Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

An Overview of Feynman's Method for Calculating Improper Integrals

Abdlgader M. Inbaig

Email: [email protected]

Libyan Authority for Scientific Research, University of Tripoli, Tripoli-Libya.

ABSTRACT

The objective of this paper is to utilize the Feynman method to find the definite integral
of some functions that do not have an elementary antiderivatives. The main idea behind
"Feynman's method" is to solve a simple differential equation instead of computing an
integral. In most of the examples have been presented, we concentrated on uncommon
integrations to demonstrate the efficiency of this method.

Keywords: Feynman's method, differentiation under the integral sign rule ,definite
integral and improper integral.

1 Introduction

The Fundamental Theorem of Calculus cannot be used to evaluate the definite integrals
of many continuous functions because their antiderivatives lack basic formulas.
Numerical integration provides a practical approach to estimating the values of these
so-called "non-elementary integrals" as long as there is a procedure (or program code)
in your computer or calculator for numerical integration (Stewart, Clegg, & Watson,
2020).

In calculus, every elementary function-based single-variable function has a derivative.

However, not every elementary function-based single-variable function can be exactly
integrated over any arbitrary integral; for instance, the antiderivative cannot be
expressed in terms of elementary functions, in order to find the exact integral for any
interval. Elliptic integral, logarithmic integral, error function, Gaussian integral, Fresnel
integral sine integral (Dirichlet integral), Exponential integral (in terms of the
exponential integral) and logarithmic integral (in terms of the logarithmic integral) are
examples of functions with non-elementary antiderivatives (Zwillinger, D. and Jeffrey,
A., 2007) and (Jeffrey, A. and Dai, H.H., 2008)

1
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

There are so many integration techniques to cover in the literature, but you will find a
few of the well-known of them for addressing the problems with integrals. The most
common approaches of integration include; Integrating Functions Using Long Division, the
Power Rule for Integration, Integration by Substitution, Trigonometric Substitution,
Integration by Parts, Integration by Partial Fractions, Integrals of Inverse Functions,
Weierstrass Substitution, and Feynman's Integration Method. These are essential to know,
but they will not always be helpful when dealing with improper integrals.

Among all the approaches that have been mentioned above, we found that the most
practical approach to calculate the actual values of the so-called "non-elementary
integrals" is provided by "Feynman's method" (Kotikov, 1991). We refer the reader for
additional information on the integration methods listed above to some appropriate
resources (Ashlock, 2019) and (Shaxobiddinova et al, 2022).

In this paper, Feynman's method will be discussed in details along with some illustrated
examples involving improper integral formulae.

2 Basic Definitions

This section presents the basic definitions that have been used in this paper: we have
introduced Leibniz Rule (LR) and the Initial Value Problem (IVP) , and presented a brief
explanation of the ,

Definition 1. (Leibniz Rule) : Let 𝑎 and 𝑏 be constants. If a function 𝑓 and its partial
𝜕𝑓
derivative 𝜕𝑥 are continuous on a rectangle 𝑅 = {(𝑡, 𝑥): 𝑎 ≤ 𝑡 ≤ 𝑏, 𝑐 ≤ 𝑥 ≤ 𝑑}, then
𝑏
𝐼(𝑥) = ∫ 𝑓 (𝑡, 𝑥 ) 𝑑𝑡
𝑎

has the derivative

𝑏
𝑑𝐼(𝑥) 𝜕𝑓(𝑡, 𝑥)
=∫ 𝑑𝑡.
𝑑𝑥 𝑎 𝜕𝑥

2
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

The Leibniz integral rule is another name for the differentiation under the integral
sign rule. A formula for solving a definite integral with functions of the differential
variable acting as long as its limits exist. This rule makes it possible to compute a
differentiation of integral without integrating the equation.

Definition 2. (Initial Value Problem) : An Initial Value Problem, also known as an IVP,
𝑑𝑦
is a problem in which we need to solve a differential equation = 𝑓(𝑥, 𝑦) with a
𝑑𝑥

solution that satisfies 𝑦(𝑥0 ) = 𝑦0.

