Advanced Integration

Techniques

Advanced approaches for solving many complex integrals using special functions, some
transformations and complex analysis approaches

Third Version

ZAID ALYAFEAI
YEMEN

mailto:[email protected]
Contents

1 Differentiation under the integral sign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.1 Basic Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.1.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3 Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.4 Inverse Laplace transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.4.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.5 Interesting results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.5.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.5.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.5.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.5 Extension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.5.1 Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.5.2 Reduction formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.6 Other Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.6.1 Euler Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.6.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.6.3 Weierstrass Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.7 Laurent expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.9 More values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.10 Legendre Duplication Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1
 3.11 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.12 Euler’s Reflection Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.13 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.14 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Beta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1 Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1.1 First integral formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1.2 Second integral formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1.3 Geometric representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.9 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.10 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5 Digamma function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.3 Difference formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.3.1 First difference formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.3.2 Second difference formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.5 Series Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.6 Some Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.8 Integral representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.8.1 First Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.8.2 Second Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.8.3 Third Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.8.4 Fourth Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2
 5.9 Gauss Digamma theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.10 More results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.11 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.12 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.13 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.14 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5.15 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6 Zeta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.2 Bernoulli numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.3 Relation between zeta and Bernoulli numbers . . . . . . . . . . . . . . . . . . . . . . . . 57

6.4 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6.5 Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6.6 Hurwitz zeta and polygamma functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.6.2 Relation between zeta and polygamma . . . . . . . . . . . . . . . . . . . . . . 59

6.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7 Dirichlet eta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

7.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

7.2 Relation to Zeta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

7.3 Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

8 Polylogarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

8.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

8.2 Relation to other functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

8.3 Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

8.4 Square formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

8.5 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

8.6 Dilogarithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

8.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

8.6.2 First functional equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

8.6.3 Second functional equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

8.6.4 Third functional equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

8.6.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3
 8.6.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

8.6.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

8.6.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

9 Ordinary Hypergeometric function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

9.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

9.2 Some expansions using the hypergeomtric function . . . . . . . . . . . . . . . . . . . . 73

9.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

9.4 Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

9.5 Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

9.6 Special values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

10 Error Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.2 Complementary error function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.3 Imaginary error function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.4 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.5 Relation to other functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

10.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

10.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

10.9 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

10.10 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

11 Exponential integral function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

11.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

11.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

11.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

11.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

11.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

11.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

11.7 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

12 Complete Elliptic Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

12.1 Complete elliptic of first kind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

12.2 Complete elliptic of second kind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

12.3 Hypergeometric representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4
 12.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

12.5 Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

12.6 Special values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

12.7 Differentiation of elliptic integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

13 Euler sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

13.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

13.2 Generating function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

13.3 Integral representation of Harmonic numbers . . . . . . . . . . . . . . . . . . . . . . . . 99

13.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

13.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

13.6 General formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

13.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

13.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

13.9 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

13.10 Relation to polygamma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

13.11 Integral representation for r=1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

13.12 Symmetric formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

13.13 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

14 Sine Integral function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

14.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

14.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

14.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

14.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

14.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

14.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

14.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

14.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

15 Cosine Integral function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

15.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

15.2 Relation to Euler constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

15.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

15.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

15.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

5
 15.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

15.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

15.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

16 Integrals involving Cosine and Sine Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

16.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

16.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

17 Logarithm Integral function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

17.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

17.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

17.3 Find the integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

17.4 Find the integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

17.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

17.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

18 Clausen functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

18.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

18.2 Duplication formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

18.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

18.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

19 Clausen Integral function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

19.1 Definiton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

19.2 Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

19.3 Duplication formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

19.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

19.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

19.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

19.7 Second Integral representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

19.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

20 Barnes G function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

20.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

20.1.1 Functional equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

20.2 Reflection formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

20.3 Values at positive integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

20.4 Relation to Hyperfactorial function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6
 20.5 Loggamma integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

20.6 Glaisher-Kinkelin constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

20.7 Relation to Glaisher-Kinkelin constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

20.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

20.9 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

20.10 Relation to Howrtiz zeta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

20.11 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

21 Complex Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

21.1 Introduction to complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

21.2 Polar representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

21.3 Complex functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

21.3.1 Exponential function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

21.3.2 Sine and Cosine and hyperbolic functions . . . . . . . . . . . . . . . . . . . . 158

21.3.3 Complex logarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

21.4 Taylor and Laurent expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

21.5 Poles and residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

21.6 Integration around paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

21.7 Bounds on integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

21.8 Contours around poles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

22 Real integrals using contour integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

22.1 Trigonometric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

22.1.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

22.2 Integrating around an ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

22.2.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

22.3 Creating crazy integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

22.3.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

22.3.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

22.3.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

22.4 Trigonometric functions with rationals of polynomials . . . . . . . . . . . . . . . . . . 181

22.4.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

22.5 Integration along contours with detours . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

22.5.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

22.6 Integrals of functions with branch cuts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

7
 22.6.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

22.6.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

22.6.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

22.6.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

22.6.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

22.6.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

22.7 Rectangular contours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

22.7.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

22.7.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

22.7.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

22.8 Triangular contours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

22.8.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

22.9 Residue at infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

22.9.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

22.10 Inverse of Laplace transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

22.10.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

22.11 Infinite sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

22.11.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

8
 
ÕækQË@ á ÔgQË@  Õæ„
 é<Ë@

 .

9
Acknowledgement

I want to offer my sincerest gratitude to all those who supported me during my journey to finish this

book. Especially my parents, sisters and friends who supported the idea of this book. I also want to

thank my Math teachers at King Fahd University because I wouldn’t be able to learn the advanced

without having knowledge of the elementary. I also want to extend my thanks to all my friends on the

different math forums like MMF, MHB and stack exchange without them I wouldn’t be learning any

thing.

Reviewers

A special thank for Mohammad Nather Shaaban for reviewing some parts of the book.

What is new?

The new version is all about contour integration using the concepts from complex analysis. One might

deviate from such approaches because of the heavy theory behind them but I tried to give a brief

overview of the theory before delving deep into approaches.

The future work

I have a plan to add many other sections. Basically I’ll try to focus on transformations like Mellin and

fourier transforms. Also many other functions like the Jacobi theta function and q-series.

10
Introduction

This book is a summary of working on advanced integrations for around five years. It collects many

examples that I gathered during that period. The approaches taken to solve the integrals aren’t neces-

sarily the only and best methods but they are offered for the sake of explaining the topic. Most of the

content of this book I already wrote on mathhelpboards.com during the past three years but I thought

that publishing it using a pdf would be easier to read and distribute. The motivation behind this book is

to allow those who are interested in solving complicated integrals to be able to use the different methods

to solve them efficiently. When I started learning about these techniques I would suffer to get enough

information about all the required approaches so I tried to collect every thing in just one book. You are

free to distribute this book and use any of the methods to solve the integrals or use the same techniques.

The methods used are not necessarily new or ground-breaking but as I said they introduce the concept

as easy as possible.

To follow this book you have to be know the basic integration techniques like integration by parts, by

substitution and by partial fractions. I don’t assume that the readers know any other stuff from any

other topics or advanced courses from mathematics. Usually the details that require deep knowledge of

analysis or advanced topics are left or just touched upon lightly to give the reader some hints but not

going into details.

After reading this book you should be able to solve many advanced integrals that you might face

in engineering courses. I hope you enjoy reading this book and if you have any suggestions, com-

ments or correction I will be happy to recieve them through my email mailto:[email protected]

or this email mailto:[email protected]. Also I am avilable as a staff member at http://www.

mathhelpboards.com if you have some questions that I could reply to you directly using Latex.

11
1 Differentiation under the integral sign

This is one of the most commonly used techniques to solve a numerous number of questions.

Assume that we have the following function of two variables

b
∫ f (x, y) dx
a

Then we can differentiate with respect to y provided that f is continuous and has a partial continuous

derivative on a chosen interval

b
F ′ (y) = ∫ fy (x, y) dx
a

Now using this in many problems is not that clear you have to think a lot to get the required answer

because many integrals are usually in one variable so you need to introduce the second variable and

assume it is a function of two variables.

1.1 Example

Assume we want to solve the following integral

1 x2 − 1
∫ dx
0 log(x)
That seems very difficult to solve but using this technique we can solve it easily. The crux move is to

decide where to put the second variable! So the problem with the integral is that we have a logarithm

in the denominator which makes the problem so difficult to tackle! Remember that we can get a natural

logarithm if we differentiate exponential functions i.e F (a) = 2a ⇒ F ′ (a) = log(2) ⋅ 2a

Applying this to our problem

1 xa − 1
F (a) = ∫ dx
0 log(x)
Now we take the partial derivative with respect to a

1 ∂ xa − 1 1 1
F ′ (a) = ∫ ( ) dx = ∫ xa dx =
0 ∂a log(x) 0 a+1
Integrate with respect to a

F (a) = log (a + 1) + C

12
To find the value of the constant put a = 0

F (0) = log(1) + C Ô⇒ C = 0

This implies that

1 xa − 1
∫ dx = log (a + 1)
0 log(x)
By this powerful method we were not only able to solve the integral we also found a general formula for

some a where the function is differentiable in the second variable.

To solve our original integral put a = 2

1 x2 − 1
∫ dx = log (2 + 1) = log(3)
0 log(x)

1.2 Example

Find the following integral

π
2 x
∫ dx
0 tan x
So where do we put the variable a here? that doesn’t seem to be straight forward , how do we proceed ?

Let us try the following

π
2 arctan(a tan(x))
F (a) = ∫ dx
0 tan(x)
Now differentiate with respect to a

π
1
F ′ (a) = ∫
2
dx
0 1 + (a tan(x))2
It can be proved that

π
2 1 π
∫ dx =
0 1 + (a tan(x))2 2(1 + a)
Now Integrate both sides

π
F (a) = log(1 + a) + C
2
Substitute a = 0 to find C = 0

13
 π
2 arctan(a tan(x)) π
∫ dx = log(1 + a)
0 tan(x) 2
Put a = 1 in order to get our original integral

π
2 x π
∫ dx = log(2)
0 tan(x) 2

1.3 Example
∞ sin(x)
∫ dx
0 x
This problem can be solved by many ways , but here we will try to solve it by differentiation. So as I

showed in the previous examples it is generally not easy to find the function to differentiate. Actually

this step might require trial and error techniques until we get the desired result, so don’t just give up if

an approach doesn’t work!

Let us try this one

∞ sin(ax)
F (a) = ∫ dx
0 x
If we differentiated with respect to a we would get the following

∞
F ′ (a) = ∫ cos(ax) dx
0

But unfortunately this integral doesn’t converge, so this is not the correct one. Actually, the previous

theorem will not work here because the integral is improper.

So let us try the following

∞ sin(x)e−ax
F (a) = ∫ dx
0 x
Take the derivative

∞
F ′ (a) = − ∫ sin(x)e−ax dx
0

Use integration by parts twice

∞ −1
F ′ (a) = − ∫ sin(x)e−ax dx =
0 a2 + 1
Integrate both sides

14
 F (a) = − arctan(a) + C

To find the value of the constant take the limit as a grows large

π
C = lim F (a) + arctan(a) =
a→∞ 2
So we get our F (a) as the following

π
F (a) = − arctan(a) +
2
For a = 0 we have

∞ sin(x) π
∫ =
0 x 2

15
2 Laplace Transform

2.1 Basic Introduction

Laplace transform is a powerful integral transform. It can be used in many applications. For example, it

can be used to solve Differential Equations and its rules can be used to solve integration problems.

The basic definition of Laplace transform

∞
F (s) = L(f (t)) = ∫ e−st f (t) dt
0

This integral will converge when

Re(s) > a , ∣f (t)∣ ≤ M eat

Let us see the Laplace transform for some functions

2.1.1 Example

Find the Laplace transform of the following functions

1. f (t) = 1

∞ 1
F (s) = ∫ e−st dt =
0 s

2. For f (t) = tn where n ≥ 0

We can prove using integration by parts

∞ n!
F (s) = ∫ e−st tn dt =
0 sn+1

3. For the geometric function f (t) = cos(at), Use integration by parts

∞ s
F (s) = ∫ e−st cos(at) dt =
0 s2 + a2

16
2.2 Example

Find the following integral

∞
∫ e−2t t3 dt
0

We can directly use the formula in the previous example

∞ n!
∫ e−st tn dt =
0 sn+1
Here we have s = 2 and n = 3

∞ 3! 3
∫ e−2t t3 dt = =
0 23+1 8

2.3 Convolution

Define the following integral

t
(f ∗ g)(t) = ∫ f (s)g(t − s) ds
0

Then we have the following

L ((f ∗ g)(t)) = L(f (t))L(g(t))

2.4 Inverse Laplace transform

So, basically you are given F (s) and we want to get f (t) this is denoted by

L(f (t)) = F (s) Ô⇒ L−1 (F (s)) = f (t)

2.4.1 Example

Find the inverse Laplace transform of

1. F (s) = 1
s3

We use the results applied previously

2! 1 1
L(t2 ) = 3
⇒ L(t2 ) = 3
s 2 s

17
 Now take the inverse to both sides

t2 1
= L−1 ( 3 )
2 s

2. F (s) = s
s2 +4

we can use the Laplace of cosine to deduce

s
cos(2t) = L−1 ( )
s2 + 4

Exercises

Find the Laplace transform

sin(at)

Find the inverse Laplace

1
sn+1

2.5 Interesting results

2.5.1 Example

Prove the following

1 Γ(x)Γ(y)
β(x, y) = ∫ tx−1 (1 − t)y−1 dt =
0 Γ(x + y)
β is the Beta function and Γ is the Gamma function. We will take enough time and examples to explain both

functions in the next sections.

proof

We need convolution rule we described earlier

Let us choose some functions f and g

f (t) = tx , g(t) = ty

Hence we get

t
(tx ∗ ty ) = ∫ sx (t − s)y ds
0

So by the convolution rule we have the following

18
 L (tx ∗ ty ) = L(tx )L(ty )

We can now use the Laplace of the power

x! ⋅ y!
L (tx ∗ ty ) =
sx+y+2
Notice that we need to find the inverse of Laplace L−1

x! ⋅ y! x! ⋅ y!
L−1 (L(tx ∗ ty )) = L−1 ( ) = tx+y+1
sx+y+2 (x + y + 1)!
So we have the following

x! ⋅ y!
(tx ∗ ty ) = tx+y+1
(x + y + 1)!
By definition we have

x! ⋅ y! t
tx+y+1 = ∫ sx (t − s)y ds
(x + y + 1)! 0

Now put t = 1 we get

x! ⋅ y! 1
= ∫ sx (1 − s)y ds
(x + y + 1)! 0

By using that n! = Γ(n + 1) we deduce that

1 Γ(x + 1)Γ(y + 1)
∫ sx (1 − s)y ds =
0 Γ(x + y + 2)
which can be written as

1 Γ(x)Γ(y)
∫ sx−1 (1 − s)y−1 ds =
0 Γ(x + y)

2.5.2 Example

Prove the following

∞ f (t) ∞
∫ dt = ∫ L(f (t)) ds
0 t 0

proof

we know from the definition

19
 ∞ ∞ ∞
∫ L(f (t)) ds = ∫ (∫ e−st f (t) dt) ds
0 0 0

Now by the Fubini theorem we can rearrange the double integral

∞ ∞
∫ f (t) (∫ e−st ds) dt
0 0

The integral inside the parenthesis

∞ 1
∫ e−st ds =
0 t
Now substitute this value in the integral

∞ f (t)
∫ dt
0 t

2.5.3 Example

Find the following integral

∞ sin(t)
∫ dt
0 t
This is not the first time we see this integral and not the last . We have seen that we can find it using

differentiation under the integral sign.

Let us use the previous example

∞ sin(t) ∞
∫ dt = ∫ L(sin(t)) ds
0 t 0

We can prove that

1
L(sin(t)) =
s2 + 1
Substitute in our integral

∞ ds π
∫ = tan−1 (s)∣s=∞ − tan−1 (s)∣s=0 =
0 1 + s2 2

20
3 Gamma Function

The gamma function is used to solve many interesting integrals, here we try to define some basic prop-

erties, prove some of them and take some examples.

3.1 Definition
∞
Γ(x + 1) = ∫ e−t tx dt
0

For the first glance that just looks like the Laplace Transform, actually they are closely related.

So let us for simplicity assume that x = n where n ≥ 0 (is an integer )

∞
Γ(n + 1) = ∫ e−t tn dt
0

We can use the Laplace transform

∞ n!
∫ e−t tn = ∣s=1 = n!
0 sn+1
So we see that there is a relation between the gamma function and the factorial. We will assume for the

time being that the gamma function is defined as the following

n! = Γ(n + 1)

This definition is somehow limited but it will be soon replaced by a stronger one.

3.2 Example

Find the following integrals

∞
∫ e−t t4 dt
0

By definition this can be replaced by

∞
∫ e−t t4 dt = Γ(4 + 1) = 4! = 24
0

21
3.3 Example

Solving the following integrals

1.
∞
e−t t dt
2
∫
0

We need a substitution before we go ahead, so let us start by putting x = t2 so the integral becomes

1 ∞ −x 21 − 12 1 1
∫ e x ⋅ x dt = Γ(1 + 0) =
2 0 2 2

2.
1
∫ log(t) t2 dt
0

we use the substitution t = e− 2

x

1 1 ∞ − 3x
∫ log(t) t2 dt = − ∫ e 2 ⋅ x dx
0 4 0

Using another substitution t = 3x

2

−1 ∞ −x −Γ(2) −1
∫ e x dx = =
9 0 9 9

It is an important thing to get used to the symbol Γ. I am sure that you are saying that this seems

elementary, but my main aim here is to let you practice the new symbol and get used to solving some

problems using it.

3.4 Exercises

Prove that

Γ(5) ⋅ Γ(2) 1
=
Γ(7) 30
Find the following integral

∞
e− 60 t t20 dt
1
∫
0

22
3.5 Extension

For simplicity we assumed that the gamma function only works for positive integers. This definition

was so helpful as we assumed the relation between gamma and factorial. Actually, this restricts the

gamma function, we want to exploit the real strength of this function. Hence, we must extend the gamma

function to work for all real numbers except for some values. Actually we will see soon that we can

extend it to work for all complex numbers except where the function has poles.

3.5.1 Theorem

Using the integral representation we can extend the gamma function to x > −1.

proof

We need only consider the case when −1 < x < 0.

Near infinity we have the following

∞ ∞
∣∫ e−t tx dt∣ ≤ ∫ e−t dt < ∞
0 

Near zero when x = −z we have the following

 e−t  1
∣∫ dt∣ ∼ ∫ dt < ∞
0 tz 0 tz

3.5.2 Reduction formula

Γ(x + 1) = xΓ(x)

This can be proved through integration by parts for x > 0. Actually this representation allows us to

extend the gamma function for all real numbers for non-negative integers. In terms of complex analysis

this function is analytic except at non-positive integers where it has poles.

3.6 Other Representations

3.6.1 Euler Representation

z
1 ∞ (1 + k )
1
nz n k
Γ(z) = lim ∏ = ∏
k=1 k + z z k=1 1 + kz
n→∞ z

proof

Note that

n
Γ(z + n + 1) = Γ(z + 1) ∏ (k + z)
k=1

23
Which indicates that

n
Γ(z + n + 1)
∏ (k + z) =
k=1 zΓ(z)
Also note that

n
∏ k = n!
k=1

Hence we have

nz n k nz × n!
lim ∏ = Γ(z) lim
k=1 k + z
n→∞ z n→∞ Γ(z + n + 1)

Hence we must show that

nz × n!
lim =1
n→∞ Γ(z + n + 1)

Note that by Stirling formula

√
Γ(z + n + 1) ∼ 2π(n + z)n+z+1/2 e−(n+z)

and

√
n! ∼ 2πnn+1/2 e−n

Hence we have by

√
nz × ( 2πnn+1/2 e−n ) nn
lim √ = lim =1
n→∞ 2π(n + z)n+z+1/2 e−(n+z) n→∞ (n + z)n e−z
Note that
z n
lim (1 + ) = ez
n→∞ n
To prove the other product formula note that

z
n
1 z ∏nk=1 (1 + k)
∏ (1 + ) = n = (n + 1)z ∼ nz
k=1 k ∏k=1 k z
Hence we deduce

z z
∏k=1 (1 + k1 ) n 1 1 ∞ (1 + k )
n 1
nz n k
lim ∏ = lim ∏ = ∏
k=1 k + z k=1 1 + k z k=1 1 + kz
n→∞ z z
n→∞ z

24
3.6.2 Example

Prove that

∞
Γ(x)Γ(y) z z
= ∏ [(1 + ) (1 − )]
Γ(x + z)Γ(y − z) k=0 x+k y+k
proof

Start by

nz n k
Γ(z) = lim ∏
k=1 k + z
n→∞ z

We have

x y

Γ(x)Γ(y) ( nx ∏nk=1 k
k+x
) ( ny n
∏k=1 k
k+y
)
= lim
Γ(x + z)Γ(y − z) n→∞ ( nx+z ∏n
k=1 k+x+z ) ( y−z
ny−z
x+z
k
∏k=1
n k
k+y−z
)
By simplifications we have

Γ(x)Γ(y) (x + z)(y − z) n (k + x + z)(k + y − z)

= lim ∏
Γ(x + z)Γ(y − z) n→∞ xy k=1 (k + x)(k + y)
This simplifies to

∞
Γ(x)Γ(y) (k + x + z)(k + y − z) ∞ z z
=∏ = ∏ [(1 + ) (1 − )]
Γ(x + z)Γ(y − z) k=0 (k + x)(k + y) k=0 x+k y+k

3.6.3 Weierstrass Representation

e−γz ∞ z −1 z/n
Γ(z) = ∏ (1 + ) e
z n=1 n
where γ is the Euler constant

proof

Take logarithm to the Euler representation

n n
z
log zΓ(z) = lim z ∑ (log (1 + k) − log(k)) − ∑ log (1 + )
n→∞
k=1 k=1 k
Note the alternating sum

n
∑ (log (1 + k) − log(k)) = log(n + 1)
k=1

Hence we have

25
 n
z
log zΓ(z) = lim z log(n + 1) − ∑ log (1 + )
n→∞
k=1 k
Now we can use the harmonic numbers

n
1
Hn = ∑
k=1 n
Add and subtract zHn+1

n
z −1 z z
log zΓ(z) = lim z log(n + 1) − zHn+1 + ∑ [log (1 + ) + ] +
n→∞
k=1 k k n +1
The last term goes to zero and by definition we have the Euler constant

γ = lim Hn − log(n)
n→∞

Hence the first term is the Euler constant

∞
z −1 z
log zΓ(z) = −zγ + ∑ log (1 + ) +
k=1 k k
By taking the exponent of both sides

∞
z −1 z
zΓ(z) = e−γz ∏ (1 + ) e k
k=1 k

3.7 Laurent expansion

1 1
Γ(z) = − γ + (γ 2 + ζ(2))z + O(z 2 )
z 2!
proof

Note that f (z) = Γ(z + 1) has a Maclurain expansion near 0

∞
Γ(k) (1) k
Γ(z + 1) = ∑ z
k=0 k!
For the first term

f (0) = Γ(1 + 0) = 1

For the second term

f ′ (0)
= Γ′ (1)
1!

