24/01/2024, 16:01 Double Integrals

Math 21a Handouts

Janet Chen

4 Double Integrals
First, let's remember the definite integral from single-variable calculus.

4.1 Review of the single-variable definite integral

In single-variable calculus, the definite integral is an operation involving two
pieces of data, a single-variable function f and a closed interval [a, b]. We're
b

used to visualizing the definite integral ∫ f (x) dx as a signed area, but the
a

definite integral is really defined to be a limit of Riemann sums. Here's a recap

b

of that definition; ∫ f (x) dx is the result of the following process:

a

1. Slice the interval [a, b] into pieces of equal width Δx.

2. For each slice, pick a point x in the slice 1 and calculate f (x)Δx.
3. Sum all of these values. We can write this sum as ∑ f (x)Δx . (A
one x per slice

sum of this form is called a single-variable Riemann sum.)

4. Take the limit of such Riemann sums as Δx → 0.

The resulting limit is the value of ∫ f (x) dx .

a

Of course, we never actually calculate definite integrals this way (we instead use
the Fundamental Theorem of Calculus), but this definition of the definite
integral is important because it enables us to recognize situations in which
definite integrals may be useful.

4.2 The double integral

For short, we often refer to a “single-variable definite integral” simply as a
single integral. Analagously, the double integral is an operation involving two
pieces of data, a 2-variable function f (x, y) and a 2-dimensional region R in R . 2

We write the double integral of f (x, y) over R using the symbol

∬ f (x, y) dA . Like the single integral, the double integral is defined to be a
R

limit of Riemann sums. More specifically, the double integral is the result of the
following process:

1. Slice the region R into small pieces. For example, we could slice R into
rectangles of width Δx and height Δy. (Of course, since R need not have
straight sides, not all of the pieces will be complete rectangles, as shown
below.)

https://people.math.harvard.edu/~jjchen/math21a/handouts/double-integrals.html 1/5
24/01/2024, 16:01 Double Integrals

2. For each slice, pick a point (x, y) in the slice and calculate f (x, y)ΔA,
where ΔA is the area of the slice. 2
3. Sum all of these values. We write this sum as ∑∑ f (x, y) ΔA .
one (x,y) per slice

4. Take the limit of this sum as the sizes of the pieces go to 0 (with
rectangular pieces, this means we take the limit as Δx → 0 and Δy → 0 ).

The resulting limit 3 is the value of ∬ f (x, y) dA .

R

There are a few important questions we'll be working to answer in the next few
classes:

1. Why is this idea useful?

2. Practically speaking, how do we calculate a double integral? (Although a
single-variable definite integral is defined as a limit of Riemann sums,
when we compute a definite integral, we never actually compute Riemann
sums and take a limit; the same is true for double integrals.)

4.3 Calculating double integrals

To understand how to calculate a double integral, we'll take a deep dive into an
example. Let R be the region in R above y = 0, below y = ln x, and to the left
2

of x = e ; here's a sketch of the region:

2

We'd like to understand how to evaluate ∬ f (x, y) dA for an arbitrary 2-

R

variable function f . The key is that this double integral can be rewritten in terms
of single integrals. To understand why this is, let's go back to the definition of
double integrals, which says ∬ f (x, y) dA is the result of a process:
R

We first slice the region R into small pieces; here, we'll use rectangles of width
Δx and height Δy: 4

https://people.math.harvard.edu/~jjchen/math21a/handouts/double-integrals.html 2/5
24/01/2024, 16:01 Double Integrals
On each slice, we pick a point (x, y) and calculate f (x, y)ΔA where ΔA is the
area of the slice. Since most of the pieces are rectangles of width Δx and height
Δy, they have ΔA = ΔxΔy; even for the pieces that are not complete

rectangles, it's reasonable to approximate ΔA by ΔxΔy, as this approximation

will become more accurate once we take the limit as Δx → 0 and Δy → 0.

We sum these values of f (x, y)ΔA, one value per slice, and ∬ f (x, y) dA is the
R

limit of these sums as Δx → 0 and Δy → 0 . That is,

∬ f (x, y) dA = lim lim ∑∑ f (x, y)ΔxΔy (4.1)

Δx→0 Δy→0
R
one (x,y) per slice

The key insight is that, in the double Riemann sum, we can add up the values
f (x, y)ΔxΔy in any order we want. So, for instance, we could add up the values

in each column first, get one result per column, and add up those results. So,
we can rewrite (4.1) as

⎛ ⎞
∬ f (x, y) dA = lim lim ∑ ∑ f (x, y)Δy Δx (4.2)
Δx→0 Δy→0 ⎝ ⎠
R
one x per column one y per slice in column of x

Let's think about the inner sum, where we've fixed a value of x and are
summing f (x, y)Δy for one y per slice in the column. Here's a picture of one
such column:

The sum ∑ f (x, y)Δy is a Riemann sum in y, so when we take

one y per slice in column of x

the limit of this Riemann sum as Δy → 0 , we'll get a definite integral of the form
?

