Quant Academy Math · Probability · Quant Finance
Phase 3 - Lesson 3.3

Multiple Integrals and Change of Variables

Integrating over regions of the plane and space, iterating with Fubini, and rescaling with the Jacobian determinant.

⏱ 55 min● Intermediate🔗 Prereqs: 3.2
↖ Phase 3 hub
Builds on: Partial derivatives and the Jacobian from 3.2.
Leads to: Change of variables underlies joint densities and the Gaussian normalization in probability (Phase 7).

Learning Objectives

Click a status chip to cycle: Not started → In progress → Studied → Practiced → Needs review → Mastered.

Key Vocabulary

Double integral
The limit \(\iint_R f\,dA\) of Riemann sums of \(f\) over a plane region \(R\).
Iterated integral
Evaluation \(\int\!\!\int f\,dy\,dx\) one variable at a time, inner limits possibly depending on the outer.
Fubini’s theorem
For integrable \(f\), the double integral equals either iterated integral, so the order may be swapped.
Jacobian determinant
The factor \(|\det J|\) by which a change of variables rescales area/volume elements.
Region of integration
The domain \(R\); its description (as x- or y-simple) sets the inner and outer limits.
Polar coordinates
The map \(x=r\cos\theta,\ y=r\sin\theta\) with area element \(dA=r\,dr\,d\theta\).

Intuition & Motivation

Intuition
A double integral adds up a quantity spread over a two-dimensional region - a volume under a surface, a total mass, or a probability. You compute it by slicing: integrate along one direction to collapse the region to a single variable, then integrate that. The change-of-variables formula is the multivariable substitution rule; the single-variable factor \(du/dx\) is replaced by the Jacobian determinant, which measures how much a coordinate map stretches area.

Double integrals and Fubini

Over a rectangle \(R=[a,b]\times[c,d]\), Fubini’s theorem lets us iterate in either order:

\[\iint_R f\,dA=\int_a^b\!\!\int_c^d f(x,y)\,dy\,dx=\int_c^d\!\!\int_a^b f(x,y)\,dx\,dy.\] (3.4)

Over a non-rectangular region we describe it as y-simple (\(g_1(x)\le y\le g_2(x)\)) or x-simple and let the inner limits depend on the outer variable. Swapping order then requires re-reading the region, not just flipping symbols.

Worked Example - Swapping the order of integration
1
Consider \(\int_0^1\!\!\int_x^1 e^{y^2}\,dy\,dx\). The inner antiderivative of \(e^{y^2}\) has no elementary form.
2
Re-describe the triangular region \(\{0\le x\le 1,\ x\le y\le 1\}\) as \(\{0\le y\le 1,\ 0\le x\le y\}\).
3
Swap: \(\int_0^1\!\!\int_0^y e^{y^2}\,dx\,dy=\int_0^1 y\,e^{y^2}\,dy\) since the inner integral is just \(e^{y^2}\cdot y\).
4
Substitute \(u=y^2\): \(\tfrac12\int_0^1 e^u\,du=\tfrac12(e-1)\). Choosing the right order made an impossible integral trivial.

Change of variables and the Jacobian

Theorem - Change of variables
If \(T:(u,v)\mapsto(x,y)\) is a smooth injective map with Jacobian determinant \(\det J_T=\partial(x,y)/\partial(u,v)\), then \(\iint_R f\,dx\,dy=\iint_{T^{-1}(R)} f(T(u,v))\,|\det J_T|\,du\,dv\).

The absolute value of the determinant is the local area-scaling factor: a coordinate map that doubles areas contributes a factor of 2. For polar coordinates \(\det J=r\), giving the familiar \(dA=r\,dr\,d\theta\).

The Gaussian integral

The most important integral in quantitative finance normalizes the normal distribution. It has no elementary antiderivative in one variable, yet a polar trick evaluates it exactly:

Worked Example - Evaluating \(I=\int_{-\infty}^{\infty} e^{-x^2}\,dx\)
1
Square it and write as a double integral: \(I^2=\int_{-\infty}^{\infty}\!\!\int_{-\infty}^{\infty} e^{-(x^2+y^2)}\,dx\,dy\).
2
Switch to polar: \(x^2+y^2=r^2\) and \(dA=r\,dr\,d\theta\), so \(I^2=\int_0^{2\pi}\!\!\int_0^{\infty} e^{-r^2}\,r\,dr\,d\theta\).
3
Inner: \(\int_0^\infty e^{-r^2}r\,dr=\tfrac12\) (substitute \(s=r^2\)). Outer: \(\int_0^{2\pi}d\theta=2\pi\).
4
Hence \(I^2=\pi\) and \(I=\sqrt{\pi}\). Rescaling gives \(\int e^{-x^2/2}dx=\sqrt{2\pi}\), the normal constant.