26
To find the derivative, note that by the Weierstrass representation

∞
z −1 z
log Γ(z) = −γz − log(z) + ∑ log (1 + ) +
n=1 n n
By taking the derivaive we have

Γ′ (z) 1 ∞
1
= −γ − + z ∑
Γ(z) z k=1 k(z + k)
Hence we have
∞
1
Γ′ (1) = −γ − 1 + ∑ = −γ − 1 + 1 = −γ
k=1 k(1 + k)

For the third term

f ′′ (0) Γ′′ (1)

=
2! 2
Taking the second derivative

Γ′′ (z)Γ(z) − (Γ′ (z))2 1 ∞

1
= + ∑
Γ (z)
2 z 2
k=1 (z + k)
2

Which indicates that

∞
1
Γ′′ (1) = (Γ′ (z))2 + 1 + ∑ = γ 2 + ζ(2)
k=1 (1 + k) 2

Hence we deduce that

1 2
Γ(z + 1) = 1 − γz + (γ + ζ(2))z 2 + O(z 3 )
2!
Dividing by z we get our result.

3.8 Example

Find the integral

∞ e−t
∫ √ dt
0 t
Now according to our definition this is equal to Γ ( 21 ) but this value can be represented using elementary

functions as follows
√
Let us first make a substitution t=x

∞
e−x dx
2
2∫
0

27
Now to find this integral we need to do a simple trick, start by the following

∞ 2 ∞ ∞
e−x dx) = (∫ e−x dx) ⋅ (∫ e−x dx)
2 2 2
(∫
0 0 0

Since x is a dummy variable we can put

∞ ∞
e−x dx) ⋅ (∫ e−y dy)
2 2
(∫
0 0

Now since they are two independent variables we can do the following

∞ ∞
e−(x +y 2 )
2
∫ ∫ dy dx
0 0

Now by polar substitution we get

π
∞
e−r r dr dθ
2 2
∫ ∫
0 0

The inner integral is 21 , hence we get

π
2 1 π
∫ dθ =
0 2 4
So we have

∞ 2
π
e−x dx) =
2
(∫
0 4
Take the square root to both sides

∞
√
π
e−x dx =
2
∫
0 2

∞ √
e−x dx =
2
2∫ π
0

So we have our result

1 √
Γ( ) = π
2

3.9 More values

We can use the reduction formula and the value of Γ(1/2) to deduce other values. Assume that we want

to find

28
 3
Γ( )
2
If we used this property we get

√
1 1 1 π
Γ (1 + ) = ⋅ Γ ( ) =
2 2 2 2
Not all the time the result will be reduced to a simpler form as the previous example. For example we

don’t know how to express Γ( 41 ) in a simpler form but we can approximate its value

1
Γ ( ) ≈ 3.6256 ⋯
4
Hence we just solve some integrals in terms of gamma function since we don’t know a simpler form.

For example solve the integral

∞
e−t t 4 dt
1
∫
0

we know by definition of gamma function that this reduces to

∞ 5 Γ ( 41 )
e−t t 4 dt = Γ ( ) =
1
∫
0 4 4
√
We have seen that Γ ( 21 ) = π but what about Γ ( −1
2
)?

By the reduction formula

1 −1 −1
Γ (1 − ) = Γ( )
2 2 2
so we have that

1 √
Γ (− ) = −2 π
2
Then we can prove that any fraction where the denominator equals to 2 and the numerator is odd can be

reduced into

2n + 1 1
Γ( ) = C Γ( ) , C ∈ Q , n ∈ Z
2 2

29
3.10 Legendre Duplication Formula
1 (2n)! √
Γ ( + n) = n π
2 4 n!
proof

For the proof we use induction by assuming n ≥ 0. If n = 0 we have our basic identity

1 √
Γ( ) = π
2
Now we need to prove that

1 (2n)! √ 1 (2n + 2)! √

Γ ( + n) = n π Ô⇒ Γ ( + n + 1) = n+1 π
2 4 n! 2 4 (n + 1)!
Now we use the reduction formula

1 1 + 2n 1
Γ ( + n + 1) = Γ ( + n)
2 2 2
By the inductive step we have

1 + 2n 1 1 + 2n (2n)! √
Γ ( + n) = ⋅ n π
2 2 2 4 n!
We can multiply and divide by 2n + 2

1 + 2n (2n)! √ 2n + 2 (2n + 2)! √

⋅ n π⋅ = n+1 π
2 4 n! 2n + 2 4 (n + 1)!

3.11 Example
√
∞ e−t cosh(a t)
∫ √ dt
0 t
we have a hyperbolic function

We know that we can expand cosh using power series

∞
x2n
cosh(x) = ∑
n=0 (2n)!
√
Let x = a t

√ ∞
a2n ⋅ tn
cosh(a t) = ∑
n=0 (2n)!

Substituting back in the integral we have

30
 ∞ ∞
a2n ⋅ tn
∫ e−t ∑ √ dt
0 n=0 (2n)! t

Now since the series is always positive we can swap the integral and the series

∞ ∞
a2n
e−t tn− 2 dt]
1
∑ [∫
n=0 (2n)! 0

Hence we have by using the gamma function

∞ a2n Γ ( 21 + n)
∑
n=0 (2n!)
Using LDF (Legendre Duplication Formula) we get

∞
a2n (2n)! √
∑ ( n π)
n=0 (2n!) 4 n!
By further simplification

√ ∞
a2n
π ∑ n
n=0 4 n!

Now that looks familiar since we know that

∞
zn
∑ = ez
n=0 n!
a2 √
Putting z = 4
and multiplying by π we get

2 n
√ ∞ ( a4 ) √ a2
π∑ = πe 4

n=0 n!
So we have finally that

√
∞ e−t cosh(a t) √ a2
∫ √ dt = πe 4
0 t

3.12 Euler’s Reflection Formula

π
Γ (z) Γ (1 − z) = , ∀z ∉ Z
sin (πz)
proof

We have to use the sine infinite product formula

−1
π 1 ∞ z2
= ∏ (1 − 2 )
sin(πz) z n=1 n

31
Now we start by noting that

Γ(z)Γ(1 − z) = −zΓ(z)Γ(−z)

Now using the Weierstrass formula we have

e−γz ∞ z −1 z/n eγz ∞ z −1 −z/n

−zΓ(z)Γ(−z) = −z ⋅ ∏ (1 + ) e ⋅ ∏ (1 − ) e
z n=1 n −z n=1 n
This simplifies to

−1
1 ∞ z2 π
∏ (1 − 2 ) =
z n=1 n sin(πz)

3.13 Example

Find the following

1.
3 1
Γ( ) Γ( )
4 4

The first example we can write

1 1
Γ (1 − ) Γ ( )
4 4

Now by ERF (Euler reflection formula) we have the following

1 1 π √
Γ (1 − ) Γ ( ) = = 2π
4 4 sin ( 4 )
π

2.
1+i 1−i
Γ( ) Γ( )
2 2

Using the same idea for the second one

1+i 1+i
Γ( ) Γ(1 − )
2 2

This expression simplifies to

π π
=
sin ( π(1+i) ) cos ( iπ
2
)
2

32
 By geometry to hyperbolic conversions we get

π π
= π sech ( )
cosh ( 2 )
π 2

3.14 Example

Find the integral

a+1
∫ log Γ(x)dx
a

Let the following

a+1
f (a) = ∫ log Γ(x)dx
a

Differentiate both sides

f ′ (a) = log Γ(1 + a) − log Γ(a) = log(a)

Integrate both sides

f (a) = a log(a) − a + C

Let a → 0

We have

1
C=∫ log Γ(x) dx
0

By the reflection formula

1 1 1 1
∫ log Γ(x) dx = ∫ log π dx − ∫ log sin(πx) dx − ∫ log Γ(1 − x) dx
0 0 0 0

Which implies that

1 1 1 1
2∫ log Γ(x) dx = ∫ log π dx − ∫ log sin(πx) dx = log(2π) − ∫ log ∣2 sin(πx)∣ dx
0 0 0 0

Note that this is the Clausen Integral

1 2 2π 2
∫ log ∣2 sin(πx)∣ dx = ∫ log ∣2 sin(x/2)∣ dx = cl2 (2π) = 0
0 π 0 π

33
Hence we finalize by

1 1
∫ log Γ(x) dx = log(2π)
0 2

34
4 Beta Function

4.1 Representations

4.1.1 First integral formula

1
∫ tx−1 (1 − t)y−1 dt = B(x, y)
0

It is related to the gamma function through the identity

Γ(x)Γ(y)
β(x, y) =
Γ(x + y)
We have proved this identity earlier when we discussed convolution.

We shall realize the symmetry of beta function that is to say

β(x, y) = β(y, x)

Beta function has many other representations all can be deduced through substitutions

4.1.2 Second integral formula

∞ tx−1
B(x, y) = ∫ dt
0 (1 + t)x+y

4.1.3 Geometric representation

π
2
B(x, y) = 2 ∫ cos2x−1 (t) sin2y−1 (t) dt
0

The proofs are left to the reader as practice.

4.2 Example

Prove the following

∞ 1 π
∫ dz =
0 z2 + 1 2
proof
√
Put z = t

1 ∞ t2
−1

∫ dt
2 0 t+1

35
We can use the second integral representation by finding the values of x and y

−1 1
x−1= ⇒ x=
2 2

1
x+y =1 ⇒ y =
2
Hence we have

√ √
1 ∞ t2 B( 21 , 12 ) Γ( 12 ) Γ( 12 ) π⋅ π π
−1

∫ dt = = = =
2 0 (t + 1) 2 2 2 2

4.3 Example
∞ 1
∫ dz
0 (z 2 + 1)2
Using the same substitution as the previous example we get

1 ∞ t2
−1

∫ dt
2 0 (t + 1)2
Then we can find the values of x and y

−1 1
x−1= ⇒ x=
2 2

3
x+y =2 ⇒ y =
2
Then

1 ∞ t2 B( 12 , 32 ) Γ( 21 ) Γ( 32 ) Γ( 21 ) Γ( 12 ) π
−1

∫ dt = = = =
2 0 (t + 1)2 2 2 4 4

4.4 Example

Find the generalization

∞ 1 1
∫ dx , ∀ n >
0 (x2 + 1)n 2
Using the same substitution again

1 ∞ t2
−1

∫ dt
2 0 (t + 1)n

36
Then we can find the values of x and y

−1 1
x−1= ⇒ x=
2 2

1
x+y =n ⇒ y =n−
2
Then

1 ∞ t2 Γ( 12 ) ⋅ Γ (n − 21 )
−1

∫ dt =
2 0 (t + 1)n 2Γ(n)
Now by LDF

√ √
1 (2n − 2)! π Γ (2n − 1) π
Γ (n − ) = n−1 =
2 4 (n − 1)! 4n−1 Γ(n)
Substituting in our integral we have the following

1 ∞ t2 2π ⋅ Γ (2n − 1)
−1

∫ dt =
2 0 (t + 1) n 4n ⋅ Γ2 (n)

∞ 1 π ⋅ Γ (2n − 1)
∫ dx = 2n−1 2
0 (x2 + 1)n 2 ⋅ Γ (n)
It is easy to see that for n ∈ Z+ we get a π multiplied by some rational number.

4.5 Example
1 zn (2n)!!
∫ √ dz = 2 ⋅
0 1−z (2n + 1)!!
Where the double factorial !! is defined as the following

⎧
⎪
⎪
⎪
⎪n ⋅ (n − 2) ⋯ 5 ⋅ 3 ⋅ 1 ; if n is odd
⎪
⎪
⎪
⎪
⎪
⎪
n!! = ⎨n ⋅ (n − 2) ⋯ 6 ⋅ 4 ⋅ 2 ; if n is even
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪1 ; if n = 0
⎩
The integral in hand can be rewritten as

1
z n ⋅ (1 − z)− 2 dz
1
∫
0

We find the variables x and y

37
 x−1=n ⇒ x=n+1

−1 1
y−1= ⇒ y=
2 2
This can be written as

1 1
z n ⋅ (1 − z)− 2 dz = B(n + 1, )
1
∫
0 2
By some simplifications

√
1 Γ ( 12 ) Γ(n + 1) π Γ(n + 1)
B (n + 1, ) = =
2 Γ (n + 2 )
3
(n + 12 )Γ (n + 12 )
Now you shall realize that we must use LDF

√ √
πΓ(n + 1) 2 π n! 2 ⋅ 22n (n!)2
= √ =
(n + 21 )Γ (n + 12 ) (2n + 1) π (2n)!
4n n!
(2n)!
Now we should separate odd and even terms in the denominator

22n (n ⋅ (n − 1) ⋯ 3 ⋅ 2 ⋅ 1)2
2⋅
(2n ⋅ (2n − 2) ⋯ 4 ⋅ 2)((2n + 1) ⋅ (2n − 1) ⋯ 3 ⋅ 1)
We insert 22n into the square to obtain

(2n ⋅ (2n − 2) ⋯ 6 ⋅ 4 ⋅ 2)2 (2n)!!

2⋅ =2⋅
(2n ⋅ (2n − 2) ⋯ 4 ⋅ 2)((2n + 1) ⋅ (2n − 1) ⋯ 3 ⋅ 1) (2n + 1)!!

4.6 Example

Find the following integral

−n
∞ x2 2

∫ (1 + ) dx
−∞ n−1
First we shall realize the evenness of the integral

−n
∞ x2 2

2∫ (1 + ) dx
0 n−1
x2
Let t = n−1

√ ∞ t− 2
1

n − 1∫ n dt
0 (1 + t) 2

38
Now we see that our integral becomes so familiar

√
√ 1 n−1 π(n − 1)Γ ( n−1 )
n − 1B ( , )= 2
2 2 Γ ( n2 )

4.7 Example

Find the following integral

∞ x−p
∫ dx
0 x3 + 1
Let us do the substitution x3 = t

∞ t− 3
p+2
1
∫ dt
3 0 t+1
Now we should find x, y

1−p
x=
3

1−p
y+x=1 ⇒ y =1−
3
so we have our beta representation of the integral

B ( 1−p , 1−p ) Γ ( 1−p ) Γ (1 − 1−p

)
3 3
= 3 3
3 3
Now we should use ERF

Γ ( 1−p ) Γ (1 − 1−p
) π π π − πp
3 3
= = csc ( )
3 3 sin ( π(1−p)
3
) 3 3

4.8 Example

Now let us try to find

π
2
√
∫ sin3 z dz
0

Rewrite as

π
2 3
∫ sin 2 z cos0 z dx
0

39
This is the Geometric representation

3 5
2x − 1 = ⇒ x=
2 4

1
2y − 1 = 0 ⇒ y =
2
Then

π
2 3 B ( 45 , 12 ) Γ ( 54 ) Γ ( 12 )
∫ sin 2 z dz = =
0 2 2Γ ( 74 )

4.9 Example

Find the following integral

(sin z)i ⋅ (cos z)−i dz

2
∫
0

This is the geometric representation

1+i
2x − 1 = i ⇒ x =
2

1−i
2y − 1 = −i ⇒ y =
2
Then

1 1+i 1−i
Γ( ) Γ( )
2 2 2
Now we see that we have to use ERF

1 1+i 1+i π π π
Γ( ) Γ (1 − )= π(1+i)
= sech ( )
2 2 2 2 sin ( 2 ) 2 2

4.10 Exercise

Prove

∞ x2m+1 m! (n − m − 2)!
∫ dx =
0 (ax + c)
2 n 2(n − 1)! am+1 cn−m−1

40
5 Digamma function

5.1 Definition
Γ′ (x)
ψ(x) =
Γ(x)
We call digamma function the logarithmic derivative of the gamma function. Using this we can define

the derivative of the gamma function.

Γ′ (x) = ψ(x) Γ(x)

5.2 Example

Find the derivative of

Γ(2x + 1)
f (x) =
Γ(x)
We can use the differentiation rule for quotients

2Γ′ (2x + 1)Γ(x) − Γ′ (x)Γ(2x + 1)

Γ2 (x)
which can be rewritten as

2Γ(2x + 1)ψ(2x + 1)Γ(x) − ψ(x)Γ(x)Γ(2x + 1) Γ(2x + 1)

= (2ψ(2x + 1) − ψ(x))
Γ2 (x) Γ(x)

5.3 Difference formulas

5.3.1 First difference formula

ψ(1 − x) − ψ(x) = π cot(πx)

proof

We know by ERF that

Γ(x)Γ(1 − x) = π csc(πx)

Now differentiate both sides

41
 ψ(x)Γ(x)Γ(1 − x) − ψ(1 − x)Γ(x)Γ(1 − x) = −π 2 csc(πx) cot(πx)

Which can be simplified

Γ(x)Γ(1 − x) (ψ(1 − x) − ψ(x)) = π 2 csc(πx) cot(πx)

Further simplifications using ERF results in

ψ(1 − x) − ψ(x) = π cot(πx)

5.3.2 Second difference formula

1
ψ(1 + x) − ψ(x) =
x
proof

Let us start by the following

Γ(1 + x)
=x
Γ(x)
Now differentiate both sides

Γ(1 + x)
(ψ(1 + x) − ψ(x)) = 1
Γ(x)
Which simplifies to

Γ(x) 1
ψ(1 + x) − ψ(x) = =
Γ(1 + x) x

5.4 Example

Find the following integral

∞ log(x)
∫ dx
0 (1 + x2 )2
Consider the general case

∞ xa
∫ dx
0 (1 + x2 )2
Use the following substitution x2 = t

42
 a−1
1 ∞ t 2
∫ dt
2 0 (1 + t)2
By the beta function this is equivalent to

a−1
1 ∞ t 2 1 a+1 a+1 1 a+1 a+1
∫ dt = B ( ,2 − ) = Γ( ) Γ (2 − )
2 0 (1 + t)2 2 2 2 2 2 2
Differentiate with respect to a

a−1
1 ∞ log(t) t 2 1 a+1 a+1 a+1 a+1
F ′ (a) = ∫ dt = Γ ( ) Γ (2 − ) [ψ ( ) − ψ (2 − )]
4 0 (1 + t)2 4 2 2 2 2
Now put a = 0

1 ∞ log(t) t 2
−1
1 1 3 1 3
∫ dt = Γ ( ) Γ ( ) [ψ ( ) − ψ ( )]
4 0 (1 + t) 2 4 2 2 2 2
Now we use our second difference formula

1 3 1 1
ψ ( ) − ψ ( ) = − (ψ (1 + ) − ψ ( )) = −2
2 2 2 2
Also by some gamma manipulation we have

1 3 1 1 π
Γ ( ) Γ ( ) = Γ2 ( ) =
2 2 2 2 2
The integral reduces to

1 ∞ log(t) t 2
−1
π
∫ dt = −
4 0 (1 + t)2 4
Putting x2 = t we have our result

∞ log(x) π
∫ dx = −
0 (1 + x )
2 2 4

5.5 Series Representation

∞
1 x
ψ(x) = −γ − +∑
x n=1 n(n + x)
proof

We start by taking the logarithm of the Weierstrass representation of the gamma function

∞
x x
log (Γ(x)) = −γ x − log(x) + ∑ − log (1 + )+
n=1 n n

43
Now we shall differentiate with respect to x

∞ −1
1 1
ψ(x) = −γ − +∑ nx +
x n=1 1 + n n
Further simplification will result in the following

∞
1 x
ψ(x) = −γ − +∑
x n=1 n(n + x)

5.6 Some Values

Find the values of

1. ψ(1)

∞
1
ψ(1) = −γ − 1 + ∑
n=1 n(n + 1)

It should be easy to prove that

∞
1
∑ =1
n=1 n(n + 1)

Hence we have

ψ(1) = −γ

2. ψ ( 12 )

∞
1 1
ψ ( ) = −γ − 2 + ∑
2 n=1 n(2n + 1)

We need to find
∞
1
∑
n=1 n(2n + 1)

We can start by

∞
xn
∑ = − log(1 − x)
n=1 n

So we can prove easily that

44
 ∞
1
∑ = 2 − 2 log(2)
n=1 n(2n + 1)

Hence

1
ψ ( ) = −γ − 2 log(2)
2

5.7 Example

Prove that

1 1 1
∫ + dx = γ
0 log(x) 1 − x
proof

Let x = e−t

∞ 1 e−t
∫ − dt
0 et − 1 t
Let the following

∞ ts
F (s) = ∫ − ts−1 e−t dt = ζ(s + 1)Γ(s + 1) − Γ(s)
0 et − 1
Hence the limit

1 1
lim Γ(s + 1) (ζ(s + 1) − ) = lim ζ(s + 1) −
s→0 s s→0 s
Use the expansion of the zeta function

1 ∞ γn
ζ(s + 1) = +∑ (−s)n
s n=0 n!
Hence the limit is equal to γ0 = γ.

