The Choice of a Change of Variables and Typical Worked Examples
The Principles for Choosing a Change of Variables
Choosing a suitable change of variables is the key skill in computing a multiple integral. In general you consider the following two dimensions.
Dimension one: the shape of the region of integration
Shape of the region
Recommended transformation
The region after the transformation
A disk, sector or annulus
polar coordinates
a rectangle (ρ-θ)
An ellipse or elliptical sector
generalised polar coordinates
a rectangle (ρ-θ)
A solid of revolution (a cylinder, a cone and so on)
cylindrical coordinates
a box (ρ-θ-z)
A sphere, spherical sector or spherical shell
spherical coordinates
a box (r-ϕ-θ)
An ellipsoid
generalised spherical coordinates
a box
A parallelogram
a linear transformation
a rectangle
A region bounded by parabolas
the hyperbolic transformation
a rectangle
Dimension two: the form of the integrand
Feature of the integrand
Recommended transformation
Contains x2+y2 or x2+y2
polar / cylindrical coordinates
Contains x2+y2+z2 or x2+y2+z2
spherical coordinates
Contains a2x2+b2y2
generalised polar coordinates
Contains a combination of x+y and x−y
the linear transformation u=x+y,v=x−y
Contains xy or arctanxy
polar coordinates (θ=arctanxy)
The antiderivative is not elementary (such as e−x2 or xsinx)
change the order of integration or use a change of variables
A Rule of Thumb
First look at the shape of the region, and then at the form of the integrand. When both point to the same transformation, that is usually the best choice. If the two conflict, give priority to whether the integrand can be integrated at all — an integrand with no elementary antiderivative has to be handled by a transformation or by changing the order.
The General Steps of a Change of Variables
Analyse the region of integration: sketch the region and identify the types of the boundary curves
Analyse the integrand: check whether it contains special forms such as x2+y2, and whether there is any non-integrable part
Choose the transformation: pick a suitable coordinate transformation according to the region and the integrand
Compute the Jacobian: find the Jacobian determinant J of the transformation
Transform the limits: transform the original region into the region of the new coordinate system and determine the new limits
Transform the integrand: express x,y,z in terms of the new variables and substitute into the integrand
Compute the integral: evaluate the iterated integral in the new coordinate system
The Common Pitfalls
Forgetting the Jacobian: you must multiply by ∣J∣ after the transformation — this is the most common mistake
Computing the Jacobian incorrectly: in particular the r2sinϕ of spherical coordinates, where it is easy to leave out r2 or sinϕ
Transforming the limits incorrectly: after the transformation the limits have to be determined afresh; the original limits cannot simply be reused
The transformation is not one-to-one: at special points such as the origin or on a coordinate axis the Jacobian may be zero, but as long as this is a finite set of points or curves it does not affect the result of the integral
The wrong angular ranges: the ranges of θ and ϕ in polar and spherical coordinates have to be determined carefully from the region
Worked Examples
Example 1: Polar Coordinates Together with Symmetry
Compute ∬D(x2+y2)dxdy, where D is the region between the circles x2+y2=2x and x2+y2=4x.
Both circles have their centres on the x-axis, and in polar coordinates they become ρ=2cosθ and ρ=4cosθ. The region is symmetric about the x-axis.
Steps:
The region in polar coordinates: 2cosθ≤ρ≤4cosθ,−2π≤θ≤2π
The integrand: x2+y2=ρ2
Using the symmetry about the x-axis:
∬D(x2+y2)dxdy=2∫0π/2dθ∫2cosθ4cosθρ2⋅ρdρ=2∫0π/2[4ρ4]2cosθ4cosθdθ=2∫0π/241(256cos4θ−16cos4θ)dθ=2⋅60∫0π/2cos4θdθ=120⋅43⋅21⋅2π=120⋅163π=245π
Answer: 245π
Example 2: Cylindrical Coordinates with the Two-Then-One Method
Compute ∭Ωz2dv, where Ω is the part of the sphere x2+y2+z2≤R2 lying inside the cone x2+y2≤z2.
cylindrical coordinatesa spherical sectortwo then onea conical surface
Approach:
The conical surface z=x2+y2 divides the sphere into two parts; take the part inside the cone (z≥x2+y2). Use the two-then-one method.
Steps:
The curve of intersection: substituting z=x2+y2 into the sphere gives 2z2=R2, that is, z=2R
When 0≤z≤2R, the cross-section Dz:x2+y2≤z2 has area πz2
When 2R≤z≤R, the cross-section Dz:x2+y2≤R2−z2 has area π(R2−z2)
∭Ωz2dv=∫0R/2z2⋅πz2dz+∫R/2Rz2⋅π(R2−z2)dz
Computing the two terms: π∫0R/2z4dz=202πR5, and π∫R/2R(R2z2−z4)dz=π[3R2z3−5z5]R/2R
Simplifying the whole expression gives 30πR5(4−2)
Answer: 30πR5(4−2)
Example 3: A Change of Variables for a Non-Integrable Function
Compute ∬Dex+yydxdy, where D is bounded by x+y=1,x=0,y=0.
a linear transformationa non-integrable functiona triangular regiona change of variables
Approach:
The function ex+yy has no elementary antiderivative, so a change of variables is needed. Put u=x+y,v=y, and then x+yy=uv.
Steps:
The transformation: u=x+y,v=y, with the inverse transformation x=u−v,y=v
The Jacobian: J=∂(u,v)∂(x,y)=10−11=1
The transformation of the region: x≥0⇒u−v≥0⇒v≤u; y≥0⇒v≥0; and x+y≤1⇒u≤1
Compute ∭Ω(x2+y2)dv, where Ω is the region enclosed by the surface obtained by rotating the curve y2=2z,x=0 once around the z-axis, together with the planes z=2,z=8.
cylindrical coordinatesa paraboloid of revolutiontwo then onea solid of revolution
Approach:
The surface of revolution is x2+y2=2z (a paraboloid of revolution). Use cylindrical coordinates and the two-then-one method.
Steps:
The surface of revolution: x2+y2=2z, which in cylindrical coordinates is ρ2=2z
When 2≤z≤8, the cross-section Dz:x2+y2≤2z, that is, ρ≤2z