Feynman's method

Feynman’s method is often applied to integrals with a single variable. The technique
of Feynman itself is not easy to put into words. However, the idea is to artificially
introduce a second variable and then build a differential equation in terms of
derivatives with respect to the new variable. By solving this differential equation we
often can compute the original integral.

The following steps provide at least a sketch of what we mean when evaluating the
𝑏
definite integral of a function; 𝐼 = ∫𝑎 𝑓(𝑥)𝑑𝑥 using Feynman's method:

1. Consider the integral 𝐼 as a function of 𝑝; where p is an arbitrary parameter,

𝑏
𝐼(𝑝) = ∫ 𝑓(𝑥, 𝑝) 𝑑𝑥
𝑎

2. Compute the integral for some particular convenient value of 𝑝 = 𝑝0 , such that
𝐼(𝑝0 ) = 𝐼0 .

3. Differentiate the integral with respect to 𝑝.

4. Compute the definite integral with respect to 𝑥.

5. Integrate indefinitely with respect to 𝑝.

6. Use the fact that 𝐼(𝑝0 ) = 𝐼0 to compute the value of the constant of integration.

3
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

Following the previous steps, the problem will be changed from computing an integral to
solving a straightforward differential equation. We warm up with a relatively simple
examples.

3 Examples

Example 3.1.

tan−1 𝑎𝑥 −tan−1 𝑏𝑥
Consider the function 𝑓(𝑥) = , where 𝑎, 𝑏 are any constants such that
𝑥
𝑎
> 0. Let us compute the following Integral:
𝑏

∞ ∞
tan−1 𝑎𝑥 − tan−1 𝑏𝑥
𝐼=∫ 𝑓(𝑥) 𝑑𝑥 = ∫ 𝑑𝑥 (1)
0 0 𝑥

We first Introduce 𝐼 as a function of a parameter 𝑝:

∞
tan−1 𝑝𝑥 − tan−1 𝑏𝑥
𝐼(𝑝) = ∫ 𝑑𝑥 (2)
0 𝑥

Such that 𝐼(𝑎) = 𝐼, and 𝐼(𝑏) = 0.

Now, differentiate both sides of Eq. (2) with respect to 𝑝;

∞
𝑑 𝜕 tan−1 𝑝𝑥 − tan−1 𝑏𝑥 (3)
𝐼(𝑝) = ∫ ( ) 𝑑𝑥
𝑑𝑝 0 𝜕𝑝 𝑥

∞
𝑑 1 𝜕
𝐼(𝑝) = ∫ (tan−1 𝑝𝑥 )𝑑𝑥 (4)
𝑑𝑝 0 𝑥 𝜕𝑝

∞
𝑑 1 𝑥
𝐼(𝑝) = ∫ ( ) 𝑑𝑥 (5)
𝑑𝑝 2
0 𝑥 1 + (𝑝𝑥)

We can now calculate the integration by taking 𝑢 = 𝑝𝑥, and 𝑑𝑢 = 𝑝 𝑑𝑥.

𝑑 1 ∞ 1
𝐼(𝑝) = ∫ ( ) 𝑑𝑢 (6)
𝑑𝑝 𝑝 0 1 + 𝑢2

Then,

4
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

𝑑 1
𝐼(𝑝) = 𝑢 |∞ (7)
0
𝑑𝑝 𝑝

Let us now return 𝑢 back to its original value 𝑝𝑥;

𝑑 1
𝐼(𝑝) = 𝑝𝑥 |∞ (8)
0
𝑑𝑝 𝑝

By substituting, we have;

𝑑 𝜋
𝐼(𝑝) = ; 𝐼(𝑏) = 0 (9)
𝑑𝑝 2𝑝

By solving the initial value problem in Eq. (9), we get;

𝜋
𝐼(𝑝) = ln 𝑝 + 𝑐 (10)
2

We can find the value of 𝑐, by replacing 𝑝 with 𝑏 in Eq. (10);