5.8 Integral representations

5.8.1 First Integral representation

∞ e−z − (1 + z)−a
ψ(a) = ∫ dz
0 z
We begin with the double integral

45
 ∞ t ∞ e−z − e−tz
∫ ∫ e−xz dx dz = ∫ dz
0 1 0 z
Using fubini theorem we also have

t ∞ t 1
∫ ∫ e−xz dz dx = ∫ dx = log t
1 0 1 x
Hence we have the following

∞ e−z − e−tz
∫ dz = log(t)
0 z
We also know that

∞
Γ′ (a) = ∫ ta−1 e−t log t dt
0

Hence we have

∞ ∞ e−z − e−tz ∞ ∞ ta−1 e−t e−z − ta−1 e−t(z+1)

Γ′ (a) = ∫ ta−1 e−t (∫ dz) dt = ∫ ∫ dz dt
0 0 z 0 0 z
Now we can use the fubini theorem

∞ ∞ ta−1 e−t e−z − ta−1 e−t(z+1)

Γ′ (a) = ∫ ∫ dt dz
0 0 z

∞ 1 −z ∞ a−1 −t ∞
Γ′ (a) = ∫ (e ∫ t e dt − ∫ ta−1 e−t(z+1) dt) dz
0 z 0 0

But we can easily deduce using Laplace that

∞
∫ ta−1 e−t(z+1) dt = Γ(a) (z + 1)−a
0

Aslo we have

∞
∫ ta−1 e−t dt = Γ(a)
0

Hence we can simplify our integral to the following

∞ e−z − (1 + z)−a
Γ′ (a) = Γ(a) ∫ dz
0 z

Γ′ (a) ∞ e−z − (1 + z)−a

= ψ(a) = ∫ dz
Γ(a) 0 z

46
5.8.2 Second Integral representation
1 1 − xs
ψ(s + 1) = −γ + ∫ dx
0 1−x
proof

This can be done by noting that

∞
s
ψ(s + 1) = −γ + ∑
n=1 n(n + s)
It is left as an exercise to prove that

∞ 1 1 − xs
s
∑ =∫ dx
n=1 n(n + s) 0 1−x

5.8.3 Third Integral representation

∞ e−t e−(at)
ψ (a) = ∫ − dt
0 t 1 − e−t
proof

Let e−t = x

1 1 xa−1
∫ − − dx
0 log(x) 1 − x
By adding and subtracting 1

1 1 1 1 1 − xa−1
−∫ + dx + ∫ dx
0 log(x) 1 − x 0 1−x
Using the second integral representation

1 1 1
−∫ + dx + γ + ψ(a)
0 log(x) 1 − x
We have already proved that

1 1 1
∫ + dx = γ
0 log(x) 1 − x
Finally we get

∞ e−t e−(at)
∫ − dt = −γ + γ + ψ(a) = ψ(a)
0 t 1 − e−t

47
5.8.4 Fourth Integral representation

Prove that

1 ∞ t
ψ(z) = log(z) − − 2∫ dt ; Rez > 0
2z 0 (t2 + z 2 )(e2π − 1)
We prove that

∞ t 1
2∫ dt = log(z) − − ψ(z)
0 (t2 + z 2 )(e2π − 1) 2z
First note that

2
= coth(πt) − 1
e2πt −1
Also note that

1 2t ∞ 1
coth(πt) = + ∑ 2 2
πt π k=1 k + t
Hence we conclude that

2t 1 2t2 ∞ 1
= − t + ∑
e2πt − 1 π π k=1 k 2 + t2
Substitute the value in the integral

∞ 1 1 2t2 ∞ 1
∫ { −t+ ∑ } dt
0 t2 +z π
2 π k=1 k 2 + t2
The first integral

1 ∞ 1 1
∫ dt =
π 0 t +z
2 2 2z
Since the second integral is divergent we put

N t 1
∫ dt = log(N 2 + z 2 ) − log(z)
0 t2 +z 2 2
Also for the series

2 N ∞ t2 N
1
∑∫ dt = ∑
π k=1 0 (t2 + z 2 )(t2 + k 2 ) k=1 k + z
Which simplifies to

48
 N N
1 1 N z N
z
∑ =∑ −∑ = HN − ∑
k=1 k + z k=1 k k=1 k(k + z) k=1 k(k + z)

Now take the limit

N
1 z
lim − log(N 2 + z 2 ) + log(z) + HN − ∑
N →∞ 2 k=1 k(k + z)
Or

∞
z
lim HN − log(N ) + log(z) − ∑
N →∞ k=1 k(k + z)
This simplifies to

∞
z 1
log(z) + γ − ∑ = log(z) − − ψ(z)
k=1 k(k + z) z
Collecting the results we have

∞ t 1 1 1
2∫ dt = log(z) + − − ψ(z) = log(z) − − ψ(z)
0 (t2 + z 2 )(e2π − 1) 2z z 2z

5.9 Gauss Digamma theorem

Let p/q be a rational number with 0 < p < q then

q/2−1
p π p 2πpk πk
ψ ( ) = −γ − log(2q) − cot ( π) + 2 ∑ cos ( ) log [sin ( )]
q 2 q k=1 q q
proof

The proof is omitted.

5.10 More results

Assume that p = 1 and q > 1 is an integer then

q/2−1
1 π π 2πk πk
ψ ( ) = −γ − log(2q) − cot ( ) + 2 ∑ cos ( ) log [sin ( )]
q 2 q k=1 q q
So for example

49
 1 1 √
ψ ( ) = (−6γ − π 3 − 9 log(3))
3 6
1 1
ψ ( ) = (−2γ − π − 6 log(2))
4 2
1 1√ 3
ψ ( ) = −γ − 3π − 2 log(2) − log(3)
6 2 2

5.11 Example

Find the following integral

∞
∫ e−at log(t) dt
0

We start by considering

∞
F (b) = ∫ e−at tb dt
0

Now use the substitution x = at we get

1 ∞ x b
F (b) = ∫ e−x ( ) dx
a 0 a
We can use the gamma function

1 ∞ b
Γ(b + 1)
−x x
F (b) = ∫ e ( ) dx =
a 0 a ab+1
Now differentiate with respect to b

1 ∞ x x b Γ(b + 1)ψ(b + 1) log(a)Γ(b + 1)

F ′ (b) = ∫ e−x log ( ) ( ) dx = −
a 0 a a ab+1 ab+1
Now put b = 0 and at = x

∞ ψ(1) − log(a) γ + log(a)

∫ e−at log(t) dt = =−
0 a a

5.12 Example

Prove the following

1 (1 − xa )(1 − xb )(1 − xc ) Γ(b + c + 1)Γ(c + a + 1)Γ(a + b + 1)

∫ dx = log { }
0 (1 − x)(− log x) Γ(a + 1)Γ(b + 1)Γ(c + 1)Γ(a + b + c + 1)
proof

50
First note that since there is a log in the denominator that gives as an idea to use differentiation under

the integral sign.

Let

1 (1 − xa )(1 − xb )(1 − xc )
F (c) = ∫ dx
0 (1 − x)(− log x)
Differentiate with respect to c

1 (1 − xa )(1 − xb )xc
F ′ (c) = ∫ dx
0 (1 − x)
By expanding

1 (1 − xa − xb + xa+b )xc 1 xc − xa+c − xb+c + xa+b+c

F ′ (c) = ∫ dx = ∫ dx
0 (1 − x) 0 (1 − x)
We can add and subtract one to use the second integral representation

1 (xc − 1) + (1 − xa+c ) + (1 − xb+c ) + (xa+b+c − 1)

F ′ (c) = ∫ dx
0 (1 − x)
Distribute the integral over the terms

1 1 − xc 1 1 − xa+c 1 1 − xb+c 1 1 − xa+b+c

F ′ (c) = − ∫ dx + ∫ dx + ∫ dx − ∫ dx
0 1−x 0 1−x 0 1−x 0 1−x
Which simplifies to

F ′ (c) = −ψ(c + 1) + ψ(a + c + 1) + ψ(b + c + 1) − ψ(a + b + c + 1)

Integrate with respect to c

F (c) = − log [Γ(c + 1)] + log [Γ(a + c + 1)] + log [Γ(b + c + 1)] − log [Γ(a + b + c + 1)] + e

Which reduces to

Γ(a + c + 1)Γ(b + c + 1)
log [ ]+e
Γ(c + 1)Γ(a + b + c + 1)
Now put c = 0 we have

Γ(a + 1)Γ(b + 1)
0 = log [ ]+e
Γ(a + b + 1)
The constant

51
 Γ(a + 1)Γ(b + 1)
e = − log [ ]
Γ(a + b + 1)
So we have the following

Γ(a + c + 1)Γ(b + c + 1) Γ(a + 1)Γ(b + 1)

F (c) = log [ ] − log [ ]
Γ(c + 1)Γ(a + b + c + 1) Γ(a + b + 1)
Hence we have the result

1 (1 − xa )(1 − xb )(1 − xc ) Γ(b + c + 1)Γ(c + a + 1)Γ(a + b + 1)

∫ dx = log { }
0 (1 − x)(− log x) Γ(a + 1)Γ(b + 1)Γ(c + 1)Γ(a + b + c + 1)

5.13 Example

Find the following integral

∞ 1 dx
∫ (e−bx − )
0 1 + ax x
Let us first use the substitution t = ax

∞ 1 dt
(e− a −
bt
∫ )
0 1+t t
Add and subtract e−t

∞ 1 dt
(e−t − e−t + e− a −
bt
∫ )
0 1+t t
Separate into two integrals

∞ e− a − e−t
bt
∞ 1 dt
−t
∫ (e − ) +∫ dt
0 1+t t 0 t
The first integral is a representation of the Euler constant when a = 1

∞ 1 dt
∫ (e−t − ) = −γ
0 1+t t
We also proved

e− a − e−t
bt
∞ b a
∫ dt = − log ( ) = log ( )
0 t a b
Hence the result

∞ 1 dx a
∫ (e−bx − ) = log ( ) − γ
0 1 + ax x b

52
5.14 Example

Find the following integral

∞ 1
∫ e−ax ( − coth(x)) dx
0 x
By using the exponential representation of the hyperbolic functions

∞ 1 1 + e−2x
∫ e−ax ( − ) dx
0 x 1 − e−2x
Now let 2x = t so we have

∞ 1 1 + e−t
e−( 2 ) ( −
at
∫ ) dt
0 t 2(1 − e−t )

e−( 2 ) e−( 2 ) + e(− 2 −t)

at at at
∞
∫ − dt
0 t 2(1 − e−t )
By adding and subtracting some terms

e−t + e−( 2 ) − e−t e−( 2 ) + e−( 2 ) − e− 2 + e−( 2 )−t

at at at at at
∞
∫ − dt
0 t 2(1 − e−t )
Separate the integrals

e−t e−( 2 ) ∞ e−( 2 ) − e(− 2 −t) ∞ e−( 2 ) − e−t

at at at at
∞
∫ − −t
dt + ∫ −t
dt + ∫ dt
0 t 1−e 0 2(1 − e ) 0 t
By using the third integral representation

e−t e−( 2 )
at
∞ a
∫ − −t
dt = ψ ( )
0 t 1−e 2
The second integral reduces to

e−( 2 ) − e−( 2 −t) ∞ e−( 2 )

at at at
∞ 1
∫ −t
= ∫ dt =
0 2(1 − e ) 0 2 a
The third integral

e−( 2 ) − e−t
at
∞ a
∫ dt = − log ( )
0 t 2
By collecting the results

∞ 1 a a 1
∫ e−ax ( − coth(x)) dx = ψ ( ) − log ( ) +
0 x 2 2 a

53
5.15 Example

Prove that

∞ x dx 1 1 a 1 a 2
∫ = ψ( + )− ψ( )− a
0 x2 + 1 sinh(ax) 2 2 2π 2 2π π
proof

∞ x dx ∞ x dx
∫ =∫
0 x2 + 1 sinh(ax) 0 x + a sinh(x)
2 2

∞ ∞ sin(xt)
=∫ ∫ e−at dt dx
0 0 sinh(x)
∞ ∞ sin(xt)
=∫ e−at ∫ dx dt
0 0 sinh(x)
π ∞ π
= ∫ e−at tanh ( t) dt
2 0 2
∞ 2
−zx
=∫ e tanh(x) dt ; z = a
0 π
∞ e−zx (1 − e−2x )
=∫ dx
0 e−2x + 1

By splitting the integral we have

∞ e−zx ∞
∫ −2x
dx = ∑ ∫ e−x(2n+z) dx
0 e +1 n≥0 0
(−1)n
=∑
n≥0 2n + z
1 1 z z
= (ψ ( + ) − ψ ( ))
4 2 4 4

∞ −e−x(z+2) (−1)n
−∫ dx = − ∑
0 e−2x + 1 n≥0 z + 2 + 2n
1 1 z z
= − (−ψ ( + ) + ψ (1 + ))
4 2 4 4

Hence we have

∞ x dx 1 1 z z z
∫ = (2ψ ( + ) − ψ (1 + ) − ψ ( ))
0 x2 + 1 sinh(ax) 4 2 4 4 4
1 1 z 1 z
= ψ( + )− ψ( )−z
2 2 4 2 4
1 1 a 1 a 2
= ψ( + )− ψ( )− a
2 2 2π 2 2π π

54
Let a = π/2

∞ x dx 1 1 1 1 1
∫ = ψ( + )− ψ( )−1
0 x2 + 1 sinh( π2 x) 2 2 4 2 4
π
= cot(π/4) − 1
2
π
= −1
2

55
6 Zeta function

Zeta function is one of the most important mathematical functions. The study of zeta function isn’t

exclusive to analysis. It also extends to number theory and the most celebrating theorem of Riemann.

6.1 Definition
∞
1
ζ(s) = ∑ s
n=1 n

6.2 Bernoulli numbers

We define the Bernoulli numbers Bk as

∞
x Bk k
= ∑ x
e − 1 k=0 k!
x

Now let us derive some values for the Bernoulli numbers , rewrite the power series as

∞
Bk k
x = (ex − 1) ∑ x
k=0 k!

By expansion

1 2 1 3 1 4 B2 2 B3 3
x = (x + x + x + x + ⋯) ⋅ (B0 + B1 x + x + x + ⋯)
2! 3! 4! 2! 3!
By multiplying we get

B0 2 B0 B2 B1 3 B0 B1 B2 B3 4
x = B0 x + (B1 + )x + ( + + )x + ( + + + )x + ⋯
2! 3! 2! 2! 4! 3! 2! 2! 3!
By comparing the terms we get the following values

1 1 1
B0 = 1 , B 1 = − + B2 = , B 3 = 0 , B 4 = − , ⋯
2 6 30
Actually we also deduce that

B2k+1 = 0 , ∀ k ∈ Z+

56
6.3 Relation between zeta and Bernoulli numbers

According to Euler we have the following relation

22k−1 2k
ζ(2k) = (−1)k−1 B2k π
(2k)!
proof

We start by the product formula of the sine function

sin(z) ∞ z2
= ∏ (1 − 2 2 )
z n=1 n π
Take the logarithm to both sides

∞
z2
log(sin(z)) − log(z) = ∑ log (1 − )
n=1 n2 π 2
By differentiation with respect to z

∞ z
1 2 2
cot(z) − = −2 ∑ n πz2
z n=1 1 − n2 π 2

By simple algebraical manipulation we have

∞
z2 ⎛ 1 ⎞
z cot(z) = 1 − 2 ∑ 2 π2 ⎝ z2 ⎠
n=1 n 1 − π 2 n2
Now using the power series expansion

∞
1 1
z2
=∑ z 2k , ∣z∣ < π n
1− π 2 n2 k=0 n 2k π 2k

z2 ⎛ 1 ⎞ ∞ 1 ∞
1
= ∑ 2k+2 2k+2 z 2k+2 = ∑ 2k 2k z 2k
n π ⎝ 1 − π2 n2 ⎠ k=0 n
2 2 z 2
π k=1 n π

So the sums becomes

∞ ∞
1 z 2k
z cot(z) = 1 − 2 ∑ ∑ 2k π 2k
n=1 k=1 n

Now if we invert the order of summation we have

∞ ∞ ∞
1 z 2k ζ(2k) 2k
z cot(z) = 1 − 2 ∑ ∑ 2k π 2k
= 1 − 2 ∑ 2k
z
k=1 n=1 n k=1 π

Euler didn’t stop here, he used power series for z cot(z) using the Bernoulli numbers.

Start by the equation

57
 ∞
x Bk k
= ∑ x
e − 1 k=0 k!
x

By putting x = 2iz we have

∞
2iz Bk k
= ∑ (2iz)
e2iz − 1 k=0 k!
Which can be reduced directly to the following by noticing that B2k+1 = 0

∞
22k 2k
z cot(z) = 1 − ∑ (−1)k−1 B2k z
k=1 (2k)!
The result is immediate by comparing the two different representations.

6.4 Exercise

Find the values of

ζ(4) , ζ(6) , B5 , B6

6.5 Integral representation

1 ∞ ts−1
ζ(s) = ∫ dt
Γ(s) 0 et − 1
proof

Start by the integral representation

∞ e−t ts−1
∫ dt
0 1 − e−t
Using the power expansion

∞
1
= ∑ e−nt
1 − e−t n=0
Hence we have

∞ ∞
∫ e−t ts−1 ( ∑ e−nt ) dt
0 n=0

By swapping the series and integral

∞ ∞ ∞
1
∑∫ ts−1 e−(n+1)t dt = Γ(s) ∑ = Γ(s)ζ(s)
n=0 (n + 1)
0 s
n=0

58
6.6 Hurwitz zeta and polygamma functions

Hurwitz zeta is a generalization of the zeta function by adding a parameter .

6.6.1 Definition
∞
1
ζ(a, z) = ∑ ; ζ(a, 1) = ζ(a)
n=0 (n + z)a
Let us define the polygamma function as the function produced by differentiating the Digamma function

and it is often denoted by

ψn (z) ∀ n ≥ 0

We define the digamma function by setting n = 0 so it’s denoted by ψ0 (z) .

Other values can be found by the following recurrence relation

ψn′ (z) = ψn+1 (z)

So we have

ψ1 (z) = ψ0′ (z)

6.6.2 Relation between zeta and polygamma

∀n≥1

ψn (z) = (−1)n+1 n! ζ(n + 1, z)

22n−2 2n
ψ2n−1 (1) = (−1)n−1 B2n π
n
proof

We have already proved the following relation

1 ∞ z
ψ0 (z) = −γ − +∑
z n=1 n(n + z)
This can be written as the following

∞
1 1
ψ0 (z) = −γ + ∑ −
k=0 k + 1 k+z
By differentiating with respect to z

59
 ∞
1
ψ1 (z) = ∑
k=0 (k + z)
2

∞
1
ψ2 (z) = −2 ∑
k=0 (k + z)3

∞
1
ψ3 (z) = 2 ⋅ 3 ∑
k=0 (k + z)
4

∞
1
ψ4 (z) = −2 ⋅ 3 ⋅ 4 ∑
k=0 (k + z)
5

Continue like that to obtain

∞
1
ψn (z) = (−1)n+1 n! ∑
k=0 (k + z)n+1
We realize the RHS is just the Hurwitz zeta function

ψn (z) = (−1)n+1 n! ζ(n + 1, z)

By setting z = 1 we have an equation in terms of the ordinary zeta function

ψn (1) = (−1)n+1 n! ζ(n + 1)

Now since we already proved in the preceding section that

22k−1 2k
ζ(2k) = (−1)k−1 B2k π
(2k)!
we can easily verify the following

22n−1 2n 22n−2 2n
ψ2n−1 (1) = (2n − 1)! (−1)n−1 B2n π = (−1)n−1 B2n π
(2n)! n
This can be used to deduce some values for the polygamma function

π2 π4
ψ1 (1) = , ψ3 (1) =
6 15

Other values can be evaluated in terms of the zeta function

ψ2 (1) = −2ζ(3) , ψ4 (1) = −24ζ(5)

60
6.7 Example

Prove that

π
2 π π3
∫ x sin(x) cos(x) log(sin x) log(cos x) dx = −
0 16 192
proof

Start by the transformation x → π

2
−x

π π
2 π 2
∫ x sin(x) cos(x) log(sin x) log(cos x) dx = ∫ sin(x) cos(x) log(sin x) log(cos x) dx
0 4 0
We need to find

π
2
∫ sin(x) cos(x) log(sin x) log(cos x) dx
0

Let us start by the following

π
2 Γ(a)Γ(b)
F (a, b) = 2 ∫ sin2a−1 (x) cos2b−1 (x) dx =
0 Γ(a + b)
Now let us differentiate with respect to a

Γ(a)Γ(b) (ψ0 (a) − ψ0 (a + b))

π
∂ 2
(F (a, b)) = 4 ∫ sin2a−1 (x) cos2b−1 (x) log(sin x) dx =
∂a 0 Γ(a + b)
Differentiate again but this time with respect to b

π
∂ 2
(Fa (a, b)) = 8 ∫ sin2a−1 (x) cos2b−1 (x) log(sin x) log(cos x)dx
∂b 0

Γ(a)Γ(b) (ψ02 (a + b) + ψ0 (a)ψ0 (b) − ψ0 (a)ψ0 (a + b) − ψ0 (b)ψ0 (a + b) + ψ1 (a + b))

=
Γ(a + b)
Putting a = b = 1 we have the following

ψ02 (2) + ψ02 (1) − ψ0 (1)ψ0 (2) − ψ0 (1)ψ(2) − ψ1 (2)

π
2
∫ sin(x) cos(x) log(sin x) log(cos x)dx =
0 8

By simple algebra we arrive to

(ψ0 (2) − ψ0 (1))2 − ψ1 (2)

π
2
∫ sin(x) cos(x) log(sin x) log(cos x)dx =
0 8
We already know that ψ0 (1) = −γ and ψ0 (2) = 1 − γ

Now to evaluate ψ1 (2) , we have to use the zeta function we have already established the following

relation

61
 ∞
1
ψ1 (z) = ∑
k=0 (n + z)
2

Now putting z = 2 we have the following

∞
1
ψ1 (2) = ∑
k=0 (k + 2)2
Let us write the first few terms in the expansion

∞
1 1 1 1
∑ = 2 + 2 + 2 +⋯
k=0 (k + 2) 2 2 3 4
we see this is similar to ζ(2) but we are missing the first term

π2
ψ1 (2) = ζ(2) − 1 = −1
6
Collecting all these results together we have

π
2 1 π2
∫ sin(x) cos(x) log(sin x) log(cos x)dx = −
0 4 48
Finally we get our result

π
2 π π3
∫ x sin(x) cos(x) log(sin x) log(cos x) dx = −
0 16 192

62
7 Dirichlet eta function

Dirichlet eta function is the alternating form of the zeta function.