This estimator throws random points in a square and counts the fraction inside a quarter-disc - a stochastic double integral. It converges at rate \(1/\sqrt{N}\), the same law that governs Monte Carlo option pricing (Phase 13).

Common Mistakes to Avoid
  • Swapping the order of integration without redrawing the region - the limits change, not just the differentials.
  • Forgetting the Jacobian factor \(r\) in polar coordinates; \(dA\neq dr\,d\theta\).
  • Using the signed determinant instead of \(|\det J|\); area scaling is always nonnegative.
  • Assuming every double integral factors into a product; that only happens when \(f(x,y)=g(x)h(y)\) on a rectangle.
Quant Practitioner Tips
  • When an inner integral is intractable, try swapping the order before anything fancier.
  • Radial symmetry (functions of \(x^2+y^2\)) almost always signals polar coordinates.
  • Memorize \(\int_{-\infty}^\infty e^{-x^2/2}dx=\sqrt{2\pi}\); it is the normalizing constant behind every Gaussian in the course.

Knowledge Check

Q1 Easy
Fubini’s theorem guarantees that for an integrable \(f\) on a rectangle:
The two orders of iterated integration give the same value
The integral factors as a product
f must be continuous
The region must be a disc
Q2 Easy
Under the polar change of variables, the area element \(dx\,dy\) becomes:
\(dr\,d\theta\)
\(r\,dr\,d\theta\)
\(r^2\,dr\,d\theta\)
\(\tfrac1r\,dr\,d\theta\)
Q3 Medium
The value of \(\int_{-\infty}^\infty e^{-x^2}\,dx\) is:
\(\pi\)
\(\sqrt{\pi}\)
\(2\pi\)
\(1\)

Practical Exercise

Evaluate \(\iint_R (x+y)\,dA\) where \(R\) is the triangle with vertices \((0,0),\ (1,0),\ (0,1)\). Set up the iterated integral, then compute.

▶ Show full solution

Describe \(R\) as \(0\le x\le 1,\ 0\le y\le 1-x\):

\[\int_0^1\!\!\int_0^{1-x}(x+y)\,dy\,dx=\int_0^1\Big[xy+\tfrac{y^2}{2}\Big]_0^{1-x}dx=\int_0^1\Big(x(1-x)+\tfrac{(1-x)^2}{2}\Big)dx.\]

Expand the integrand: \(x-x^2+\tfrac12(1-2x+x^2)=\tfrac12-\tfrac12x^2\). Integrate:

\[\int_0^1\Big(\tfrac12-\tfrac12 x^2\Big)dx=\tfrac12-\tfrac16=\tfrac13.\]

So the integral equals \(\tfrac13\). By symmetry in \(x\) and \(y\) you could also split into two equal pieces as a check.

After the reveal, answer for yourself: Re-do it with the order dx dy; the symmetry of the region should give the identical answer 1/3.

Lesson Summary

Double integrals are evaluated by iteration; Fubini lets us choose or swap the order (after re-reading the region). The change-of-variables formula rescales by the Jacobian determinant \(|\det J|\), and the polar trick delivers the Gaussian integral \(\int e^{-x^2}dx=\sqrt\pi\) that normalizes every normal distribution.

Formula Sheet Additions

Change of variables
\[\iint_R f\,dx\,dy=\iint_{T^{-1}(R)} (f\circ T)\,|\det J_T|\,du\,dv\]
Jacobian determinant = local area-scaling factor.

Retrieval Practice

Close the lesson and answer from memory before checking. This is deliberate, effortful recall - the single highest-yield study action.

▶ Show retrieval prompts & answers
Q: What replaces du/dx from single-variable substitution in a 2-D change of variables?
A: The absolute Jacobian determinant |det J| of the coordinate map, the local area-scaling factor.
Q: How is the Gaussian integral evaluated?
A: Square it into a double integral, switch to polar (dA=r dr dθ), integrate to get π, so the integral is sqrt(π).

Completion Checklist

Confidence / mastery rating
Personal notes

Source References

This lesson synthesizes and paraphrases concepts from the sources below. No copyrighted text, problem sets, or solutions are reproduced. Return to the originals for full depth.

← 3.2Next: 3.4 →