𝜋
𝐼(𝑏) = 𝑙𝑛 𝑏 + 𝑐. (11)
2

Since 𝐼(𝑏) = 0, then

𝜋
0= 𝑙𝑛 𝑏 + 𝑐, (12)
2
So, we get

𝜋
𝑐=− 𝑙𝑛 𝑏 . (13)
2

By subtitling the value of 𝑐 in Eq. (10), we get:

𝜋 𝜋
𝐼 (𝑝 ) = 𝑙𝑛 𝑝 − 𝑏 (14)
2 2

𝜋 𝑝
𝐼(𝑝) = 𝑙𝑛 ( ) (15)
2 𝑏

Therefore,

5
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023
∞
tan−1 𝑎𝑥 − tan−1 𝑏𝑥 𝜋 𝑎
𝐼(𝑎) = 𝐼 = ∫ 𝑑𝑥 = 𝑙𝑛 ( ) (16)
0 𝑥 2 𝑏
𝑎
We should take in our consideration that > 0. If we put 𝑎 = 2 and 𝑏 = 8, we get
𝑏
tan−1 2𝑥−tan−1 8𝑥
𝑓(𝑥) = . Figure 1. shows the graph of the function 𝑓(𝑥) and its
𝑥
𝑥
antiderivative 𝐹(𝑥) = ∫0 𝑓(𝑡)𝑑𝑡. Notice that as 𝑥 → ∞, 𝐹(𝑥) → 𝐼. From Eq. (16) we
𝜋 2 𝜋 1
have 𝐼 = 𝑙𝑛 ( ) = 𝑙𝑛 ( ) ≅ −2.17758609.
2 8 2 4

Example 3.2.

𝑥 𝑎 −1
Let us take the function (𝑥) = ln (𝑥) ; provided that 𝑎 > −1. Since the process of finding

the definite Integral of this function is not easy, we will follow the Feynman's method:

tan−1 2𝑥−tan−1 8𝑥
Fig. 1. The graph of 𝑓(𝑥) = and its antiderivative within the domain
𝑥

0 ≤ 𝑥 ≤ 100.
6
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023
1 1
𝑥𝑎 − 1
𝐼 = ∫ 𝑓(𝑥)𝑑𝑥 = ∫ 𝑑𝑥; 𝑎 > −1 (17)
0 0 ln (𝑥)

Again, for computing the integration in Eq. (17), we Introduce 𝐼 as a function in

parameter 𝑝 as follows:

1
𝑥𝑝 − 1 (18)
𝐼(𝑝) = ∫ 𝑑𝑥;
0 𝑙𝑛 (𝑥)

Such that 𝐼(𝑎) = 𝐼, and 𝐼(0) = 0.

Now, differentiate both sides of Eq. (18) with respect to 𝑝

1
𝑑 1 𝜕 𝑝 (19)
𝐼 (𝑝 ) = ∫ (𝑥 − 1)𝑑𝑥
𝑑𝑝 0 𝑙𝑛 (𝑥) 𝜕𝑝
1
𝑑 1 (20)
𝐼(𝑝) = ∫ (𝑥 𝑝 . ( 𝑙𝑛 (𝑥) )𝑑𝑥
𝑑𝑝 0 𝑙𝑛 (𝑥)

1
𝑑 (21)
𝐼(𝑝) = ∫ 𝑥 𝑝 𝑑𝑥
𝑑𝑝 0

We can now calculate the integration by taking 𝑢 = 𝑝𝑥, and 𝑑𝑢 = 𝑝 𝑑𝑥.