7.1 Definition
∞
(−1)n−1
η(s) = ∑
n=1 ns
The alternating form of the zeta function is easier to compute once we have established the main results

of the zeta function because the alternating form is related to the zeta function through the relation

7.2 Relation to Zeta function

η(s) = (1 − 21−s ) ζ(s)

proof

We will start by the RHS

(1 − 21−s ) ζ(s) = ζ(s) − 21−s ζ(s)

Which can be written as sums of series

∞
1 1 ∞ 1
∑ s
− ∑
n=1 n 2s−1 n=1 ns

∞ ∞
1 1
∑ − 2 ∑
n=1 n s
n=1 (2n) s

Clearly we can see that we are subtracting even terms twice , this is equivalent to

∞ ∞
1 1
∑ −∑
n=1 (2n − 1) s
n=1 (2n) s

This looks easier to understand if we write the terms

1 1 1 1 1
(1 + s
+ s + ⋯) − ( s + s + s + ⋯)
3 5 2 4 6
Rearranging the terms we establish the alternating form

∞
1 1 1 (−1)n−1
1− + − + ⋯ = ∑ = η(s)
2s 3s 4s n=1 ns

63
7.3 Integral representation
∞ ts−1
η(s) Γ(s) = ∫ dt
0 et + 1
proof

Start by the RHS

∞ ts−1 ∞ e−t ts−1

∫ dt = ∫ dt
0 e +1
t 0 1 + e−t
Now using the power expansion we arrive to

∞ ∞
∫ e−t ts−1 dt ( ∑ (−1)n e−nt )
0 n=0

∞ ∞
∑ (−1) ∫
n
e−(n+1)t ts−1 dt
n=0 0

Using Laplace transform we can solve the inner integral

∞ Γ(s)
∫ e−(n+1)t ts−1 dt =
0 (n + 1)s
Hence we have the following

∞ ∞
(−1)n (−1)n−1
Γ(s) ∑ = Γ(s) ∑ = Γ(s) η(s)
n=0 (n + 1)
s ns
n=1

An easy result of the above integral

∞ t π2
∫ dt = Γ(2) η(2) =
0 et +1 12

64
8 Polylogarithm

8.1 Definition
∞
zk
Lin (z) = ∑ n
k=1 k

The name contains two parts, (poly) because we can choose different n and produce many functions and

(logarithm) because we can express Li1 (z) = − log(1 − z).

8.2 Relation to other functions

We can relate it to the Zeta function

∞
1
Lin (1) = ∑ n
= ζ(n)
k=1 k

In particular we have for n = 2

π2
Li2 (1) = ζ(2) =
6
Also we can relate it to the eta function though z = −1

∞
(−1)k
Lin (−1) = ∑ n
= −η(n)
k=1 k

Also we can relate it to logarithms by putting n = 1

∞
zk
Li 1 (z) = ∑
n=1 k

The power expansion on the left is famous

∞
zk
∑ = − log(1 − z)
k=1 k

8.3 Integral representation

z Li n (t)
Li n+1 (z) = ∫ dt
0 t
proof

Using the series representation we have

z 1 ∞ tk ∞
1 z ∞
zk
∫ ( ∑ n ) dt = ∑ n ∫ tk−1 dt = ∑ n+1 = Li n+1 (z)
0 t k=1 k k=1 k 0 k=1 k

65
8.4 Square formula

Li n (−z) + Li n (z) = 21−n Li n (z 2 )

proof

As usual we write the series representation of the LHS

∞
z k ∞ (−z)k
∑ n
+∑ n
k=1 k k=1 k

Listing the first few terms

z2 z3 z2 z3
z+ + + ⋯ + (−z + − + ⋯)
2n 3n 2n 3n
The odd terms will cancel

z2 z4 z6
2 + 2 + 2 +⋯
2n 4n 6n
Take 21−n as a common factor

∞
(z 2 )2 (z 2 )3 (z 2 )k
21−n (z 2 + + + ⋯) = 2 1−n
∑ = 21−n Li n (z 2 )
2n 3n k=1 k n

8.5 Exercise

Prove that

z log(1 − t)
Li 2 (z) = − ∫ dt
0 t

66
8.6 Dilogarithms

Of all polylogarithms Li2 (z) is the most interesting one, in this section we will see why!

8.6.1 Definition
∞ z log(1 − t)
zk
Li2 (z) = ∑ 2
= −∫ dt
k=1 k 0 t
The curious reader should try to prove the integral representation using the recursive definition we

introduced in the previous section .

8.6.2 First functional equation

−1 1 π2
Li2 ( ) + Li2 (−z) = − log2 (z) −
z 2 6
proof

We will start by the following

−1 log(1 − t)
−1
z
Li2 ( ) = −∫ dt
z 0 t
Differentiate with respect to z

1 ⎛ log (1 + z ) ⎞ log (1 + z ) log(1 + z) − log(z)

1 1
d −1
Li2 ( ) = 2 − −1
= =
dz z z ⎝ z
⎠ z z
Now integrate with respect to z

−1 −z log(1 − t) 1 1
Li2 ( )=∫ dt − log2 (z) + C = −Li2 (−z) − log2 (z) + C
z 0 t 2 2
To find the constant C let z = 1

C = 2Li2 (−1)

Now we must be aware that

−π 2
C = 2Li2 (−1) = −2η(2) =
6
Which proves the result by simple rearrangement

−1 1 π2
Li2 ( ) + Li2 (−z) = − log2 (z) −
z 2 6

67
8.6.3 Second functional equation

π2
Li2 (z) + Li2 (1 − z) = − log(z) log(1 − z) , 0 < z < 1
6
proof

Start by the following

z log(1 − t)
Li2 (z) = − ∫ dt
0 t
Now integrate by parts to obtain

z log(t)
Li2 (z) = − ∫ dt − log(z) log(1 − z)
0 1−t
By the change of variable t = 1 − x we get

z log(t) 1−z log(1 − x)

∫ dt = − ∫ dx
0 1−t 1 x
For 0 < z < 1

1 log(1 − x) 1 log(1 − x) 1−z log(1 − x)

∫ dx = ∫ dx − ∫ dx
1−z x 0 x 0 x
Now it is easy to see that

1 log(1 − x)
∫ dx = −Li2 (1) + Li2 (1 − z)
1−z x
Which implies that

Li2 (z) = Li2 (1) − Li2 (1 − z) − log(z) log(1 − z)

Li2 (z) + Li2 (1 − z) = Li2 (1) − log(z) log(1 − z)

π2
Now since Li2 (1) = ζ(2) = 6

π2
Li2 (z) + Li2 (1 − z) = − log(z) log(1 − z)
6
We can easily deduce that for z = 1
2

1 π2 1
2Li2 ( ) = − log2 ( )
2 6 2

1 π2 1 1
Li2 ( ) = − log2 ( )
2 12 2 2

68
8.6.4 Third functional equation

z 1
Li2 (z) + Li2 ( ) = − log2 (1 − z) z < 1
z−1 2
proof

Start by the following

z−1 log(1 − t)
z
z
Li2 ( ) = −∫ dt
z−1 0 t
Differentiate both sides with respect to z

d z 1 ⎛ log (1 − z−1
z
)⎞
Li2 ( )=
dz z−1 (z − 1) ⎝
2 z
z−1 ⎠
Upon simplification we obtain

d z − log(1 − z)
Li2 ( )=
dz z−1 z(z − 1)
Using partial fractions decomposition

d z log(1 − z) log(1 − z)
Li2 ( )= +
dz z−1 1−z z
Integrate both sides with respect to z

z 1
Li2 ( ) = − log2 (1 − z) − Li2 (z) + C
z−1 2
Put z = −1 to find the constant

1 1
Li2 ( ) = − log2 (2) − Li2 (−1) + C
2 2
Remember that

1 π2 1 1 π2
Li2 ( ) = − log2 ( ) , Li2 (−1) = −
2 12 2 2 12
Hence we deduce that C = 0

z 1
Li2 ( ) = − log2 (1 − z) − Li2 (z)
z−1 2
Which can be written as

z 1
Li2 ( ) + Li2 (z) = − log2 (1 − z)
z−1 2

69
8.6.5 Example

Prove that

√ √
5−1 π2 5−1
Li2 ( )= − log (
2
)
2 10 2
proof

First we add the two functional equations of this section to obtain

z 1 1
Li2 ( ) + Li2 (z 2 ) − Li2 (−z) = − log2 (1 − z)
z−1 2 2
√
Now let z = 1− 5
2

√ √ √ √
1−2 5+5 3− 5 z 5−1 3− 5
z2 = = Ô⇒ = √ =
4 2 z−1 1+ 5 2
Hence we have

√ √ √
3 3− 5 5−1 1 5+1
Li2 ( ) − Li2 ( ) = − log (
2
)
2 2 2 2 2
We already established the following functional equation

π2
Li2 (z) + Li2 (1 − z) = − log(z) log(1 − z)
6
√
Put z = 3− 5
2

√ √ √ √
3− 5 5−1 π2 3− 5 5−1
Li2 ( ) + Li2 ( )= − log ( ) log ( )
2 2 6 2 2
√
Solving the two representations for Li2 ( 5−1
2
) we get our result.

8.6.6 Example

Find the following integral

1 log(1 − x) log(x)
I =∫ dx
0 x
Integrate by parts

∞
1 Li2 (t) 1
I = − log(x)Li2 (x)∣10 + ∫ dt = Li3 (1) = ∑ 3 = ζ(3)
0 t k=1 k

70
8.6.7 Example

Evaluate the following integral

x log2 (1 − t)
∫ dt , 0 < x < 1
0 t
Integrating by parts we get the following

x log2 (1 − t) x Li (t)
2
∫ dt = − log(1 − x)Li2 (x) − ∫ dt
0 t 0 1−t
Now we are left with the following integral

x Li2 (t) 1 Li (1 − t)
2
∫ dt = ∫ dt
0 1−t 1−x t
Using the first functional equation

π2
1
6
− Li2 (t) − log(1 − t) log(t)
∫ dt
1−x t

π2 1 Li (t) 1 log(1 − t) log(t)

2
− log(1 − x) − ∫ dt − ∫ dt
6 1−x t 1−x t
The first integral

1 Li2 (t)
∫ dt = Li3 (1) − Li3 (1 − x)
1−x t
The second integral is the same as the first exercise

1 log(1 − t) log(t)
∫ = Li3 (1) + log(1 − x)Li2 (1 − x) − Li3 (1 − x)
1−x t
Collecting the results together we obtain

x Li2 (t) π2
∫ dt = − log(1 − x) − Li2 (1 − x) log(1 − x) + 2 Li3 (1 − x) − 2ζ(3)
0 1−t 6
Finally we have

x log2 (1 − t) π2
∫ dt = − log(1 − x)Li2 (x) + log(1 − x) + Li2 (1 − x) log(1 − x) − 2 Li3 (1 − x) + 2ζ(3)
0 t 6

71
8.6.8 Example

Find the following integral

a x
I(a) = ∫ dx
0 ex − 1
1
Start by the power expansion of 1−e−x

a a ∞
x
I(a) = ∫ dx = ∫ xe−x ∑ e−nx
0 ex −1 0 n=0

By swapping the integral and the summation

∞ a
I(a) = ∑ ∫ xe−x(n+1) dx
n=0 0

The integral could be solved by parts

∞
1 e−(n+1)a ae−(n+1)a
I(a) = ∑ − −
n=0 (n + 1) (n + 1)2 (n + 1)
2

Distribute the summation to obtain

I(a) = ζ(2) + a log(1 − e−a ) − Li2 (e−a )

72
9 Ordinary Hypergeometric function

Ordinary or sometimes called the Gauss hypergoemtric function is a generalization of the power expan-

sion definition. Before we start with the definition we will explain some notations.

9.1 Definition

Define the raising factorial as follows

⎧
⎪
⎪
⎪
⎪ 1 ∶n=0
(z)n = ⎨
⎪
⎪
⎪ Γ(z+n)
⎪ ∶n>0
⎩ Γ(z)
Using this definition have

∞
(a)n (b)n z n
2 F1 (a, b; c; z) =∑
n=0 (c)n n!

9.2 Some expansions using the hypergeomtric function

We can represent famous functions using the hypergeomtric function

1. Logarithm
∞ ∞ ∞
(1)n (1)n (−z)n n! z n+1
z 2 F1 (1, 1; 2; −z) = z ∑ = ∑ (−1)n z n+1 = ∑ (−1)n = log(1 + z)
n=0 (2)n n! n=0 (n + 1)! n=0 n+1

2. Power function
∞ ∞
(a)n (1)n z n (a)n n
2 F1 (a, 1; 1; z) =∑ =∑ z = (1 − z)−a
n=0 (1)n n! n=0 n!

3. Sine inverse

∞ ( 12 )n ( 12 )n z 2n+1 ∞( 12 )n z 2n+1 ∞
1 ⋅ 3 ⋅ 5⋯(2n − 1) 2n+1
z 2 F1 ( 21 , 12 ; 32 ; z 2 ) =∑ =∑ =∑ z
n=0 ( 32 )n n! n=0 2n + 1 n! n=0 2 n! (2n + 1)
n

Which can be written as

∞
(2n)! ∞ (2n )
z 2 F1 ( 21 , 12 ; 32 ; z 2 ) = ∑ z 2n+1
= ∑
n
z 2n+1 = arcsin(z)
n=0 4 n (n!) 2 (2n + 1) n=0 4 n (2n + 1)

Now we consider converting the Taylor expansion into the equivalent hypergeomtric representation

Suppose the following

73
 ∞
2 F1 (a, b; c; z) = ∑ tk z k , t 0 = 1
k=0

Now consider the ratio

tk+1 (k + a)(k + b)
= z
tk (k + c)(k + 1)
Using this definition, we can easily find the terms a, b, c.

Let us consider some examples

1. Exponential function

f (z) = ez

The power expansion is

∞
zk
f (z) = ez = ∑
k=0 k!

Hence we have

tk+1 z
=
tk k+1

Comparing to our representation we conclude

ez = 2 F1 (−, −; −; z)

2. Cosine function

∞
(−1)k z 2k
f (z) = cos(z) = ∑
k=0 (2k)!

By the same approach

tk+1 1 1 −z 2
=− z=
tk (2k + 2)(2k + 1) (k + 1) (k + 21 ) 4

Hence we have

1 −z 2
cos(z) = 2 F1 (−, −; ; )
2 4

74
 3. Power function

∞
(a)k k
f (z) = (1 − z)−a = ∑ z
k=0 k!

By the same approach

tk+1 (k + a) (k + a)(k + 1)
= z= z
tk (k + 1) (k + 1)(k + 1)

Hence we have

(1 − z)−a = 2 F1 (a, 1; 1; z)

9.3 Exercise

Find the hypergeoemtric representations of the following functions

arcsin(z), sin(z)

9.4 Integral representation

1 tb−1 (1 − t)c−b−1
β(c − b, b) 2 F1 (a, b; c; z) = ∫ dt
0 (1 − tz)a
proof

Start by the RHS

1
∫ tb−1 (1 − t)c−b−1 (1 − tz)−a dt
0

Using the expansion of (1 − tz)−a we have

∞
1 (a)k
∫ tb−1 (1 − t)c−b−1 ∑ (tz)k
0 k=0 k!
Interchanging the integral with the series

∞
(a)k k 1
∑ z ∫ tk+b−1 (1 − t)c−b−1 dt
k=0 k! 0

Recalling the beta function we have

∞
(a)k Γ(k + b)Γ(c − b) z k
∑
n=0 Γ(k + c) k!

75
Using the identity that

Γ(b)Γ(c − b)
β(c − b, b) =
Γ(c)
and

Γ(z + k)
= (z)k
Γ(z)
We deduce that

∞ ∞
(a)k Γ(k + b)Γ(c) z k (a)k (b)k z k
β(c − b, b) ∑ = β(c − b, b) ∑
k=0 Γ(b)Γ(k + c) k! k=0 (c)k k!

9.5 Transformations

1. Pfaff transformations

z
2 F1 (a, b; c; z) = (1 − z)−a 2 F1 (a, c − b; c; )
z−1

and

z
2 F1 (a, b; c; z) = (1 − z)−b 2 F1 (c − a, b; c; )
z−1

proof

Start by the integral representation

1 1 tb−1 (1 − t)c−b−1
2 F1 (a, b; c; z) = ∫ dt
β(b, c − b) 0 (1 − tz)a

By the the transformation t → 1 − t

1 (1 − t)b−1 tc−b−1 1 (1 − t)b−1 tc−b−1

∫ dt = ∫ dt
0 (1 − (1 − t)z) a 0 (1 − z + tz)a

Which can be written as

(1 − z)−a 1 z −a
∫ t c−b−1
(1 − t) b−1
(1 − t ) dt
β(b, c − b) 0 z−1

Note this is the integral representation of

76
 z
(1 − z)−a 2 F1 (a, c − b; c; )
z−1

Also using that

2 F1 (a, b; c; z) = 2 F1 (b, a; c; z)

We deduce that

z
2 F1 (a, b; c; z) = (1 − z)−b 2 F1 (c − a, b; c; )
z−1

2. Euler transformation

2 F1 (a, b; c; z) = (1 − z)c−a−b 2 F1 (c − a, c − b; c; z)

proof

In the Pfaff transformations let z → z

z−1

z
2 F1 (a, b; c; ) = (1 − z)−a 2 F1 (a, c − b; c; z)
z−1

and

z
2 F1 (a, b; c; ) = (1 − z)−b 2 F1 (c − a, b; c; z)
z−1

By equating the two transformations

(1 − z)−a 2 F1 (a, c − b; c; z) = (1 − z)−b 2 F1 (c − a, b; c; z)

Now use the transformation b → c − b

(1 − z)−a 2 F1 (a, b; c; z) = (1 − z)c−b 2 F1 (c − a, c − b; c; z)

Which can be reduced to

2 F1 (a, b; c; z) = (1 − z)c−a−b 2 F1 (c − a, c − b; c; z)

77
 3. Quadratic transformation

a a 1 z2
= (1 − z)− 2 2 F1 ( , b − ; b + ;
a
2 F1 (a, b; 2b; z) )
2 2 2 4z − 4

4. Kummer

2 F1 (a, b; c; z) = 2 F1 (a, b; 1 + a + b − c; 1 − z)

9.6 Special values

1. At z = 1
Γ(c)Γ(c − a − b)
2 F1 (a, b; c; 1) =
Γ(c − a)Γ(c − a)

Start by the integral representation at z = 1

1 1
2 F1 (a, b; c; 1) = ∫ t (1 − t)
b−1 c−b−a−1
dt
β(c − b, b) 0

Now we can use the first integral representation of the beta function

1 Γ(b)Γ(c − b − a)
2 F1 (a, b; c; 1) = ⋅
β(c − b, b) Γ(c − a)

Which could be simplified to

Γ(c) Γ(b)Γ(c − b − a) Γ(c)Γ(c − a − b)

2 F1 (a, b; c; 1) = ⋅ =
Γ(b)Γ(c − b) Γ(c − a) Γ(c − a)Γ(c − a)

2. At z = −1

Γ(1 + a − b)Γ (1 + a2 )
2 F1 (a, b; 1 + a − b; −1) =
Γ(1 + a)Γ(1 + a
2
− b)

78
10 Error Function

The error function is an interesting function that has many applications in probability, statistics and

physics.

10.1 Definition
2 x
erf(x) = √ ∫ e−t dt
2

π 0

10.2 Complementary error function

erfc(x) = 1 − erf(x)

10.3 Imaginary error function

erfi(x) = −ierf(ix)

10.4 Properties

1. The error function is odd

2 −x 2 x
e−t dt = − √ ∫ e−t dt = −erf(x)
2 2
erf(−x) = √ ∫
π 0 π 0

2. Real part and imaginary parts

erf(z) + erf(z̄)
R erf(z) =
2

erf(z) − erf(z̄)
I erf(z) =
2i

Using complex variables it can be done using erf(z̄) = erf(z)

10.5 Relation to other functions

1. Hypergeomtric function

2x 1 3
erf(x) = √ 1 F1 ( , , −x2 )
π 2 2

79
 proof

By expanding the hypergeometirc function

2x ∞ ( ) (−x2 )k
1
2x 1 3
√ 1 F1 ( , , −x2 ) = √ ∑ 23 k
π 2 2 π k=0 ( 2 ) k!
k

Which can be simplified to

2x 1 3 x ∞ (−x2 )k
√ 1 F1 ( , , −x2 ) = √ ∑ 1
π 2 2 π k=0 ( 2 + k)k!

Notice that this is actually the expanded error function

2 x 2 x ∞ (−x2 )k 2 ∞ (−x)2k+1
√ ∫ e−t dt = √ ∫ ∑
2
dt = √ ∑
π 0 π 0 k=0 k! π k=0 (2k + 1)k!

2. Incomplete Gamma function

Γ ( 21 , x2 )
erf(x) = 1 − √
π

proof

By the definition of the incomplete gamma function

1 ∞
t− 2 e−t dt
1
Γ ( , x2 ) = ∫
2 x 2

Let t = y 2

1 ∞
e−y dy
2
Γ ( , x2 ) = 2 ∫
2 x

We need to get the interval (0, x)

1 ∞ x
e−y dy − 2 ∫ e−y dy
2 2
Γ ( , x2 ) = 2 ∫
2 0 0

We have already proved that

∞ √
e−y dy =
2
2∫ π
0

Hence we have using the definition of the error function

80
 1 √ √
Γ ( , x2 ) = π − πerf(x)
2

By rearrangements we get our result.

10.6 Example

Find the integral

x 2
I =∫ et dt
0

The function has no elementary anti-derivative so we represent it using the error function.