∫ f (x, y) dy . What are the bounds of the integral? From the picture, we see
?

that we've sliced y-values from y = 0 to y = ln x (remember that we're looking

at a fixed x), so these are the bounds. That is,
ln x

lim ∑ f (x, y)Δy = ∫ f (x, y) dy.

Δy→0
one y per slice in column of x 0

Plugging this into (4.2) ,

ln x

∬ f (x, y) dA = lim ∑ (∫ f (x, y) dy) Δx (4.3)

Δx→0
R
one x per column 0

But notice that this is now the limit of a Riemann sum in x, so it can be rewritten
as an integral in x! What are the bounds of x? We're summing over one x per
column, and the columns go from x = 1 to x = e , so (4.3) becomes 2

2
e ln x

∬ f (x, y) dA = ∫ (∫ f (x, y) dy) dx (4.4)

R 1 0

Thus, we've rewritten ∬ f (x, y) dA in terms of single integrals. Let's see an

R

example of how to use this.

https://people.math.harvard.edu/~jjchen/math21a/handouts/double-integrals.html 3/5
24/01/2024, 16:01 Double Integrals

Example 4.1. Let's calculate ∬ (x

2
+ e ) dA
y
for the region R we've been
R

looking at.

▼ Solution

By (4.4) ,
2
e ln x

2 y 2 y
∬ (x + e ) dA = ∫ (∫ (x + e ) dy) dx
R 1 0

As we said already, in the inner integral, x should be thought of as

constant (because we're integrating over a column of the region, where x
is constant) and y is the variable:
2
e
y=ln x
2 y∣
= ∫ (x y + e ) dx
∣
y=0
1

2
e

2
= ∫ (x ln x + x − 1) dx
1

Using integration by parts to integrate x 2

ln x , this evaluates to:

6 4
5e e 11
2
= + − e +
9 2 18

2
e ln x

The expression ∫ (∫ f (x, y) dy) dx is called an iterated integral, and

1 0

2
e ln x

we can also write it without parentheses as ∫ ∫ f (x, y) dy dx .

1 0

If you look back at how we rewrote ∬ f (x, y) dA as an iterated integral, you'll

R

remember that it came from adding up the values f (x, y)ΔA in a particular
order: columns first. Of course, we could have instead chosen to add rows first.
Let's see what would happen if we had made that choice. Then, instead of (4.2) ,
we could have written

⎛ ⎞
∬ f (x, y) dA = lim lim ∑ ∑ f (x, y)Δx Δy (4.5)
Δx→0 Δy→0 ⎝ ⎠
R
one y per row one x per slice in row of y

Here's a picture of one row (with a fixed y):

From the picture, we can see that this row involves x-values from e (because y

the left end of the row is on y = ln x, which can be rewritten as x = e ) to e , so y 2

when we take the limit as Δx → 0 of the inner Riemann sum, we get the
2
e

definite integral ∫ f (x, y) dx . When we sum over all rows, y varies from 0 to 2,
y
e

so we end up rewriting

https://people.math.harvard.edu/~jjchen/math21a/handouts/double-integrals.html 4/5
24/01/2024, 16:01 Double Integrals
2
2 e

∬ f (x, y) dA = ∫ ∫ f (x, y) dx dy (4.6)

y
R 0 e

Example 4.2. Let's calculate ∬ (x

2
+ e ) dA
y
again using (4.6) .
R

▼ Solution

(4.6) says
2
2 e

2 y 2 y
∬ (x + e ) dA = ∫ ∫ (x + e ) dx dy
y
R 0 e

This time, in the inner integral, y is thought of as constant and x is the

variable:
2
2 x=e
3
x ∣
y
= ∫ ( + xe ∣ ) dy
3 ∣
0 x=e
y

2 6 3y
e e
2 y 2y
= ∫ ( + e e − − e ) dy
0
3 3

y=2
6 3y 2y
e e e ∣
2 y
= y + e e − − ∣
3 9 2 ∣
y=0

6 4
5e e 11
2
= + − e +
9 2 18

This is of course the same answer that we got in Example 4.1 , but notice
that this time we didn't need integration by parts. This illustrates one
reason it can be useful to have multiple ways to convert a double integral
to an iterated integral; sometimes, one iterated integral is easier to
evaluate than another.

https://people.math.harvard.edu/~jjchen/math21a/handouts/double-integrals.html 5/5