𝑑 𝑥 𝑝+1 1 (22)
𝐼(𝑝) = |
𝑑𝑝 𝑝+1 0
Then,

𝑑 1
𝐼(𝑝) = ; 𝐼(0) = 0. (23)
𝑑𝑝 𝑝+1
Taking the integral of both sides, we get;

𝐼(𝑝) = 𝑙𝑛 |𝑝 + 1| + 𝑐 (24)

To find the value of c, we replace 𝑝 with 0 as follows:

𝐼(0) = 𝑙𝑛 |0 + 1| + 𝑐 . (25)

Since 𝐼(0) = 0, then

7
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

0 = 𝑙𝑛 |1| + 𝑐 , (26)

and we get

𝑐=0 (27)

By subtitling the value of 𝑐 in Equation 24, we have:

𝐼(𝑝) = 𝑙𝑛 |𝑝 + 1| (28)

Therefore, our Integration I became

𝐼 = 𝐼(𝑎) = 𝑙𝑛 |𝑎 + 1| (29)

𝑥−1 1 1 𝑥−1
When 𝑎 = 1 , we have (𝑥) = 𝑙𝑛 𝑥 . The definite integral 𝐼 = ∫0 𝑓(𝑥)𝑑𝑥 = ∫0 𝑑𝑥 =
𝑙𝑛 𝑥
𝑥−1
𝑙𝑛 (2) ≅ 0.69314718. The graph of the function 𝑓(𝑥) = and its antiderivatives
𝑙𝑛 𝑥
𝑥
𝐹(𝑥) = ∫0 𝑓(𝑡) 𝑑𝑡 is shown in Figure 2.

𝑥−1
Fig. 2. The graph of 𝑓(𝑥) = ln 𝑥 and its antiderivative within the domain 0 ≤ 𝑥 ≤ 2.

8
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

Example 3.3.

Here, we consider the integral

𝜋
2
𝐼 = ∫ 𝑥 𝑐𝑜𝑡 (𝑥) 𝑑𝑥.
0

First, and before choosing a proper parameter 𝑝, we need to make some manipulation
with the function 𝑥𝑐𝑜𝑡 (𝑥) as follows:

𝑥
Let 𝑥 𝑐𝑜𝑡 (𝑥) = 𝑡𝑎𝑛 (𝑥) , so our integration becomes;

𝜋
2 𝑥 (30)
𝐼=∫ 𝑑𝑥
0 𝑡𝑎𝑛 (𝑥)
Now we can introduce 𝐼(𝑝) and rewrite the integral as

𝜋
2 𝑥 (31)
𝐼(𝑝) = ∫ 𝑑𝑥
0 𝑡𝑎𝑛 (𝑝𝑥)
where, 𝐼(1) = 𝐼, and 𝐼(0) = 0.

Differentiating both sides of Eq. (31) with respect to 𝑝 to have

𝜋
𝑑 2 1 𝜕 (32)
𝐼(𝑝) = ∫ ( 𝑥) ) 𝑑𝑥
𝑑𝑝 0 𝑡𝑎𝑛 (𝑥) 𝜕𝑝
𝜋
𝑑 2 1 tan 𝑥
𝐼(𝑝) = ∫ ( [ ]) 𝑑𝑥
𝑑𝑝 0 𝑡𝑎𝑛 (𝑥) 1 + 𝑝2 𝑥
(33)
𝜋
2 1
=∫ ( ) 𝑑𝑥.
0 1 + 𝑝2 𝑥
Let 𝑢 = 𝑡𝑎𝑛 𝑥 , 𝑑𝑢 = 𝑥 𝑑𝑥 (performing Integration by substitution method), we get

∞
𝑑 1 𝑑𝑢
𝐼(𝑝) = ∫ . (34)
𝑑𝑝 2 2 1 + 𝑢2
0 1+𝑝 𝑢

We can now calculate the integration by using the partial fraction method. So the
integral in the right hand side becomes

9
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023
𝑑 𝜋 1
𝐼(𝑝) = ( ), 𝐼(0) = 0. (35)
𝑑𝑝 2 𝑝+1
Taking the integral of both sides, we get

𝜋
𝐼(𝑝) = 𝑙𝑛 |𝑝 + 1| + 𝑐 (36)
2

Fig. 3. The graph of 𝑓(𝑥) = 𝑥 𝑐𝑜𝑡 (𝑥) and its antiderivative within the domain 0 ≤ 𝑥 ≤
1.6.