Consider the imaginary error function

2 ix
e−t dt
2
erfi(x) = −i √ ∫
π 0
By differentiating both sides we have

d 2 2
erfi(x) = √ ex
dx π
Hence we have

√
2 d π
ex = erfi(x)
dx 2
By integrating both sides we have

x
√
2 π
∫ et dt = erfi(x)
0 2

10.7 Example

Prove that

∞ 1
∫ erfc(x) dx = √
0 π
proof

Using the complementary error function

∞
∫ (1 − erf(x)) dx
0

81
Integrating by parts we have

2 ∞
∞
xe−x dx
2
I = x(1 − erf(x)) ]0 + √ ∫
π 0
Now we compute erf(∞)

∞
√
2 −t2 2 π
erf(∞) = √ ∫ e dt = √ × =1
π 0 π 2
So the first term will go to zero. The integral can be solved by substitution

2 ∞ 1
xe−x dx = √
2
I=√ ∫
π 0 π

10.8 Example

Prove that

√
∞ 2− 2
2
∫ erfc (x) dx = √
0 π
proof

Integrate by parts

∞
∞
I = xerfc2 (x) ]0 − 2 ∫ xerfc′ (x)erfc(x) dx
0

The first integral goes to 0

∞
I = −2 ∫ x erfc′ (x)erfc(x) dx
0

The derivative of the complementary error function

2
erfc′ (x) = (1 − erf(x))′ = − √ e−x
2

π
That results in

4 ∞
xe−x erfc(x) dx
2
I=√ ∫
π 0
Integrate by parts again

2 4 ∞
∞
I = √ e−x erfc(x) ]0 − ∫ e−2t dt
2 2

π π 0

82
At infinity the integral goes to 0. At 0 we get

2 2 2
√ e−x erfc(x)]x=0 = √ (1 − erf(0)) = √
2

π π π
The integral can be evaluated to

∞
√
−2t2 π
∫ e dt = √
0 2 2
Collecting the results together we have

√ √
2 π 4 2− 2
I=√ − √ × = √
π 2 2 π π

10.9 Example

Prove that

√
∞ π − 2 coth−1 2
∫ sin(x )erfc(x) dx =
2
√
0 4 2π
proof
√
Using the substitution x = t

1 ∞ √
sin(t)t− 2 erfc( t) dt
1
∫
2 0
Consider the function

1 ∞ √
sin(t)t− 2 erfc(a t) dt
1
I(a) = ∫
2 0
Differentiating with respect to a we have

−1 ∞ −1 1
I ′ (a) = √ ∫ sin(t)e−a t dt = √ ⋅ 4
2

π 0 π a +1
Now integrating with respect to a

−1 a dx
I(a) = √ ∫ +C
π 0 x4 + 1
To evaluate the constant we take a → ∞

−1 ∞ dx
I(∞) = √ ∫ +C
π 0 x4 + 1

83
The function has an anti-derivative and the value is

∞ dx
√
−1 π
√ ∫ =− √
π 0 x +1
4
2 2
Note that

√
π
erfc(∞) = 0 Ô⇒ C= √
2 2
Finally we get

√
−1 a dx π
I(a) = √ ∫ + √
π 0 x4 + 1 2 2
Let a = 1 in the integral

∞
√ 1 dx
π 1
I(1) = ∫ sin(x2 )erfc(x) dx = √ − √ ∫
0 2 2 π 0 x4 + 1
Also knowing that

√
1 1 dx π + 2 coth−1 2
√ ∫ = √
π 0 x4 + 1 4 2π
Hence we have the result

√
∞ π − 2 coth−1 2
∫ sin(x )erfc(x) dx =
2
√
0 4 2π

10.10 Exercise

Can you find closed forms for

∞
∫ erfc3 (x) dx =?
0

∞
∫ erfc4 (x) dx =?
0

What about

∞
∫ erfcn (x) dx =?
0

84
11 Exponential integral function

11.1 Definition
∞ e−t ∞ e−xt
E(x) = ∫ dt = ∫ dt
x t 1 t

11.2 Example

Prove that

lim [log(x) + E(x)] = −γ

x→0

proof

Integration by parts for E(x)

∞
∞
E(x) = e−t log(t) ]x + ∫ log(t)e−t dt
x

The limit at infinity goes to zero

∞
E(x) = −e−x log(x) + ∫ log(t)e−t dt
x

Hence by taking the limit

∞
lim [log(x) + E(x)] = lim (log(x) − e−x log(x)) + ∫ log(t)e−t dt
x→0 x→0 0

The first limit goes to 0

∞
lim [log(x) + E(x)] = ∫ log(t)e−t dt = ψ(1) = −γ
x→0 0

11.3 Example

Prove that for p > 0

∞ Γ(p)
∫ xp−1 E(ax) dx =
0 pap
proof

Integrating by parts we have

85
 ∞ 1 1 ∞
∞
∫ xp−1 E(ax) dx = xp E(ax) ]0 + ∫ xp−1 e−ax dx
0 p ap 0
The first limit goes to 0

1 ∞ 1 ∞ Γ(p)
∫ xp−1 e−ax dx = p ∫ xp−1 e−x dx =
ap 0 pa 0 p ap

11.4 Example

Prove the general case

∞ π Γ(p)
∫ xp−1 eax E(ax) dx = ⋅
0 sin(aπ) ap
proof

Switch to the integral representation

∞ ∞ e−t
∫ xp−1 eax ∫ dt dx
0 ax t
Use the substitution t = ax y

∞ ∞ e−ax(y−1)
∫ ∫ xp−1 dy dx
0 1 y
By switching the two integrals

∞ 1 ∞
∫ ∫ xp−1 e−ax(y−1) dx dy
1 y 0
By the Laplace identities

Γ(p) ∞ 1
∫ dy
ap 1 y(y − 1)p
Now let y = 1/x

Γ(p) 1 p−1 −p
∫ x (1 − x) dx
ap 0
Using the reflection formula for the Gamma function

Γ(p) 1 p−1 −p π Γ(p)

p ∫ x (1 − x) dx = ⋅ p
a 0 sin(aπ) a

86
11.5 Example

Prove that

∞ π2
∫ ez E 2 (z) dz =
0 6
proof

Using the integral representation

∞ ∞ e−xz e−yz
E 2 (z) = ∫ ∫ dx dy
1 1 xy

∞ ∞ ∞ ∞ e−z(x+y−1)
∫ ez E 2 (z) dz = ∫ ∫ ∫ dx dy dz
0 0 1 1 xy
Swap the integrals

∞ 1 ∞ 1 ∞
∫ ∫ ∫ e−z(x+y−1) dz dx dy
1 y 1 x 0

∞ 1 ∞ 1
∫ ∫ dx dy
1 y 1 x(x + y − 1)
The inner integral is an elementary integral

∞ 1 log(y)
∫ dx = −
1 x(x + y − 1) 1−y
The integral becomes

∞ log(y)
∫ dy
1 y(y − 1)
Now use the substitution y = 1/x

1 log(x) 1 log(1 − x) π2
−∫ dx = − ∫ dx = Li2 (1) =
0 (1 − x) 0 x 6

11.6 Example

Prove that

∞ 2Γ(p)
∫ z p−1 E 2 (z) dz = 2 F1 (p, p; p + 1; −1)
0 p2
proof

87
Consider

∞ e−t
E 2 (z) = ∫ dt
z t
By differentiation with respect to z

2 e−z E(z)
2E ′ (z)E(z) =
z
Knowing that we can return to our integral by integration by parts

2 ∞ p−1 −z
∫ z e E(z) dz
p 0
Write the integral representation

2 ∞ p−1 −z ∞ e−zt
∫ z e ∫ dtdz
p 0 1 t
Swap the two integrals

2 ∞ 1 ∞ p−1 −z(1+t)
∫ ∫ z e dz dt
p 1 t 0
The inner integral reduces to

2Γ(p) ∞ dt
∫
p 1 t(1 + t)p
Use the substitution t = 1/x

2Γ(p) 1 xp−1
∫ dx
p 0 (1 + x)p

Using the integral representation of the Hypergeometric function

1 xb−1 (1 − x)c−b−1
β(c − b, b) 2 F1 (a, b; c; z) = ∫ dx
0 (1 − xz)a
Let c = p + 1, b = p, a = p, z = −1

1 xp−1
β(1, p) 2 F1 (p, p; p + 1; −1) = ∫ dx
0 (1 + x)p
Hence the result

∞ 2Γ(p)
∫ z p−1 E 2 (z) dz = 2 F1 (p, p; p + 1; −1)
0 p2

88
Where

Γ(p) 1
β(1, p) = =
Γ(p + 1) p

11.7 Exercise

Find the integral fo n ∈ N

∞
∫ xn E 2 (x) dx
0

89
12 Complete Elliptic Integral

12.1 Complete elliptic of first kind

π
2 dθ 1 dx
K(k) = ∫ √ =∫ √ √
0 1 − k 2 sin2 θ 0 1 − x2 1 − k 2 x2

12.2 Complete elliptic of second kind

√ √
1 − k 2 x2
π
2 1
E(k) = ∫ 1 − k 2 sin2 θ dθ =∫ √ dx
0 0 1 − x2

12.3 Hypergeometric representation

π 1 1
K(k) = 2 F1 ( , , 1, k )
2
2 2 2
and

π 1 1
E(k) = 2 F1 ( , − , 1, k )
2
2 2 2
proof

Using the integral representation of the hypergeometric function

1 tb−1 (1 − t)c−b−1
β(c − b, b) 2 F1 (a, b, c, z) = ∫ dt
0 (1 − tz)a
Now use the substitution t = x2 and z = k 2

1 x2b−1 (1 − x2 )c−b−1
β(c − b, b) 2 F1 (a, b, c, k 2 ) = 2 ∫ dx
0 (1 − k 2 x2 )a
Put a = 1
2
;b= 1
2
and c = 1

1 dx 1 1 1
∫ √ √ = β(1/2, 1/2) 2 F1 ( , , 1, k 2 )
0 1 − x2 1−k x
2 2 2 2 2
By the beta function we have

1 1 1 π
β(1/2, 1/2) = Γ2 ( ) =
2 2 2 2
Hence the result

π 1 1
K(k) = 2 F1 ( , , 1, k )
2
2 2 2

90
By the same approach we have

π 1 1
E(k) = 2 F1 ( , − , 1, k )
2
2 2 2

12.4 Example

Prove that

1
∫ K(k) dk = 2G
0

G is the Catalan’s constant.

proof

Start by the integral representation

1 1 1
I =∫ ∫ √ √ dx dk
0 0 1 − x2 1 − k 2 x2
Switching the two integrals

1 1 1 1
I =∫ √ ∫ √ dk dx
0 1 − x2 0 1 − k 2 x2

1 arcsin x
I =∫ √ dx
0 x 1 − x2
Now let arcsin x = t hence we have x = sin t

π
2 t
I =∫ dt
0 sin t
The previous integral is a representation of the constant

I
G= Ô⇒ I = 2G
2

12.5 Identities

1. For k ≥ 1

√
√ ⎛ k ⎞ 1
K( k) = √ K
1 − k ⎝ k − 1⎠

and

91
 √
√ √ ⎛ k ⎞
E( k) = 1 − k E
⎝ k − 1⎠

proof

Starting by the integral representation

1 dx
K(k) = ∫ √ √
0 1 − x2 1 − k 2 x2
√
Use the substitution x = 1 − y2

1 y dy
∫ √ √ √
0 1 − y2 1 − (1 − y 2 ) 1 − k 2 (1 − y 2 )

By cancelling the terms we have

1 dy
∫ √ √
0 1 − y2 1 − k2 + k2 y2
√
Take 1 − k 2 as a common factor

1 dy
∫ √ √ √
k2
0
1 − k2 1 − y2 1− k2 −1
y2

Comparing this to the integral representation we get

√
⎛ k2 ⎞ 1
K(k) = √ K
1 − k2 ⎝ k − 1 ⎠
2

√
We can finish by k → k

√
√ ⎛ k ⎞ 1
K( k) = √ K
1 − k ⎝ k − 1⎠

Similarly for the second representation

√
√ 1 1 − kx2
E( k) = ∫ √ dx
0 1 − x2
√
By using that x = 1 − y2

√ √
√ √ 1 1 − k−1
k
y √ ⎛ k ⎞
E( k) = 1−k∫ √ dy = 1 − k E
0 1−y 2 ⎝ k − 1⎠

92
2.
√
1 2 k
K(k) = K( )
k+1 1+k

proof

Start by the Quadratic transformation

4z 1 1
2 F1 (a, b, 2b, ) = (1 + z)2a 2 F1 (a, a − b + , b + , z 2 ) .
(1 + z)2 2 2

Hence we can deduce by putting a = b = 1/2

√
2 k
K( ) = (1 + k)K(k)
1+k

Or we have

√
1 2 k
K(k) = K( )
k+1 1+k

3.
√ √
2 k 1+k 2 −k
K( )= K( )
1+k 1−k 1−k

and

√ √
2 k 1−k 2 −k
E( )= E( )
1+k 1+k 1−k

proof

Start by the following

√
2 k 1 1
K( )=∫ √ √ dx
1+k 0 1 − x2 1 − 4k
x2
(1+k)2

By some simplifications we have

√
2 k 1 1
E( ) = (1 + k) ∫ √ √ dx
1+k 0 1 − x (1 + k)2 − 4k x2
2

√
Use x = 1 − y2

1 1+k 1+k 1 1
∫ √ √ dy = ∫ √ √ dy
0 1 − y 2 (1 + k)2 − 4k (1 − y 2 ) 1−k 0 1 − y2 1 + 4k
y2
(1−k)2

93
 Hence we have

√ √
2 k 1+k 2 −k
K( )= K( )
1+k 1−k 1−k

Similarly we have

√ √
2 k 1−k 2 −k
E( )= E( )
1+k 1+k 1−k

12.6 Special values

1.
1 1
K(i) = √ Γ2 ( )
4 2π 4
and

Γ2 ( 41 ) Γ2 ( 34 )
E(i) = √ + √
4 2π 2π

proof

By definition we have

1 dx 1 dx
K(i) = ∫ √ √ =∫ √
0 1−x 1+x
2 2 0 1 − x4
√
Let x = t we have dx = 14 t
4 −3
4 dt

1 1 −3
K(i) = ∫ t 4 (1 − t) 2 dt
−1

4 0

By beta function

Γ ( 41 ) Γ ( 12 )
K(i) =
4Γ ( 34 )

By reflection formula

1 3 π √
Γ ( ) Γ ( ) = π csc ( ) = π 2
4 4 4

Γ2 ( 41 ) Γ ( 12 ) Γ2 ( 14 )
K(i) = √ = √
4π 2 4 2π

94
 Γ2 ( 41 ) Γ2 ( 34 )
E(i) = √ + √
4 2π 2π

By definition we have

√
1 1 + x2
E(i) = ∫ √ dx
0 1 − x2

Separating the two integrals

11 + x2 1 1 1 x2
E(i) = ∫ √ dx = ∫ √ dx + ∫ √ dx
0 1 − x4 0 1 − x4 0 1 − x4
√
The first integral is K(i) for the second integral use x = 4 t

1 1 34 −1 −1
Γ ( 43 ) Γ ( 12 ) Γ2 ( 34 )
∫ t (1 − t) 2 dt = = √
4 0 4Γ ( 54 ) 2π

Hence we have

Γ2 ( 3 ) Γ2 ( 1 ) Γ2 ( 3 )
E(i) = K(i) + √ 4 = √ 4 + √ 4
2π 4 2π 2π

2.
1 1 1
K ( √ ) = √ Γ2 ( )
2 4 π 4

and

1 Γ2 ( 1 ) Γ2 ( 3 )
E ( √ ) = √4 + √4
2 8 π 2 π

proof

Start by the identity

√
√ 1 ⎛ k ⎞
K( k) = √ K
1 − k ⎝ k − 1⎠

For the value k = −1

1 1
K(i) = √ K ( √ )
2 2

Using the value for K(i)

95
 1 √ 1 1
K ( √ ) = 2K(i) = √ Γ2 ( )
2 4 π 4

Similarly we have

1 Γ2 ( 1 ) Γ2 ( 3 )
E ( √ ) = √4 + √4
2 8 π 2 π

3.
√ √
√ 1+ 2 2 1
K (2 −4 + 3 2) = √ Γ ( )
4 2π 4

and

√ √
√ 2 ⎡⎢ Γ ( ) Γ ( ) ⎤⎥
2 1 2 3
E (2 −4 + 3 2) = √ ⎢ √4 + √ 4 ⎥
1 + 2 ⎢⎣ 8 π π ⎥⎦

proof

Start by the identity

√
1 2 k
K(k) = K( )
k+1 1+k
Hence we have for k = √1
2

√ √ √
√ 1+ 2 1 1+ 2 2 1
K (2 −4 + 3 2) = √ K ( √ ) = √ Γ ( )
2 2 4 2π 4
−1
For the elliptic integral of second kind using the hypergeomtric representation with a = 2
and

b= 1
2

4z
2 F1 (−1/2, 1/2, 1, ) = (1 + z)−1 2 F1 (−1/2, −1/2, 1, z 2 )
(1 + z)2

The later hypergeometric series can be written in terms of elliptic integrals using some general

contiguity relations

2
2 F1 (−1/2, −1/2, 1, z )= (2E(k) + (k 2 − 1)K(k))
2
π

So we have

√
2 k
2E(k) + (k 2 − 1)K(k) = (k + 1)E ( )
1+k

96
 For k = √1
2

√ √
√ 2 ⎡⎢ Γ ( ) Γ ( ) ⎤⎥
2 1 2 3
E (2 −4 + 3 2) = √ ⎢ √4 + √ 4 ⎥
1 + 2 ⎢⎣ 8 π π ⎥⎦

4.
√ √ √
√ π π 2− 2
K (2 −4 − 3 2) = ⋅
4 Γ2 ( 43 )

and

√
√ √ (2 + 2) (π 2 + 4Γ4 ( 34 ))
E (2 −4 − 3 2) = √
4 πΓ2 ( 34 )

proof

Start by the following identity

√ √
2 k 1+k 2 −k
K( )= K( )
1+k 1−k 1−k
√
Let x = 1/ 2
√ √ √
√ π π 2− 2
K (2 −4 − 3 2) = ⋅
4 Γ2 ( 43 )

√
√ √ (2 + 2) (π 2 + 4Γ4 ( 34 ))
E (2 −4 − 3 2) = √
4 πΓ2 ( 34 )

12.7 Differentiation of elliptic integrals

Note We should remove the variable k and denote elliptic integrals E and K once there is no confusion.

It is assumed that the variable is k when we use these symbols.

Interestingly the derivative of elliptic integrals can be written in terms of elliptic integrals

Derivative of complete elliptic integral of second kind

√
d 1 ∂
1 − k 2 x2
E=∫ ∂k
√ dx
dk 0 1 − x2

d 1 −k x2
E=∫ √ √ dx
dk 0 1 − x2 1 − k 2 x2
Adding and subtracting 1 results in

√
1 1 1 − k 2 x2 1 1 dx
∫ √ dx − ∫ √ √
k 0 1−x 2 k 0 1 − x 1 − k 2 x2
2

97
Upon realizing the relation to elliptic integrals we conclude

d E−K
E=
dk k
For the complete elliptic integral of first kind we need more work

Start by the following

d 1 1 ∂ 1
K=∫ √ [√ ] dx
dk 0 1−x2 ∂ k 1 − k 2 x2

d 1 kx2
K=∫ √ √ dx
dk 0 1 − x2 1 − k 2 x2 (1 − k 2 x2 )
Adding and subtracting 1 we have

−1 1 1 − kx2 − 1 1 1 dx K
∫ √ √ dx = ∫ √ √ −
k 0 1 − x2 1 − k 2 x2 (1 − k 2 x2 ) k 0 1 − x2 1 − k 2 x2 (1 − k 2 x2 ) k
Let us focus on the first integral

1 1
∫ √ 3
dx
0 1 − x2 (1 − k 2 x2 ) 2
√
Let x = t and we have dx = 1
√
2 t
dt

t− 2
1
1 1
∫ √ 3 dx
2 0 1 − t(1 − k 2 t) 2
Using the hypergeometric integral representation

t− 2
1
1 1 π 3 1
∫ √ 3 = 2 F1 ( , , 1, k )
2
2 0 1 − t(1 − k 2 t) 2 2 2 2
Using the linear transformation

2 F1 (a, b, c, z) = (1 − z)c−a−b 2 F1 (c − a, c − b, c, z)
√
We get by putting k ′ = 1 − k2

π 3 1 1 π 1 1 E
2 F1 ( , , 1, k ) = 2 F1 (− , , 1, k ) = ′2
2 2
2 2 2 1−k 2
2 2 2 k
So finally we get

d 1 E
K = ( ′2 − K)
dk k k

98
13 Euler sums

13.1 Definition
∞ (p)
(Hk )r
Sp r ,q = ∑
k=1 kq
Where we define the general harmonic number

k k
(p) 1 (1) 1
Hk =∑ p
; H k ≡ Hk = ∑
n=1 n n=1 n

Euler sums were greatly studied by Euler, hence the name.

13.2 Generating function

∞ Lip (x)
(p)
∑ Hk x =
k

k=1 1−x
Proof
(p)
Start by writing Hk as a sum

∞ ∞ k
(p) 1 k
∑ Hk x = ∑ ∑
k
p
x
k=1 k=1 n=1 n

By interchanging the two series we have

∞ ∞ ∞
xk 1 ∞ k
∑ ∑ p
= ∑ p ∑
x
n=1 k=n n n=1 n k=n

The inner sum is a geometric series

1 ∞ xn Lip (x)
∑ =
1 − x n=1 np 1−x
We can use this to generate some more functions by integrating.

13.3 Integral representation of Harmonic numbers

1 1 − xn
Hn = ∫ dx
0 1−x
proof

We can use the geometric series of xn

∞
1 1 − xn 1
∫ dx = ∑ ∫ xk − xn+k dx
0 1−x k=0 0

99
 ∞
1 1
∑ − = Hn
k=0 k + 1 n+k+1

13.4 Example
∞
Hn
∑ 2
= 2ζ(3)
n=1 n

proof

Using the integral representation

1 1 ∞ 1 − xn 1 ζ(2) − Li (x)
2
∫ ∑ dx = ∫ dx
0 1 − x n=1 n 2 0 1−x
Now use the functional equation

ζ(2) − Li2 (x) = Li2 (1 − x) + log(x) log(1 − x)

Hence we have

1 Li2 (1 − x) + log(x) log(1 − x)

∫ dx
0 1−x
The first integral

1 Li2 (1 − x)
∫ = Li3 (1) = ζ(3)
0 1−x
The Second integral using integration by parts

1 log(1 − x) log(x) 1 Li (x)

2
∫ dx = ∫ = ζ(3)
0 x 0 x
Finally we have

∞
Hn
∑ 2
= ζ(3) + ζ(3) = 2ζ(3)
n=1 n

13.5 Example
∞
Hk k 1
∑ 2
x = Li3 (x) − Li3 (1 − x) + log(1 − x)Li2 (1 − x) + log(x) log2 (1 − x) + ζ(3)
k=1 k 2
proof

In the general definition assume p = 1

100
 ∞
log(1 − x)
∑ Hk x = −
k

k=1 1−x
Divide by x and integrate to get

∞
Hk k 1
∑ x = Li2 (x) + log2 (1 − x)
k=1 k 2
Now divide by x and integrate again

∞
Hk k 1 x log2 (1 − t)
∑ 2
x = Li3 (x) + ∫ dt
k=1 k 2 0 t
Now let us look at the integral

x log2 (1 − t)
∫ dt
0 t
Integrating by parts

x log2 (1 − t) x Li (t)
2
∫ dt = − log(1 − x)Li2 (x) − ∫ dt
0 t 0 1−t
Use a change of variable in the integral

x Li2 (t) 1 Li (1 − t)
2
∫ dt = ∫ dt
0 1−t 1−x t
Now we can use the second functional equation of the dilogarithm

π2
1
6
− Li2 (t) − log(1 − t) log(t)
∫ dt
1−x t
Separate the integrals

π2 1 Li (t) 1 log(1 − t) log(t)

2
− log(1 − x) − ∫ dt − ∫ dt
6 1−x t 1−x t
The first integral

1 Li2 (t)
∫ dt = Li3 (1) − Li3 (1 − x)
1−x t
Use integration by parts in the second integral

1 log(1 − t) log(t)
∫ = Li3 (1) + log(1 − x)Li2 (1 − x) − Li3 (1 − x)
1−x t
Collecting the results together we obtain

101
 x Li2 (t) π2
∫ dt = − log(1 − x) − Li2 (1 − x) log(1 − x) + 2 Li3 (1 − x) − 2ζ(3)
0 1−t 6
Hence we solved the integral

x log2 (1 − t) π2
∫ dt = − log(1 − x)Li2 (x) + log(1 − x) + Li2 (1 − x) log(1 − x) − 2 Li3 (1 − x) + 2ζ(3)
0 t 6

So we have got our Harmonic sum

∞
Hk k 1 π2
∑ 2
x = Li3 (x) + (− log(1 − x)Li2 (x) + log(1 − x) + Li2 (1 − x) log(1 − x) − 2 Li3 (1 − x) + 2ζ(3))
k=1 k 2 6

∞
Hk k 1
∑ 2
x = Li3 (x) − Li3 (1 − x) + log(1 − x)Li2 (1 − x) + log(x) log2 (1 − x) + ζ(3)
k=1 k 2

13.6 General formula

∞
Hn q 1 q−2
∑ q
= (1 + ) ζ(q + 1) − ∑ ζ(k + 1)ζ(q − k)
n=1 n 2 2 k=1
This can be proved using complex analysis.

