r\) and spherical's \(\rho^2\sin\phi\) are all Jacobians, and the reciprocal rule lets you compute whichever direction — forward or inverse — is convenient.
Flashcards
Flashcard
Map a simple region S onto the awkward region R
Test your understanding
Question 1 of 5
If x = uv and y = u + v, what is the Jacobian ∂(x,y)/∂(u,v)?
Question 2 of 5
What is the Jacobian ∂(x,y)/∂(r,θ) for the polar transformation x = r cos θ, y = r sin θ?
Question 3 of 5
The transformation x = (u+v)/2, y = (u−v)/2 maps the rectangle 1 ≤ u ≤ 3, −1 ≤ v ≤ 1 onto a region R. What is the area of R?
Question 4 of 5
In the change of variables formula, what multiplies f(x(u,v), y(u,v)) inside the integral over S?
Question 5 of 5
A transformation has ∂(x,y)/∂(u,v) = −2. What is ∂(u,v)/∂(x,y)?
Standards: Beyond AP