To find the value of c, we replace 𝑝 with 0 as follows:

𝜋
𝐼(0) = (𝑙𝑛 |0 + 1| ) + 𝑐. (37)
2

But 𝐼(0) = 0, then

𝜋
0= 𝑙𝑛 (1) + 𝑐, (38)
2
and we get

𝑐=0 (39)

By subtitling the value of 𝑐 in Equation (36), we have

10
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023
𝜋
𝐼(𝑝) = 𝑙𝑛 (𝑝 + 1) (40)
2
Therefore,

𝜋
𝐼 = 𝐼(1) = 𝑙𝑛 (2) ≅ 1.08879 . (41)
2

is shown in

Figure 3. shows the graph of the function 𝑓(𝑥) = 𝑥 cot 𝑥 along with its antiderivatives
𝑥
𝐹(𝑥) = ∫0 𝑓(𝑡) 𝑑𝑡.

Table1. Using Feynman’s method for calculating the definite integrals of some
non-elementary functions

The definite The assumption The initial The Solution

Integral for parameter 𝒑 conditions as a function
of 𝒑
𝑰𝟏 𝐼1 (𝑝) 𝐼1 (4) = 𝐼1 , 𝐼1 (𝑝) = 𝜋 √𝑝
∞
∞
𝒍𝒏 (𝟏 + 𝟒𝒙𝟐 ) 𝑙𝑛 (1 + 𝑝𝑥 2 ) 𝐼1 (0) = 0
=∫ 𝒅𝒙= ∫ 𝑑𝑥
𝟎 𝒙𝟐 0 𝑥2
𝑰𝟐 𝐼2 (𝑝) 𝐼2 (1) = 𝐼2 , √2𝜋𝑝
∞ 2
𝐼2 (𝑝) =
∞ 𝟐
𝒔𝒊𝒏 (𝒙 ) 𝑠𝑖𝑛 (𝑝𝑥 ) 𝐼2 (0) = 0 2
=∫ 𝒅𝒙 =∫ 𝑑𝑥
𝟎 𝒙𝟐 0 𝑥2
𝑰𝟑 𝐼3 (𝑝) 𝐼3 (1) = 𝐼2 , 𝜋𝑝2
𝐼3 (𝑝) =
∞
𝒔𝒊𝒏𝟐 (𝒙) ∞
𝑠𝑖𝑛2 (𝑝𝑥) 𝐼3 (0) = 0 2
=∫ 𝒅𝒙 =∫ 𝑑𝑥
𝟎 𝒙𝟐 0 𝑥2
𝑰𝟒 𝐼4 (𝑝) 𝐼4 (2) = 𝐼4 , 𝐼4 (𝑝) = 𝑝
∞ ∞ −𝑝𝑥
𝒆−𝟐𝒙 𝒔𝒊𝒏 (𝒙) 𝑒 𝑠𝑖𝑛(𝑥) 𝐼4 (∞) = 0
=∫ 𝒅𝒙 = ∫ 𝑑𝑥
𝟎 𝒙 0 𝑥
𝑰𝟓 𝐼5 (𝑝) 𝐼5 (1) = 𝐼5 , 𝐼5 (𝑝)
∞
𝒄𝒐𝒔(𝒙) ∞
𝑐𝑜𝑠 (𝑝𝑥) 𝐼5 (0) = 2
𝜋 𝑝 𝑝
=∫ 𝒅𝒙 =∫ 𝑑𝑥 𝜋 𝑠𝑖𝑛 (2) 𝑐𝑜𝑠 (2 )
𝟎 𝟏 + 𝒙𝟒 0 1 + 𝑥4 =
√2
2√2 𝑒 2

11
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

𝑰𝟔 𝐼6 (𝑝) 𝐼6 (1) = 𝐼7 , 𝐼6 (𝑝)