13.7 Example
1 log2 (1 − x) log(x) π4
∫ =−
0 x 180
proof

Using the generating function

∞
log(1 − x)
∑ Hk x
k−1
=−
k=1 x(1 − x)
By integrating both sides

∞
Hk k 1
∑ x = Li2 (x) + log2 (1 − x)
k=1 k 2
Or

∞
Hk k
log2 (1 − x) = 2 ∑ x − 2Li2 (x)
k=1 k

plugging this in our integral we have

102
 1 ∞
Hk k log(x)
2∫ (∑ x − Li2 (x)) dx
0 k=1 k x
Which simplifies to

∞ 1 1
Hk log(x)
∫ log(x)x dx − 2 ∫ Li2 (x)
k−1
2∑ dx
k=1 k 0 0 x
The second integral

1 log(x) 1 Li (x)
3
−2 ∫ Li2 (x) dx = 2 ∫ dx = 2ζ(4)
0 x 0 x
The first integral

∞ 1
Hk k−1
2∑ ∫ x log(x) dx
k=1 k 0

Using integration by parts twice and the general formula for q = 3

∞
Hk
−2 ∑ 3
= −5ζ(4) + ζ 2 (2)
k=1 k

Finally we get

1 log2 (1 − x) log(x)
∫ = −5ζ(4) + ζ 2 (2) + 2ζ(4) = ζ 2 (2) − 3ζ(4)
0 x

13.8 Example

Show that

∞ log t log(a2 + b2 ) b
∫ e−at sin(bt) dt = − ( + γ) arctan ( )
0 t 2 a
Proof

We can start by the following integral

∞
I(s) = ∫ ts−1 e−at sin(bt)dt
0

By using the the expansion of the sine function

∞ ∞
(−1)n (bt)2n+1
I(s) = ∫ ts−1 e−at ∑
0 n=0 Γ(2n + 2)
By swapping the summation and integration

103
 ∞
(−1)n (b)2n+1 ∞ 1 ∞ (−1)n (b)2n+1 Γ(s + 2n + 1)
I(s) = ∑ ∫ ts+2n e−at dt = s ∑
n=0 Γ(2n + 2) 0 a n=0 Γ(2n + 2)a2n+1
By differentiating and plugging s = 0 we have
∞ ∞
(−1)n ψ0 (2n + 1) b 2n+1 (−1)n b 2n+1
I ′ (0) = ∑ ( ) − log(a) ∑ ( )
n=0 2n + 1 a n=0 2n + 1 a

Now use that ψ(n + 1) + γ = Hn

∞
(−1)n H2n − γ b 2n+1 b
I ′ (0) = ∑ ( ) − log(a) arctan ( )
n=0 2n + 1 a a

∞
(−1)n H2n b 2n+1 b
I ′ (0) = ∑ ( ) − (γ + log(a)) arctan ( )
n=0 2n + 1 a a
Now we look at the harmonic sum

∞ ∞ 1 1 − t2k
∑ (−1) H2k x = ∑ (−1) x ∫
k 2k k 2k
dt
k=0 k=0 0 1−t
Use the integral representation

1 1 ∞
∑ (−1) x (1 − t ) dt
k 2k 2k
∫
0 1 − t k=0
Swap the series and the integral

1 1 ∞
∑ (−1) (x − (xt) ) dt
k 2k 2k
∫
0 1 − t k=0
Evaluate the geometric series

1 1 1 1 −x2 1 (1 − t2 )
∫ ( − ) dt = ∫ dt
0 1 − t 1 + x2 1 + t 2 x2 1 + x2 0 (1 − t)(1 + t2 x2 )
which simplifies to

−x2 1 1+t −x2 1 1 t

∫ dt = (∫ + dt)
1 + x 0 (1 + t x )
2 2 2 1+x 2 0 1 + t 2 x2 1 + t 2 x2
Evaluating the integrals

−1
(2x arctan(x) + log(1 + x2 ))
2(1 + x2 )
Using this we conclude by integrating

∞
(−1)k H2k 2k 1
∑ x = − log(1 + x2 ) arctan(x)
k=0 2k + 1 2

104
Hence the following

∞
(−1)k H2k b 2k+1 1 a2 + b2 b
∑ ( ) = − log ( ) arctan ( )
k=0 2k + 1 a 2 a2 a
Substituting that in our integral

∞ log t 1 a2 + b2 b
∫ e−at sin(bt) dt = − ( log ( 2
) + γ + log(a)) arctan ( )
0 t 2 a a
log(a2 + b2 ) b
= −( + γ) arctan ( )
2 a

13.9 Example

1 Lip (x) Liq (x) p−1

1 p+q−2
∫ dx = ∑ (−1)n−1 ζ(p − n + 1)ζ(q + n) − ∑ (−1) ζ(n + 1)ζ(p + q − n)
p−1
0 x n=1 2 n=1
p+q
+ (−1) (1 +
p−1
) ζ(p + q + 1)
2

proof

We can see that

1 Lip (x) Liq (x) ∞ ∞

1
∫ dx = ∑ ∑ q p
0 x k=1 n=1 k n (n + k)
Let us first look at the following

∞
1 Hk
C (α, k) = ∑ ; C (1, k) =
n=1 nα (n + k) k
This can be solved using

∞
1 1 1
C (α, k) = ∑ ( − )
n=1 k n
α−1 n n+k
1 1
= ζ(α) − C (α − 1, k)
k k
1 1 1
= ζ(α) − 2 ζ(α − 1) + 2 C (α − 2, k)
k k k
⋅

⋅
1 1 ζ(2) 1
= ζ(α) − 2 ζ(α − 1) + ⋯ + (−1)α α−1 + α−1 C (α − (α − 1), k)
k k k k
α−1
ζ(α − n + 1) Hk
= ∑ (−1)n−1 n
+ (−1)α−1 α
n=1 k k

105
Hence we have the general formula

α−1
ζ(α − n + 1) Hk
C (α, k) = ∑ (−1)n−1 n
+ (−1)α−1 α
n=1 k k
β
Dividing by k and summing w.r.t to k

∞ ∞
C (α, k) α−1 Hk
∑ β
= ∑ (−1)n−1 ζ(α − n + 1)ζ(β + n) + (−1)α−1 ∑ α+β
k=1 k n=1 k=1 k
Now we use the general formula

∞
Hn q 1 q−2
∑ q
= (1 + ) ζ(q + 1) − ∑ ζ(k + 1)ζ(q − k)
n=1 n 2 2 k=1
Hence we have

∞
Hn α+β 1 α+β−2
∑ α+β
= (1 + ) ζ(α + β + 1) − ∑ ζ(k + 1)ζ(α + β − k)
n=1 n 2 2 k=1
And the generalization is the following formula

∞
C (α, k) α−1 1 α+β−2
∑ = ∑ (−1) n−1
ζ(α − n + 1)ζ(β + n) − ∑ (−1) ζ(n + 1)ζ(α + β − n)
α−1

k=1 kβ n=1 2 n=1

α+β
+ (−1)α−1 (1 + ) ζ(α + β + 1)
2

We conclude by putting that

∞ ∞ ∞
1 C (p, k)
∑∑ q np (n + k)
= ∑
k=1 n=1 k k=1 kq

13.10 Relation to polygamma

We can relate the generalized harmonic number to the polygamma function

(p) ψp−1 (k + 1)
Hk = ζ(p) + (−1)p−1
(p − 1)!
proof

k ∞
(p) 1 1
Hk =∑ p
= ζ(p) − ∑ p
n=1 n n=k+1 n
Now change the index in the sum n = i + k + 1

k ∞
(p) 1 1
Hk =∑ = ζ(p) − ∑
n=1 n p
i=0 (i + k + 1)p

106
We know that

ψp−1 (k + 1) ∞ 1
(−1)p =∑ p≥1
(p − 1)! i=0 (i + k + 1)
p

Hence we have

(p) ψp−1 (k + 1)
Hk = ζ(p) + (−1)p−1
(p − 1)!
We can use that to obtain a nice integral representation.

13.11 Integral representation for r=1

∞ (p) 1 Li (x) log(x)p−1
Hk 1 q
∑ = ζ(p)ζ(q) + (−1) p
∫ dx
k=1 k q (p − 1)! 0 1−x
proof

Note that

1 1 − xa
ψ0 (a + 1) = ∫ dx
0 1−x
By differentiating with respect to a , p times we have

∂ 1 1 − xa
ψp (a + 1) = ∫ dx
∂ap 0 1 − x

1 xa log(x)p
ψp (a + 1) = − ∫ dx
0 1−x
Let a = k

1 xk log(x)p−1
ψp−1 (k + 1) = − ∫ dx
0 1−x
Use the relation to polygamma

1 1 xk log(x)p−1
(p)
Hk = ζ(p) + (−1)p ∫ dx
(p − 1)! 0 1−x
Now divide by k q and sum with respect to k

∞ (p) 1 Li (x) log(x)p−1

H 1 q
∑ kq = ζ(p)ζ(q) + (−1)
p
∫ dx
k=1 k (p − 1)! 0 1−x

107
13.12 Symmetric formula
∞ (p)
∞ H (q)
Hk
∑ q
+ ∑ k
p
= ζ(p)ζ(q) + ζ(p + q)
k=1 k k=1 k

proof

Take the leftmost series and swap the finite and infinite sums

∞ ∞ ∞ ∞ ∞
1 1 1 1 1 i−1 1
∑∑ p q
= ∑ ∑ p q
− ∑ p ∑ q
i=1 k=i i k i=1 k=1 i k i=1 i k=1 k

The second sum can be written as

∞ ∞
1 i−1 1 1 i
1 1
∑ p ∑ q = ∑ p (∑ q − p )
i=1 i k=1 k i=1 i k=1 k i
By separating and changing the index we get

∞ (q)
Hk
∑ p
− ζ(p + q)
k=1 k

Hence we have

∞ (p) ∞ H (q)
H
∑ kq = ζ(p)ζ(q) − ∑ kp + ζ(p + q)
k=1 k k=1 k

∞ (p)
∞ H (q)
Hk
∑ q
+ ∑ kp = ζ(p)ζ(q) + ζ(p + q)
k=1 k k=1 k

For the special case p = q = n

∞ (n)
Hk ζ 2 (n) + ζ(2n)
∑ n
=
k=1 k 2

13.13 Example
∞ (3)
Hk 11ζ(5)
∑ 2
= − 2ζ(2)ζ(3)
k=1 k 2
proof

Using the symmetric formula

∞ (3) ∞ H (2)
H
∑ k2 = ζ(2)ζ(3) + ζ(5) − ∑ k3
k=1 k k=1 k

Using the integral formula on the second sum

∞ (2) 1 Li (x) log(x)

Hk 3
∑ = ζ(2)ζ(3) + ∫ dx
k=1 k 3 0 1−x

108
Using integration by parts on the integral

1 Li3 (x) log(x) 1 Li (x)Li (1 − x)

2 2
∫ dx = − ∫ dx
0 1−x 0 x
Let us think of solving

1 Li2 (x)Li2 (1 − x)
∫ dx
0 x
Using the duplication formula

Li2 (1 − x) = ζ(2) − Li2 (x) − log(x) log(1 − x)

1 Li2 (x)(ζ(2) − Li2 (x) − log(x) log(1 − x))

∫ dx
0 x
The first integral

1 Li2 (x)
ζ(2) ∫ dx = ζ(2)ζ(3)
0 x
The third integral

1 Li2 (x) log(x) log(1 − x) 1 1 Li22 (x)

∫ dx = ∫ dx
0 x 2 0 x
Finally we get

1 Li3 (x) log(x) 3 1 Li22 (x)

∫ dx = ∫ dx − ζ(2)ζ(3)
0 1−x 2 0 x
So

∞ (2)
Hk 3 1 Li22 (x)
∑ 3
= ∫ dx
k=1 k 2 0 x
Hence we finally get that

∞ (3)
Hk 3 1 Li22 (x)
∑ 2
= ζ(2)ζ(3) + ζ(5) − ∫ dx
k=1 k 2 0 x
Let us solve the integral

1 Li22 (x)
∫ dx
0 x
By series expansion

109
 ∞ ∞ 1 ∞ ∞
1 1
∑∑ ∫ x n+k−1
dx = ∑ ∑
n=1 k=1 (nk) n=1 k=1 (nk) (n + k)
2 0 2

By some manipulations we get

∞
1 ∞ k ∞
1 ∞ 1 ∞
1 ∞ 1
∑ 3 ∑ 2 (n + k)
= ∑ 3 ∑ 2
− ∑ 3 ∑ n(n + k)
k=1 k n=1 n k=1 k n=1 n k=1 k n=1

This can be simplified to conclude that

∞
1 Li22 (x) Hk
∫ dx = ζ(2)ζ(3) − ∑ 4
0 x k=1 k

Now using that

∞
Hk
∑ 4
= 3ζ(5) − ζ(2)ζ(3)
k=1 k

Hence

1 Li22 (x)
∫ dx = 2ζ(2)ζ(3) − 3ζ(5)
0 x
Finally we get

∞ (3)
H 3 11ζ(5)
∑ k2 = ζ(2)ζ(3) + ζ(5) − (2ζ(2)ζ(3) − 3ζ(5)) = − 2ζ(2)ζ(3)
k=1 k 2 2

110
14 Sine Integral function

14.1 Definition

We define the following

z sin(x)
Si(z) = ∫ dx
0 x
A closely related function is the following

∞ sin(x)
si(z) = − ∫ dx
z x
These functions are related through the equation

π
Si(z) = si(z) +
2
A closely related function is the sinc function

⎧
⎪
⎪
⎪
⎪ x=0
⎪1
sinc(x) = ⎨
⎪
⎪
⎪
⎪
⎪
sin(x)
x≠0
⎩ x
Using that we conclude

d
Si(x) = sinc(x)
dx
For integration we have

∫ Si(x) dx = cos(x) + x Si(x) + C

14.2 Example

Show that

∞ π
∫ sin(x)si(x) dx = −
0 4
proof

Using integration by parts we get

∞ sin(x) cos(x) 1 ∞ sin(2x)

−∫ dx = − ∫ dx
0 x 2 0 x

111
Let 2x = t

1 ∞ sin(t) π
− ∫ dx = −
2 0 t 4

14.3 Example

Prove

∞ Γ(α) πα
∫ xα−1 si(x) dx = − sin ( )
0 α 2
proof

Using the integral representation

∞ ∞ sin(t)
−∫ xα−1 ∫ dt dx
0 x t
Let xy = t

∞ ∞ sin(xy)
−∫ xα−1 ∫ dy dx
0 1 y
Switching the integrals we get

∞ 1 ∞
−∫ ∫ xα−1 sin(xy) dx dy
1 y 0
Now let xy = t

∞ 1 ∞
−∫ ∫ tα−1 sin(t) dt dy
1 y α+1 0

The Mellin transform of the sine function is defined as

∞ πs
Ms (sin(x)) = ∫ xs−1 sin(x) dx = Γ(s) sin ( )
0 2
Hence we conclude that

πα ∞ 1 Γ(α) πα
−Γ(α) sin ( )∫ α+1
=− sin ( )
2 1 y α 2

112
14.4 Example

Show that

∞ arctan(α)
∫ e−α x si(x) dx = −
0 α
proof

Use the integral representation

∞ ∞ sin(t)
−∫ e−α x ∫ dt dx
0 x t
Let xy = t

∞ ∞ sin(xy)
−∫ e−α x ∫ dy dx
0 1 y
Switching the integrals

∞ 1 ∞
−∫ ∫ e−α x sin(xy) dx dy
1 y 0
The inner integral is the laplace transform of the sine function

a
Ls (sin(at)) =
s2 + a2
Hence we conclude that

∞ 1 arctan(α)
−∫ dy = −
1 y 2 + α2 α

14.5 Example

Prove the following

∞
∫ si(x) log(x) dx = γ + 1
0

proof We know that

∞ Γ(α) πα
∫ xα−1 si(x) dx = − sin ( )
0 α 2
Differentiate with respect to α

113
 ∞ Γ(α) πα Γ(α)ψ(α) πα πΓ(α) πα
∫ xα−1 si(x) log(x)dx = 2
sin ( ) − sin ( ) − cos ( )
0 α 2 α 2 2α 2
Let α → 1

∞
∫ si(x) log(x)dx = 1 − ψ(1) = 1 − (−γ) = 1 + γ
0

14.6 Example

Find the integral

∞
∫ si(x) sin(px) dx
0

solution

Using integration by parts we get

∞
si(x) cos(px) 1 ∞ sin(x)
[− ] + ∫ cos(px) dx
p 0
p 0 x
Taking the limits

si(x) cos(px) si(0) π

lim − = =
x→0 p p 2p

si(x) cos(px)
lim − =0
x→∞ p
Hence we get

π 1 ∞ sin(x)
− + ∫ cos(px) dx
2p p 0 x
The integral

∞ sin(x) 1 ∞ sin((p + 1)x) − sin((p − 1)x)

∫ cos(px) dx = ∫ dx
0 x 2 0 x
Separate the integrals

1 ∞ sin((p + 1)x) 1 ∞ sin((p − 1)x)

I= ∫ dx − ∫ dx
2 0 x 2 0 x
If p − 1 > 0 we get

π π
I= − =0
4 4

114
If p − 1 < 0

1 ∞ sin((p + 1)x) 1 ∞ sin((1 − p)x) π

I= ∫ dx + ∫ dx =
2 0 x 2 0 x 2
If p = 1 we have

1 ∞ sin(2x) π
I= ∫ dx + 0 =
2 0 x 4
Finally we get

⎧
⎪
⎪
⎪
⎪− 2p
π
p>1
⎪
⎪
⎪
∞ ⎪
⎪
⎪
∫ si(x) sin(px)dx = ⎨− π p=1
0 ⎪
⎪
⎪
4p
⎪
⎪
⎪
⎪
⎪
⎪
⎪0 p<1
⎩

14.7 Example

Prove that for 0 < a < 2

∞ π
∫ si2 (x) cos(ax) dx = log(a + 1)
0 2a
proof

Using integration by parts we get

∞
si2 (x) sin(ax) 2 ∞ si(x) sin(x)
[ ] − ∫ sin(ax) dx
a 0
a 0 x
Taking the limits

si2 (x) sin(ax)

lim =0
x→0 a

si2 (x) sin(ax)

lim =0
x→∞ a
Let the integral

∞ si(x) sin(x)
I(a) = ∫ sin(ax) dx
0 x
Differentiate with respect to a

∞
I ′ (a) = ∫ si(x) sin(x) cos(ax) dx
0

115
Now use the product to sum trigonometric rules

1 ∞
I ′ (a) = ∫ si(x)(sin((a + 1)x) − sin((a − 1)x)) dx
2 0
From the previous exercise we have

∞ −π
∫ si(x) sin((a + 1)x)dx = ;a>0
0 4(a + 1)

∞
∫ si(x) sin((a + 1)x)dx = 0 ; a < 2
0

Hence we conclude that for 0 < a < 2

π
I ′ (a) = −
4(a + 1)
Integrate with respect to a

π
I(a) = − log(a + 1) + C
4
Let a → 0

I(0) = 0 + C → C = 0

Hence we have

∞ si(x) sin(x) π
∫ sin(ax) dx = − log(a + 1)
0 x 4
Which implies that

∞ −2 π π
∫ si2 (x) cos(ax) dx = (− log(a + 1)) = log(a + 1)
0 a 4 2a

14.8 Example

Find the integral, for a ≠ 1

∞
∫ si(x) cos(ax) dx
0

solution

Use integration by parts to obtain

116
 1 ∞ sin(x) sin(ax)
∫ dx
a 0 x
Let the integral

∞ sin(x) sin(ax)
I(t) = ∫ e−tx dx
0 x
Differentiate with respect to t

∞
I ′ (t) = − ∫ e−tx sin(x) sin(ax) dx
0

Use product to sum rules

1 ∞ −tx
I ′ (t) = ∫ e (cos((a + 1)x) − cos((a − 1)x)) dx
2 0
Now we can use the Laplace transform

1 t t
I ′ (t) = ( 2 − 2 )
2 t + (a + 1) 2 t + (a − 1)2
Integrate with respect to t

1 t2 + (a + 1)2
I(t) = − log ( 2 )+C
4 t + (a − 1)2
After verifying the constant goes to 0, we have

∞ sin(x) sin(ax) 1 t2 + (a + 1)2

∫ e−tx dx = − log ( 2 )
0 x 4 t + (a − 1)2
Let t → 0

∞ sin(x) sin(ax) 1 a+1 2

∫ dx = − log ( )
0 x 4 a−1
We conclude that

∞ 1 a+1 2
∫ si(x) cos(ax) dx = − log ( )
0 4a a−1

117
15 Cosine Integral function

15.1 Definition

Define

∞ cos(t)
ci(x) = − ∫ dt
x t
A related function is the following

x 1 − cos(t)
Cin(x) = ∫ dt
0 t
The derivative is

d cos(x)
ci(x) =
dx x
The integral

∫ ci(x) dx = xci(x) − sin(x) + C

15.2 Relation to Euler constant

Prove that

lim Cin(z) − log z = γ

z→∞

proof

Write the integral representation

z 1 − cos(t)
lim ∫ dt − log z
z→∞ 0 t
Can be written

z 1 − cos(t) z 1 ∞ 1 cos(t)
lim ∫ dt − ∫ dt = ∫ − dt
z→∞ 0 t 0 1+t 0 t(1 + t) t
This is equivalent to

∞ ts−1
lim ∫ − ts−1 cos(t)dt
s→0 0 (1 + t)
The first integral

118
 ∞ ts−1
∫ = Γ(s)Γ(1 − s)
0 (1 + t)
The second integral

∞
∫ ts−1 cos(t)dt = Γ(s) cos(πs/2)
0

Hence, it reduces to evaluating the limit

lim Γ(s)Γ(1 − s) − Γ(s) cos(πs/2)

s→0

Using Γ(s + 1) = sΓ(s)

Γ(1 − s) − cos(πs/2)
lim
s→0 s
Use L’Hospital rule

lim −Γ(1 − s)ψ(1 − s) + (π/2) sin(πs/2) = −ψ(1) = γ

s→0

15.3 Example

Prove the following

Cin(x) = −ci(x) + log(x) + γ

Start by

x 1 − cos(t)
Cin(x) = ∫ dt
0 t
Rewrite as

∞ 1 − cos(t) ∞ 1 − cos(t)
Cin(x) = ∫ dt − ∫ dt
0 t x t
Which simplifies to

z 1 − cos(t)
Cin(x) = lim [∫ dt − log(z)] − ci(x) + log(x)
z→∞ 0 t
The limit goes to the Euler Maschorinit constant

Cin(x) = γ − ci(x) + log(x)

119
15.4 Example

Find the integral

∞
∫ ci(x) cos(px) dx
0

solution

Using integration by parts we get

∞
ci(x) sin(px) 1 ∞ cos(x)
[ ] − ∫ sin(px) dx
p 0
p 0 x
Taking the limits

ci(x) sin(px)
lim =0
x→0 p

ci(x) sin(px)
lim =0
x→∞ p
Hence we get

1 ∞ cos(x)
− ∫ sin(px) dx
p 0 x
The integral

∞ cos(x) 1 ∞ sin((p − 1)x) + sin((p + 1)x)

∫ sin(px) dx = ∫ dx
0 x 2 0 x
Separate the integrals

1 ∞ sin((p + 1)x) 1 ∞ sin((p − 1)x)

I= ∫ dx + ∫ dx
2 0 x 2 0 x
If p − 1 > 0 we get

π π π
I= + =
4 4 2
If p − 1 < 0

π π
I= − =0
4 4
If p = 1 we have

120
 1 ∞ sin(2x) π
I= ∫ dx + 0 =
2 0 x 4
Finally we get

⎧
⎪
⎪
⎪
⎪− 2p
π
p>1
⎪
⎪
⎪
∞ ⎪
⎪
⎪
∫ ci(x) cos(px) dx = ⎨− π p=1
0 ⎪
⎪
⎪
4p
⎪
⎪
⎪
⎪
⎪
⎪
⎪0 p<1
⎩

15.5 Example

Find for p > 1

∞
∫ ci(px)ci(x) dx
0

solution

Let

∞
I(p) = ∫ ci(px)ci(x) dx
0

Differentiate with respect to p

1 ∞
I ′ (p) = ∫ cos(px)ci(x) dx
p 0
If p > 1 from the previous example we conclude that

1 −π π
I ′ (p) = ( )=− 2
p 2p 2p
Integrate with respect to p

π
I(p) = +C
2p
Take the limit p → ∞, so C = 0.

15.6 Example

Prove that

∞ Γ(α) απ
∫ xα−1 ci(x) dx = − cos ( )
0 α 2

121
proof

Use the integral representation

∞ ∞ cos(t)
−∫ xα−1 ∫ dt dx
0 x t
Let t = yx

∞ ∞ cos(yx)
−∫ xα−1 ∫ dy dx
0 1 y
Switch the integrals

∞ 1 ∞
−∫ ∫ xα−1 cos(yx) dx dy
1 y 0
Using the Mellin transform we get

απ ∞ 1 Γ(α) απ
−Γ(α) cos ( )∫ 1+α
dy = − cos ( )
2 1 y α 2

15.7 Example

Prove that

∞ π
∫ ci(x) log(x) dx =
0 2
proof

From the previous example we know

∞ Γ(α) απ
∫ xα−1 ci(x) dx = − cos ( )
0 α 2
Differentiate with respect to α

∞ Γ(α) απ Γ(α)ψ(α) απ π Γ(α) απ

∫ xα−1 ci(x) log(x) dx = cos ( ) − cos ( ) + sin ( )
0 α2 2 α 2 2 α 2
Take the limit α → 1

∞ π π π
∫ ci(x) log(x) dx = 0 − 0 + sin ( ) =
0 2 2 2

122
15.8 Example

Show that

∞ 1 √
∫ ci(x)e−αx dx = − log 1 + α2
0 α
proof

Use the integral representation

∞ ∞ cos(yx)
−∫ e−αx ∫ dy dx
0 1 y
Switch the integrals

∞ 1 ∞
−∫ ∫ e−αx cos(yx) dx dy
1 y 0
Use the Laplace transformation

∞ α 1 1 √
−∫ dy = − log(1 + α2 ) = − log 1 + α2
1 y(α2+y )
2 2α α

123
16 Integrals involving Cosine and Sine Integrals

16.1 Example

Find the integral

∞
∫ si(qx)ci(x) dx
0

solution

Using the integral representation

∞ ∞ cos(yx)
−∫ si(qx) ∫ dy dx
0 1 y
Switch the integrals

∞ 1 ∞
−∫ ∫ si(qx) cos(yx) dx dy
1 y 0
We also showed that

∞ 1 a+1
∫ si(x) cos(ax) dx = − log ( )
0 2a a−1
Let a = y/q

∞ q y+q
∫ si(x) cos(yx/q) dx = − log ( )
0 2y y−q
Let x = tq

∞ 1 y+q
∫ si(qt) cos(yt) dx = − log ( )
0 2y y−q
Substitue the value of the itnegral

1 ∞ 1 y+q
∫ log ( ) dy
2 1 y 2 y−q
We can prove that the anti-derivative

∞
log(y) 1 1 y+q
[ − log(y 2 − q 2 ) − log ( )]
q 2q 2y y−q 1
Which simplifies

∞
1 y2 − q2 1 y+q
[− log ( )− log ( )]
2q y2 2y y−q 1

124
The limit y → ∞

1 y2 − q2 1 y+q
lim log ( )+ log ( )=0
y→∞ 2q y 2 2y y−q
The limit y → 1
1 1 1+q
log (1 − q 2 ) + log ( )
2q 2 1−q
Can be written as

1 2 1 1+q 2
log (1 − q 2 ) + log ( )
4q 4 1−q

16.2 Example

Prove that

∞ ci(αx) 1
∫ dx = − {si(αβ)2 + ci(αβ)2 }
0 x+β 2
proof

Let the following

∞ ci(αx)
I(α) = ∫ dx
0 x+β
Differentiate with respect to α

1 ∞ cos(αx)
I ′ (α) = ∫ dx
α 0 x+β
Let x + β = t

1 ∞ cos(α(t − β))
I ′ (α) = ∫ dt
α β t
Use trignometric rules

1 ∞ cos(αt) cos(αβ) + sin(αt) sin(αβ)

I ′ (α) = ∫ dt
α β t
Separate the integrals

cos(αβ) ∞ cos(αt) sin(αβ) ∞ sin(αt)

I ′ (α) = ∫ dt + ∫ dt
α β t α β t
This simplifies to

125
 cos(αβ) sin(αβ)
I ′ (α) = − ci(αβ) − si(αβ)
α α
Integrate with respect to α

1
I(α) = − {si(αβ)2 + ci(αβ)2 } + C
2
If α → ∞ we have C = 0.

126
17 Logarithm Integral function

17.1 Definition

Define =

x dt
li(x) = ∫
0 log(t)
The derivative is

d 1
li(z) =
dz log(z)
The integral

∫ li(z) dz = zli(z) − Ei(2 log z)

By using integration by parts

z x
∫ li(z) dz = zli(z) − ∫ dx
0 log(x)
In the integral let −2 log(x) = t

∞ e−t
∫ li(z) dz = zli(z) + ∫ dt = zli(z) − Ei(2 log z)
−2 log(z) t

17.2 Example

Prove that

1
∫ li(x) dx = − log(2)
0

proof

Let the following

1 x e−a log(t) dt
I(a) = ∫ ∫ dx
0 0 log(t)
Differentiate with respect to a

1 x
I ′ (a) = − ∫ ∫ t−a dt dx
0 0

127
 1 1
I ′ (a) = 1−a
∫ x dx
a−1 0
Which reduces to

1 1 1
I ′ (a) = = −
(a − 1)(2 − a) 2 − a 1 − a
Integrate with respect to a

1−a
I(a) = log ( )+C
2−a
Take the limit a → ∞ we get C = 0

1 x e−a log(t) dt 1−a

∫ ∫ dx = log ( )
0 0 log(t) 2−a
Let a → 0

1 1
∫ li(x) dx = log ( ) = − log(2)
0 2

17.3 Find the integral

1
∫ xp−1 li(x) dx
0

solution

Let the following

1 x e−a log(t) dt
I(a) = ∫ xp−1 ∫ dx
0 0 log(t)
Differentiate with respect to a

1 x
I ′ (a) = − ∫ xp−1 ∫ t−a dt dx
0 0

1 1
I ′ (a) = ∫ x
p−a
dx
a−1 0
Which reduces to

1 1 1 1
I ′ (a) = = { − }
(a − 1)(p − a + 1) p p − a + 1 1 − a
Integrate with respect to a

128
 1 1−a
I(a) = log ( )+C
p p−a+1
Take the limit a → ∞ we get C = 0

1 x e−a log(t) dt 1 1−a

∫ xp−1 ∫ dx = log ( )
0 0 log(t) p p−a+1
Let a → 0

1 x e−a log(t) dt 1 1 1
∫ xp−1 ∫ dx = log ( ) = − log (p + 1)
0 0 log(t) p p+1 p

17.4 Find the integral

1 1
∫ li ( ) sin(a log(x)) dx
0 x
proof

Let the following

e−b log(t) dt
1
1 x
I(b) = ∫ sin(a log(x)) ∫ dx
0 0 log(t)
Differentiate with respect to b

1
1
I ′ (b) = − ∫ t−b dt dx
x
sin(a log(x)) ∫
0 0

1 1
I ′ (b) = b−1
∫ x sin(a log(x)) dx
b−1 0
Let log(x) = −t

1 ∞
I ′ (b) = ∫ e−tb sin(at) dt
1−b 0
Using the Laplace transform

a
I ′ (b) =
(b − 1)(a2 + b2 )
Integrate with respect to b

a log(a2 + b2 ) − a log(b − 1)2 + 2 arctan(b/a)

I(b) = +C
2a2 + 2
Let b → ∞

129
 π
0= +C
2(a2 + 1)
Hence we have

e−b log(t) dt a log(a2 + b2 ) − a log(b − 1)2 + 2 arctan(b/a)

1
1 x π
∫ sin(a log(x)) ∫ dx = −
0 0 log(t) 2a2 + 2 2(a2 + 1)
Let b → 0

1 a log(a2 ) π 1 π
∫ sin(a log(x))li(x) dx = − = (a log(a) − )
0 2a2 + 2 2(a2 + 1) a2 + 1 2

17.5 Example

Find the integral

1 li(x) 1
∫ logp−1 ( ) dx
0 x x
proof

Let the following

1 1 x e−a log(t) 1
I(a) = ∫ [∫ dt] logp−1 ( ) dx
0 x 0 log(t) x
Differentiate with respect to a

1 1 x 1
I ′ (a) = − ∫ [∫ t−a dt] logp−1 ( ) dx
0 x 0 x

1 1
p−1 1
I ′ (a) = −a
∫ x log ( ) dx
a−1 0 x
Let − log(x) = t

1 ∞
I ′ (a) = ∫ e−(1−a)t tp−1 dt
a−1 0

Γ(p) Γ(p)
I ′ (a) = − =−
(1 − a)(1 − a)p (1 − a)p+1
Integrate with respect to a

Γ(p)
I(a) = −
p(1 − a)p

130
Let a → 0, Hence

1 li(x) 1 Γ(p)
∫ logp−1 ( ) dx = −
0 x x p

17.6 Example

Prove that

∞ 1 π
∫ li ( ) logp−1 (x) dx = − Γ(p)
1 x sin(πp)
proof

Let the following

∞
I(a) = ∫ li (x−a ) logp−1 (x) dx
1

Differentiate with respect to a

x−a
−a
d d x dt
li (x−a ) = ∫ =
da da 0 log(t) a
Hence we have

1 ∞
I ′ (a) = ∫ x−a logp−1 (x) dx
a 1
Let log(x) = t

1 ∞
I ′ (a) = ∫ e−(a−1)t tp−1 dt
a 0
Using the Laplace transform

1
I ′ (a) = Γ(p)
a(a − 1)p
Take the integral

∞ ∞ 1
∫ I ′ (a)da = Γ(p) ∫ da
1 1 a(a − 1)p
The left hand-side

∞ 1
I(∞) − I(1) = Γ(p) ∫ da
1 a(a − 1)p

131
Now since I(∞) = 0

∞ 1
I(1) = −Γ(p) ∫ da
1 a(a − 1)p
Which implies that

∞ 1 ∞ 1
∫ li ( ) logp−1 (x) dx = −Γ(p) ∫ da
1 x 1 a(a − 1)p
Now let t = a − 1

∞ t−p
∫ dt
0 t+1
Using the beta integral x + y = 1 and x − 1 = −p which implies that x = 1 − p, y = p

Hence we have

∞ t−p π
∫ dt = β(p, 1 − p) = Γ(p)Γ(1 − p) =
0 t+1 sin(πp)
Finally we get

∞ 1 π
∫ li ( ) logp−1 (x) dx = − Γ(p)
1 x sin(πp)

132
18 Clausen functions

18.1 Definition

Define

⎧
⎪
⎪
⎪ ∞ sin(kθ)
⎪∑k=1
⎪ km
m is even
clm (θ) = ⎨
⎪
⎪
⎪ ∞
⎪ cos(kθ)
⎩∑k=1
⎪ km
m is odd

18.2 Duplication formula

clm (2θ) = 2m−1 (clm (θ) − (−1)m clm (π − θ))

proof

If m is even then

∞ ∞
sin(kπ − kθ) sin(kθ)
clim (π − θ) = ∑ m
= − ∑ (−1)k
k=1 k k=1 km
This implies

∞
sin(kθ) ∞ sin(kθ) 1 ∞ sin(2kθ)
cl2 (θ) + clim (π − θ) = ∑ (−1)k m
+∑ m
= m−1 ∑
k=1 k k=1 k 2 k=1 km
This implies that

clm (2θ) = 2m−1 (clm (θ) − clm (π − θ))

If m is odd then

∞
cos(kπ − kθ) ∞ cos(kθ)
clim (π − θ) = ∑ m
= ∑ (−1)k
k=1 k k=1 km

∞
cos(kθ) ∞ k cos(kθ) 1 ∞ sin(2kθ)
∑ + ∑ (−1) = ∑
k=1 km k=1 km 2m−1 k=1 k m
Which implies that

clm (2θ) = 2m−1 (clm (θ) + clm (π − θ))

Collecting the results we have

clm (2θ) = 2m−1 (clm (θ) − (−1)m clm (π − θ))

133
18.3 Example

Find the integral, for m is even

π
∫ clm (θ)dθ
0

solution

Using the series representation

π ∞ sin(kθ)
∫ ∑ dθ
0 k=1 km
Swap the integral and the series

∞ π
1
∑ m ∫ sin(kθ)dθ
k=1 k 0

The integral

π 1 π
−(−1)k + 1
∫ sin(kθ)dθ = − [ cos(kθ)] =
0 k 0 k
We get the summation

∞
−(−1)k + 1
∑ = ζ(m + 1) + η(m + 1)
k=1 k m+1
Now use that

η(s) = (1 − 21−s )ζ(s)

∞
−(−1)k + 1
∑ m+1
= ζ(m + 1) + (1 − 2−m )ζ(m + 1) = ζ(m + 1)(2 − 2−m )
k=1 k

18.4 Example

Find the integral for m is even

∞
∫ clm (θ)e−nθ dθ
0

Using the series representation

∞ ∞ sin(kθ) −nθ
∫ ∑ e dθ
0 k=1 km

134
Swap the integral and the series

∞ ∞
1
∑ m ∫ sin(kθ)e−nθ dθ
k=1 k 0

Using the Laplace transform we have

∞
1
∑ m−1 (k 2 + n2 )
k=1 k
Add and subtract k 2 and divide by n2

1 ∞ k 2 + n2 − k 2
∑
n2 k=1 k m−1 (k 2 + n2 )
Distribute the numerator

1 1 ∞ 1
ζ(m − 1) − ∑ m−3 2
n 2 2
n k=1 k (k + n2 )
Continue this approach to conclude that

j
1 (−1)j ∞ 1
∑(−1) − (2l − +
l−1
ζ(m 1)) ∑ m−(2j+1) 2
l=1 n 2l n 2j
k=1 k (k + n2 )
Let m − 2j − 1 = 1 which implies that j = m/2 − 1

m/2−1
1 (−1)m/2−1 ∞ 1
∑ (−1) − (2l − +
l−1
ζ(m 1)) ∑
l=1 n2l nm−2 k=1 k(k 2 + n2 )
Now let us look at the sum

∞ ∞
1 1 1 1
∑ =∑ { − }
k=1 k(k 2 + n ) k=1 2ink k − in k + in
2

Which can be written as

∞
1 1 ∞ 1 in −in
∑ = ∑ { + }
k=1 k(k + n ) 2n k=1 k k + in k − in
2 2 2

According to the digamma function

∞
1 1
∑ = 2 {γ + ψ(1 + in) + ψ(1 − in) + γ}
k=1 k(k 2 + n ) 2n
2

which simplifies to

∞
1 ψ(1 − in) + ψ(1 + in) + 2γ
∑ 2 + n2 )
=
k=1 k(k 2n2

135
Now we we can verify ψ(1 − in) = ψ(1 + in)

Which suggests that

ψ(1 + in) + ψ(1 − in) = 2R {ψ(1 + in)}

Hence we have the sum

∞
1 2R {ψ(1 + in)} + 2γ R {ψ(1 + in)} + γ)
∑ 2 + n2 )
= =
k=1 k(k 2n2 n2
This concludes to

m/2−1
ζ(m − (2l − 1)) R {ψ(1 + in)} + γ
∑ (−1) + (−1)m/2−1
l−1

l=1 n2l nm

136
19 Clausen Integral function

19.1 Definiton

We define

sin(kx)
cl2 (x) = ∑
k=1 k2

19.2 Integral representation

θ φ
cl2 (θ) = − ∫ log [2 sin ( )] dφ
0 2
proof

Start by the following

∞
eikθ ∞ cos(kθ) ∞
sin(kθ)
Li2 (eiθ ) = ∑ 2
= ∑ 2
+ i ∑
k=1 k k=1 k k=1 k2
By the integral definition of the dilogarithm

eiθ log(1 − x)
Li2 (eiθ ) − ζ(2) = − ∫ dx
1 x
Let x = eiφ

θ
Li2 (eiθ ) − ζ(2) = −i ∫ log(1 − eiφ )dφ
0

Let us look at the following

1 − eiφ = 1 − cos(φ) − i sin(φ) = 2 sin2 (φ/2) − 2i sin(φ/2) cos(φ/2)

Which simplifies to

1 − eiφ = 2 sin(φ/2) [sin(φ/2) − i cos(φ/2)] = 2 sin(φ/2)e−(i/2)(π−φ)

Hence our integral

θ
Li2 (eiθ ) − ζ(2) = −i ∫ log [2 sin(φ/2)e−(i/2)(π−φ) ] dφ
0

Use the complex logarithm properties

θ 1 1
Li2 (eiθ ) − ζ(2) = −i ∫ log [2 sin(φ/2)] dφ + (π − θ)2 − π 2
0 4 4

137
By equating the imaginary parts we have our result.

We can see the special value

∞
π sin(kπ/2) ∞ (−1)k
cl2 ( ) = ∑ =∑ =G
k2 k=0 (2k + 1)
2 2
k=1

Where G is the Catalan’s constant.