𝝅 𝜋
𝐼7 (0) = 0 𝜋
𝟐 2 (𝑝 𝑡𝑎𝑛 𝑥 ) = 𝑙𝑛 (𝑝
= ∫ 𝒙 𝒄𝒐𝒕 𝒙 𝒅𝒙 =∫ 𝑑𝑥 2
𝟎 0 𝑡𝑎𝑛 𝑥
+ 1)
𝑰𝟕 𝐼7 (𝑝) 𝐼7 (0) = 𝐼7 , 𝐼7 (𝑝)
𝝅
𝒍𝒏 (𝟏 + 𝒄𝒐𝒔 𝒙 ) 𝜋
𝑙𝑛 (1 + 𝑐𝑜𝑠 𝑝 𝑐𝑜𝑠 𝑥 𝐼) (𝜋 ) = 0 𝜋
=∫ =
𝒅𝒙∫ 7 𝑑𝑥
2
= 𝜋( − 𝑝)
𝒄𝒐𝒔 𝒙 𝑐𝑜𝑠 𝑥 2
𝟎 0

𝑰𝟖 𝐼8 (𝑝) 𝐼8 (1) = 𝐼8 , 𝐼8 (𝑝)

∞ ∞ 𝜋
=∫
𝟐
𝒆−𝒙 𝒄𝒐𝒔 𝒙 𝒅𝒙 =∫ 𝑒 −𝑥
2 𝐼8 ( 2 ) = 0 √𝜋 −𝑝2
= 𝑒 4
𝟎 0 2
𝑐𝑜𝑠 𝑝𝑥 𝑑𝑥
𝑰𝟗 𝐼9 (𝑝) 𝐼9 (0) = 𝐼9 𝐼9 (𝑝)
∞
∞
𝒔𝒊𝒏 (𝒙𝟑 ) −𝑝𝑥 3
𝑠𝑖𝑛 (𝑥 3 ) , 𝐼9 (∞) = 0 −1 𝜋
=∫ 𝒅𝒙 =∫ 𝑒 𝑑𝑥 = 𝑝+
𝟎 𝒙 0 𝑥 3 6

In Table 1. We present a few examples of improper integrals that have been solved
using the Feynman’s method. It could be noticed that the Feynman method was much
simpler to use in the time when it was difficult to find real solutions for some integrals,
even with computer software.

4 Conclusions and Remarks

The Feynman method has been used a lot when dealing with difficult integrations of
physical and mathematical problems. It is known that the ability of generalization is
one of the most important skills for solving mathematical problems, and this was
accomplished using this method; since we can get the solution as formula or a function
of a variable 𝑝, and this allows us to produce so many solutions based on the value of
𝑝.

The main idea behind " Feynman's method " is to solve a simple differential equation
instead of computing an integral. This method can be used to test the convergence and
divergence series. Getting the actual value of the integration is better than getting the
approximate value; for example, Taylor series can be used to solve the definite integral,
but unlike Feynman, it does not represent the actual value.

12
Academy journal for Basic and Applied Sciences (AJBAS) Volume 5 # 4 August 2023

The only drawback to this approach is that when you begin to solve the integral, it is
hard to predict the order of the differential equation. There is no clear algorithm to
construct a function of a parameter 𝑝; each problem has unique assumptions about 𝑝.
Therefore, designing a program to create 𝐼(𝑝) is difficult. The illustrated examples in
this paper demonstrated how well Feynman's method works to solve challenging
integrations. The majority of the examples presented in Table 1. shows the results of using
Feynman's method for formulae involve improper integral.

References

Ashlock, D. (2019). Methods of Integration II. In D. Ashlock, Fast Start Integral Calculus
(pp. 151--179). Springer.

Jeffrey, A., & Dai, H. H. (2008). Handbook of mathematical formulas and integrals.
Elsevier.

Kotikov, A. (1991). Differential equations method: the calculation of vertex-type

Feynman diagrams. Physics Letters B, 259, 314-322.

Shaxobiddinova, Z., & others. (2022). INTEGRATION OF FUNCTIONS: THE FEYNMAN

METHOD. Science and Innovation, 1, 554--562.

Stewart, J., Clegg, D. K., & Watson, S. (2020). Calculus. Cengage Learning.

Zwillinger, D., & Jeffrey, A. (Eds.). (2007). Table of integrals, series, and products.
Elsevier

13