19.3 Duplication formula

Prove the following

cl2 (2θ) = 2(cl2 (θ) − cl2 (π − θ))

proof

We provide a proof using the integral representation

2θ t
cl2 (2θ) = − ∫ log [2 sin ( )] dt
0 2
Let t = 2φ

θ
−2 ∫ log [2 sin (φ)] dφ
0

Use the double angle identity

θ φ φ
−2 ∫ log [4 sin ( ) cos ( )] dφ
0 2 2
Separate the logarithms

θ φ θ φ
−2 ∫ log [2 sin ( )] dφ − 2 ∫ log [2 cos ( )] dφ
0 2 0 2
We can verify that

θ φ
cl2 (π − θ) = ∫ log [2 cos ( )] dφ
0 2
Hence

cl2 (2θ) = 2(cl2 (θ) − cl2 (π − θ))

Using that we get the value

138
 3π π
cl2 (3π) = 2cl2 ( ) − 2cl2 (− )
2 2
Since cl2 (3π) = 0
3π π π
cl2 ( ) = cl2 (− ) = −cl2 ( ) = −G
2 2 2

19.4 Example

Prove that

2π π5
∫ cl2 (x)2 dx =
0 90
Using the series representation

∞ ∞ 2π
1
∑∑ ∫ sin(kx) sin(nx)dx
k=1 n=1 (nk)
2 0

Consider the integral

2π 1 2π
∫ sin(kx) sin(nx) dx = ∫ cos((k − n)x) − cos((k + n)x) dx
0 2 0
We have two cases

If n = k then

1 2π
∫ 1 − cos(2nx) dx = π
2 0
If n ≠ k

2π
1 2π 1 sin((k − n)x) sin((k + n)x)
∫ cos((k − n)x) − cos((k + n)x) dx = [ − ] =0
2 0 2 k−n k+n 0

Hence we have

⎧
⎪
⎪
⎪
⎪ n≠k
2π ⎪0
∫ sin(kx) sin(nx) dx = ⎨
0 ⎪
⎪
⎪
⎪
⎪π n=k
⎩
We can write the series as

∞ ∞ ∞ ∞
1 1 1
∑∑ = ∑ + ∑
n=1 k=1 (nk) n≠k (nk)
2 2 4
n=1 n

Now since the integral n ≠ k goes to zero the result reduces to

139
 ∞
1 π5
π∑ 4
= πζ(4) =
n=1 n 90

19.5 Example

Prove that

π/2 7 π2
∫ x log(sin x) dx = ζ(3) − log 2
0 16 8
proof

π π
2 2 π2
I =∫ x log(sin x) dx = ∫ x log(2 sin x) − log(2)
0 0 8
The integral reduces to

π π
2 1 2
∫ x log(2 sin x)dx = ∫ Cl2 (2θ) dθ
0 2 0
π ∞
1 2 sin(2nθ)
= ∫ ∑ dθ
2 0 n=1 n2
1 (−1)n 1 1
=− ∑ 3
+ ∑ 3
4 n=1 n 4 n=1 n
7
= ζ(3)
16

Collecting that together we have

7 π2
I= ζ(3) − log(2)
16 8

19.6 Example

Prove that

π/4 π
∫ x cot(x) dx = − log(2) + G/2
0 8
proof

Start by integration by parts

π/4 π π/4
∫ x cot(x) dx = log(2) − ∫ log(sin x) dx
0 8 0

In the integral

140
 π/4
∫ log(sin x) dx
0

Let x → t/2

1 π/2
∫ log(sin t/2) dt
2 0
Which can be written as

1 π/2 1 π/2
∫ log(2 sin t/2) dt − ∫ log(2) dt
2 0 2 0
Using the Clausen integral function we get

1 π
− cl2 (π/2) − log(2)
2 4
Note that cl2 (π/2) = G

We deduce that

π/4 π
∫ log(sin x) dx = −G/2 − log(2)
0 4
Collecting the results we have

π/4 π π π
∫ x cot(x) dx = log(2) + G/2 − log(2) = − log(2) + G/2
0 8 4 8

19.7 Second Integral representation

Prove that

1 log(x)
cl2 (θ) = − sin(θ) ∫ dx
0 x2 − 2 cos(θ)x + 1
proof

Note that

x2 − 2 cos(θ)x + 1 = x2 − (eiθ + e−iθ )x + 1 = (x − eiθ )(x − e−iθ )

This implies

1 1 1 1
= iθ −iθ
{ − }
x2 − 2 cos(θ)x + 1 e − e x−e iθ x − e−iθ

141
Note that eiθ − e−iθ = 2i sin(θ)

1 1 1 1
= { − }
x2 − 2 cos(θ)x + 1 2i sin(θ) x − e iθ x − e−iθ
Now use the geometric series

∞ ∞
1 −1 k −i(k+1)θ
= { ∑ x e − ∑ xk ei(k+1)θ }
x2 − 2 cos(θ)x + 1 2i sin(θ) k=0 k=0

∞
1 1
= k−1
∑ x sin(kθ)
x2 − 2 cos(θ)x + 1 sin(θ) k=1
That implies

1 ∞ 1
log(x)
− sin(θ) ∫ dx = − ∑ sin(kθ) ∫ xk−1 log(x) dx = cl2 (θ)
0 x − 2 cos(θ)x + 1
2
k=1 0

19.8 Example

Find the value of

2π
cl2 ( )
3
proof

Use the second integral representation

√
3 1 log(x)
cl2 (2π/3) = − ∫ dx
2 0 x2 + x + 1
Use that

x3 − 1 = (x − 1)(x2 + x + 1)

Hence

√
3 1 x−1
cl2 (2π/3) = − ∫ log(x)dx
2 0 x3 − 1
Let x3 = t

1 1 t1/3−1 (t1/3 − 1)
cl2 (2π/3) = − √ ∫ log(t)dx
6 3 0 t−1
Note that

142
 1 xs−1
ψ ′ (s) = ∫ log(x) dx
0 1−x
We deduce that

1
cl2 (2π/3) = − √ (ψ ′ (2/3) − ψ ′ (1/3))
6 3

143
20 Barnes G function

20.1 Definition
z + z 2 (1 + γ) ∞ z n z2
G(z + 1) = (2π)z/2 exp (− ) ∏ {(1 + ) exp ( − z)}
2 n=1 n 2n

20.1.1 Functional equation

Prove that

G(z + 1) = Γ(z)G(z)

proof

From the series representation we have

G(z + 1) √ γ ∞ k+z k 2z − 1 − 2k
= 2π exp (−z − γz + ) ∏ ( ) exp ( ).
G(z) 2 k=1 k + z − 1 2k
This can be written as

G(z + 1) √ γ ∞ k+z k 1 + 2k z e−γz ∞ z −1 ( z )

= z 2π exp (−z + ) ∏ ( ) exp (− ) (1 + ) ∏ (1 + ) e k
G(z) 2 k=1 k + z − 1 2k k z k=1 k
Use the definition of the gamma function

G(z + 1) √ γ ∞ k+z k 1 + 2k z
= zΓ(z) 2π exp (−z + ) ∏ ( ) exp (− ) (1 + )
G(z) 2 k=1 k + z − 1 2k k
It suffices to prove that

√ γ ∞ k+z k 1 + 2k z
z 2π exp (−z + ) ∏ ( ) exp (− ) (1 + ) = 1
2 k=1 k + z − 1 2k k
or

∞
k+z k 1 + 2k z exp (z − γ2 )
∏( ) exp (− ) (1 + ) = √
k=1 k+z−1 2k k z 2π
Start by

N
k+z k 1 + 2k z
lim ∏ ( ) exp (− ) (1 + )
N →∞ k=1 k+z−1 2k k
Notice

144
 ∏k=1 (k + z)k ∏k=1 (1 + kz )
N N
N
k+z k z
∏( ) (1 + ) =
k=1 k+z−1 k ∏k=1 (k + z − 1)k
N

∏k=1 (k + z)k ∏k=1 (k + z)

N N
= −1
zN ! ∏Nk=1 (k + z)
k+1

−1 N −1
(N + z)N +1 ∏N
k=1 (k + z) ∏k=1 (k + z)
k
= −1
k=1 (k + z)
zN ! ∏N k+1

−1
(N + z)N +1 ∏N
k=1 (k + z)
k+1
= −1
k=1 (k + z)
zN ! ∏N k+1

(N + z)N +1
=
zN !

The second product

N
1 + 2k N
1 + 2k
) = e− 2 HN −N
1
∏ exp (− ) = exp (− ∑
k=1 2k k=1 2k
Hence we have the following
(N + z)N +1
e− 2 HN −N
1

zN !
According to Stirling formula we have

(N + z)N +1 (N + z)N +1 1
e− 2 HN −N ∼ e− 2 HN −N
1 1
×√
zN ! z(N /e)N
2πN
By some simplifications we have

e− 2 (HN −log N ) exp (− γ2 + z)

1
z z N 1
(1 + ) × (1 + ) × √ ∼ √
z N N 2π z 2π
Where we used that

lim Hn − log(n) = γ
n→∞

and

z N
lim (1 + ) = ez
n→∞ N

20.2 Reflection formula

G(1 − z) sin(πz) cl2 (2πz)
log { } = z log ( )+
G(1 + z) π 2π

145
proof

Start by the series expansion

exp ( z−z 2(1+γ) ) ∏∞

2 n
−z/2 2

G(1 − z) (2π) n=1 {(1 − n ) exp ( 2n + z)}

z z
=
G(1 + z) (2π)z/2 exp (− z+z2 (1+γ) ) ∏∞ {(1 + z )n exp ( z2 − z)}
2 n=1 n 2n

This simplifies to

∞
G(1 − z) (n − z)n 2z
= (2π)−z ez ∏ e
G(1 + z) n=1 (n + z)
n

Take the log of both sides

∞
G(1 − z) (n − z)n 2z
log { } = −z log(2π) + z + log { ∏ e }
G(1 + z) n=1 (n + z)
n

Let the following

∞ ∞
(n − z)n 2z
f (z) = log { ∏ e } = ∑ n log(n − z) − n log(n + z) + 2z
n=1 (n + z)
n
n=1

Differentiate with respect to z

∞ ∞
−n n −n(n + z) − n(n − z) + 2(n2 − z 2 )
f ′ (z) = ∑ − +2= ∑
n=1 n − z n+z n=1 n2 − z 2
Hence we have

∞
−2z 2
∑
n=1 n − z
2 2

Now we can use the following

∞
−2z 2
zπ cot πz = 1 + ∑
n=1 n − z
2 2

Hence we conclude that

f ′ (z) = zπ cot πz − 1

Integrate with respect to z

z
f (z) = ∫ xπ cot(πx) dx − z
0

Hence we have

146
 G(1 − z) z
log { } = −z log(2π) + ∫ zπ cot(πx) dx
G(1 + z) 0

Now use integration by parts for the integral

z z
∫ xπ cot(πx) dx = z log(sin πz) − ∫ log(sin πx)dx (1)
0 0
z
= z log(2 sin πz) − ∫ log(2 sin πx)dx (2)
0
1 2πz x
= z log(2 sin πz) − ∫ log (2 sin ) dx (3)
2π 0 2
cl2 (2πz)
= z log(2 sin πz) + (4)
2π
(5)

That implies

G(1 − z) cl2 (2πz) sin(πz) cl2 (2πz)

log { } = z log(2 sin πz) − z log(2π) + = z log ( )+
G(1 + z) 2π π 2π

20.3 Values at positive integers

Prove that

n−1
G(n) = ∏ Γ(k)
k=1

proof

It can be proved by induction. For G(1) = 1, suppose

n−1
G(n) = ∏ Γ(k)
k=1

We want to show

n
G(n + 1) = ∏ Γ(k)
k=1

By the functional equation

n−1 n
G(n + 1) = Γ(n)G(n) = Γ(n) ∏ Γ(k) = ∏ Γ(k)
k=1 k=1

147
20.4 Relation to Hyperfactorial function

We define the hyperfactorial function as

n
H(n) = ∏ k k
k=1

Prove for n is a positive integer

(n!)n
G(n + 1) =
H(n)
proof

We can prove it by induction for n = 0 we have, G(1) = 1,

suppose that

Γ(n)n−1
G(n) =
H(n − 1)
we want to show that

Γ(n)n−1
G(n + 1) = Γ(n)G(n) = Γ(n)
H(n − 1)
Notice that

n
n−1
∏ k k H(n)
H(n − 1) = ∏ k k = =
nn nn
We deduce that

Γ(n)n × nn (n!)n
G(n + 1) = Γ(n)G(n) = =
H(n) H(n)

20.5 Loggamma integral

Prove that

z z z(z − 1)
∫ log Γ(x)dx = log(2π) + + z log Γ(z) − log G(z + 1)
0 2 2
proof

Take the log to the series representation

z z + z 2 (1 + γ) ∞ z z2
log G(z + 1) = log(2π) − + ∑ n log (1 + ) + −z
2 2 n=1 n 2n

148
Let the following

∞
z z2
f (z) = ∑ n log (1 + )+ −z
n=1 n 2n
Differentiate with respect to z

∞ ∞
n z z2
f ′ (z) = ∑ + −1= ∑
n=1 z + n n n=1 n(n + z)

Now use the following

1 ∞ z
ψ(z) = −γ − +∑
z n=1 n(n + z)
which implies that

∞
z2
∑ = zψ(z) + γz + 1
n=1 n(n + z)

Hence we have

f ′ (z) = zψ(z) + γz + 1

Integrate with respect to z

z γz 2
f (z) = ∫ xψ(x)dx + +z
0 2
which implies that

z γz 2
f (z) = z log Γ(z) − ∫ log Γ(x)dx + +z
0 2
Hence we have

z z + z 2 (1 + γ) z γz 2
log G(z + 1) = log(2π) − + z log Γ(z) − ∫ log Γ(x)dx + +z
2 2 0 2
By some rearrangements we have

z z z(z − 1)
∫ log Γ(x)dx = log(2π) + + z log Γ(z) − log G(z + 1)
0 2 2

149
20.6 Glaisher-Kinkelin constant

We define the Glaisher-Kinkelin constant as

H(n)
A = lim
n→∞ nn2 /2+n/2+1/12 e−n2 /4

20.7 Relation to Glaisher-Kinkelin constant

Prove that

G(n + 1) e1/12
lim =
n→∞ (2π)n/2 nn /2−1/12 e−3n /4
2 2
A
proof

Use the relation to the hyperfactorial function

(n!)n
lim
n→∞ H(n)(2π)n/2 nn /2−1/12 e−3n /4
2 2

Now use the Stirling approximation

2
+n/2 −n2 +1/12
(n!)n ∼ (2π)n/2 nn e

Hence we deduce that

(2π)n/2 nn +n/2 e−n +1/12

2 2
1
lim ×
n→∞ H(n) (2π)n/2 nn2 /2−1/12 e−3n2 /4
By simplifications we have

2
/2+n/2+1/12 −n2 /4
nn e e1/12
e1/12 lim =
n→∞ H(n) A

20.8 Example

Prove that

π2
ζ ′ (2) = (log(2π) + γ − 12 log A)
6
We already proved that

G(n + 1) 1
log [ ]∼ − log A
(2π)n/2 nn2 /2−1/12 e−3n2 /4 12

150
Let the following

G(n + 1)
f (n) = log [ ]
(2π)n/2 nn2 /2−1/12 e−3n2 /4
Use the series representation of the Barnes functions

⎡ ⎤
⎢ (2π)n/2 exp (− n+n 2(1+γ) ) ∏∞
2
n k n2
⎢ k=1 {(1 + k ) exp ( 2k − n)} ⎥
⎥
f (n) = log ⎢ ⎥
⎢ (2π) n/2 nn2 /2−1/12 e−3n2 /4 ⎥
⎢ ⎥
⎣ ⎦
Which reduces to

n + n2 (1 + γ) ∞ n n2 n2 1 3n2
f (n) = − + ∑ {k log (1 + ) + − n} − ( − ) log(n) +
2 k=1 k 2k 2 12 4
Differentiate with respect to n

1 n 1 3n
f ′ (n) = − − n − γn + nψ(n) + γn + 1 − n log(n) − + +
2 2 12n 2
Note that we already showed that

d ∞ n n2
∑ {k log (1 + ) + − n} = nψ(n) + γn + 1
dn k=1 k 2k
By simplifications we have

1 1
f ′ (n) = nψ(n) − n log(n) + +
12n 2
Now use that

1 ∞ zdz
ψ(n) = log(n) − − 2∫ dz
2n 0 (n2 + z 2 )(e2πz − 1)
Hence we deduce that

∞ nzdz 1
f ′ (n) = −2 ∫ dz +
0 (n2 + z 2 )(e2πz − 1) 12n
Integrate with respect to n

∞ z log(n2 + z 2 ) 1
f (n) = − ∫ dz + log(n) + C
0 (e 2πz − 1) 12
Take the limit n → 0

1 ∞ z log(z 2 )
C = lim f (n) − log(n) + ∫ dz
n→0 12 0 (e2πz − 1)

151
Hence we have the limit

G(n + 1) 1 G(n + 1)
lim − log(n) = lim log =0
n→0 (2π)n/2 nn2 /2−1/12 e−3n2 /4 12 n→0 (2π) nn2 /2 e−3n2 /4
n/2

Hence we see that

∞ z log(z)
C = 2∫ dz
0 (e2πz − 1)
Finally we have

∞ z log(n2 + z 2 ) 1 ∞ z log(z)
f (n) = − ∫ dz + log(n) + 2 ∫ dz
0 (e 2πz − 1) 12 0 (e2πz − 1)

z2
∞ z log (1 + n2
) ∞ z 1 ∞ z log(z)
f (n) = − ∫ dz − log(n2 ) ∫ dz + log(n) + 2 ∫ dz
0 (e2πz − 1) 0 (e2πz − 1) 12 0 (e2πz − 1)
Also we have

∞ z 1
∫ dz =
0 (e2πz − 1) 24
That simplifies to

z2
∞ z log (1 + n2
) ∞ z log(z)
f (n) = − ∫ dz + 2 ∫ dz
0 (e2πz − 1) 0 (e2πz − 1)
Take the limit n → ∞

∞ z log(z) 1
2∫ dz = − log A
0 (e 2πz − 1) 12
Now use that

∞ z log(z) ∞ z log(z) 1
2∫ dz = 2 ∫ × dz
0 (e 2πz − 1) 0 e 2πz 1 − e−2πz
∞ ∞
=2∑∫ e−2πz(n+1) z log(z) dz
n=0 0
∞
ψ(2) − log(2π) + log(n)
=∑
n=1 2π 2 n2
(ψ(2) − log(2π))ζ(2) + ζ ′ (2)
=
2π 2

Hence we conclude that

1
ζ ′ (2) = (log(2π) − ψ(2))ζ(2) + 2π 2 ( − log A) = ζ(2)(log(2π) + γ − 12 log A)
12

152
20.9 Example

Prove that

1
ζ ′ (−1) = − log A
12
proof

Start by

m
m1−s m−s sm−s−1
ζ(s) = lim ( ∑ k −s − − + ) , Re(s) > −3.
m→∞
k=1 1−s 2 12
Differentiate with respect to s

m
m1−s m1−s m−s m−s−1 m−s−1
ζ ′ (s) = lim (− ∑ k −s log(k) − + log(m) + log(m) + − log(m))
m→∞
k=1 (1 − s) 2 1−s 2 12 12
Now let s → −1

m
m2 m2 m 1 1
ζ ′ (−1) = lim (− ∑ k log(k) − + log(m) + log(m) + − log(m))
m→∞
k=1 4 2 2 12 12
Take the exponential of both sides

2
/2+m/2−1/12 −m2 /4 2
/2+m/2−1/12 −m2 /4
ζ ′ (−1) mm e mm e e1/12
e =e 1/12
lim ∑m k log(k)
=e 1/12
lim =
m→∞ e k=1 m→∞ H(m) A
We conclude that

1
ζ ′ (−1) = − log A
12

20.10 Relation to Howrtiz zeta function

Prove that

log G(z + 1) − z log Γ(z) = ζ ′ (−1) − ζ ′ (−1, z)

proof

Start by the following

z −s z 1−s ∞ sin(s arctan(x/z))

ζ(s, z) = + + 2∫ dx
2 s−1 0 (z 2 + x2 )s/2 (e2πx − 1)
Take the derivative with respect to s and s → −1

153
 z log(z) z 2 log(z) z 2 ∞ x log(x2 + z 2 ) + 2z arctan(x/z)
ζ ′ (−1, z) = − + − +∫ dx
2 2 4 0 (e2πx − 1)
Now use that

1 ∞ x
ψ(z) = log(z) − − 2∫ dx
2z 0 (z 2 + x2 )(e2πx − 1)
Which implies that

∞ 2zx 1
∫ dx = z log(z) − − zψ(z)
0 (z 2 + x2 )(e2πx − 1) 2
By taking the integral

∞ x log(x2 + z 2 ) − x log(x2 ) z z z
∫ dx = ∫ x log(x) dx − ∫ xψ(x) dx −
0 (e 2πx − 1) 0 0 2
Which simplifies to

∞ x log(x2 + z 2 ) z2 1 2 z z
′
∫ dx = ζ (−1) − + z log(z) − z log Γ(z) + ∫ log Γ(x) dx −
0 (e 2πx − 1) 4 2 0 2
Also we have

∞ x 1
2∫ dx = log(z) − − ψ(z)
0 (x2 + z 2 )(e2πx − 1) 2z
By integration we have

∞ arctan(x/z) log(z)
2∫ dx = z + − z log(z) + log Γ(z) + C
0 (e2πx − 1) 2
Let z → 1 to evaluate the constant

∞ arctan(x/z) log(z) 1
2∫ dx = z + − z log(z) + log Γ(z) − log(2π)
0 (e 2πx − 1) 2 2
Multiply by z

∞ z arctan(x/z) z log(z) z
2∫ dx = z 2 + − z 2 log(z) + z log Γ(z) − log(2π)
0 (e2πx − 1) 2 2
Substitute both integrals in our formula

z z z(1 − z) ′ ′
∫ log Γ(x) dx = log(2π) + − ζ (−1) + ζ (−1, z)
0 2 2
We also showed that

154
 z z z(1 − z)
∫ log Γ(x) dx = log(2π) + + z log Γ(z) − log G(z + 1)
0 2 2
By equating the equations we get our result.

20.11 Example

Prove that

1
G ( ) = 21/24 π −1/4 e1/8 A−3/2
2
proof

We know that

log G(z) + log Γ(z) − z log Γ(z) = ζ ′ (−1) − ζ ′ (−1, z)

Note that

1
ζ (s, ) = (2s − 1)ζ(s)
2
Which implies that

1 log(2) 1
ζ ′ (−1, ) = ζ(−1) − ζ ′ (−1)
2 2 2
Hence we have

1 1 1 3 log(2)
log G ( ) + log Γ ( ) = ζ ′ (−1) − ζ(−1)
2 2 2 2 2
Using that we have

1
G ( ) = 21/24 π −1/4 e 2 ζ (−1)
3 ′

2
Note that

1
ζ(−1) = −
12
This can be proved by the functional equation of the zeta function.

155
21 Complex Analysis

21.1 Introduction to complex numbers

The idea of complex numbers originated from trying to solve polynomials like those in the form x2 +1 = 0.

If we do a simple algebra then it is easy to induce that x2 = −1 but we know that there is no real number

whose square is a negative number. Which implies that this polynomial has no solutions in the set of

real numbers. Hence, we need to expand the set of real numbers R to the set of complex numbers C. By

definition a complex number can be written in the form z = x + iy where x, y ∈ R. We say the real part of

R(z) = x and the imaginary part I(z) = y. We can think of i as a special symbol with the property i2 = −1
√
or i = −1. Now let us define some algebraic operations.

Let a = x1 + iy1 and b = x2 + iy2 be two complex numbers then we define the following operations

a ± b = (x1 + x2 ) ± i(y1 + y2 )

a × b = x1 x2 − y1 y2 + i(x1 y2 + x2 y1 )

For division we need that b ≠ 0 which by definition means both x2 and y2 can’t be zero at the same

time. Also we need to define the complex conjugate which is ā = x1 − iy1 then we can easily see that

a × ā = x21 + y12 . Using that we define division for a and b ≠ 0 as

a a b̄ ab̄
= × =
b b b̄ x22 + y22

21.2 Polar representation

We usually define complex numbers as the following

z = reiφ = r(cos(φ) + i sin(φ))

√
Where we define r = ∣z∣ = x2 + y 2 as the modulus and φ called the argument or arg(z). Geometrically

this can be seen as in Figure 1.

This representation allows a better representation of both multiplication and division. Let z1 = r1 eiφ1

and z2 = r2 eiφ2 be two complex numbers. Then this implies

z1 × z2 = r1 r2 ei(φ1 +φ2 )

z1 r1 i(φ1 −φ2 )
= e
z2 r2

156
 I{z}

z = x + iy

∣z∣

φ R{z}

Figure 1: Polar representation of a complex number

The absolute value of complex number or the length of the vector representation is an important property

of a complex number. Note that ∣z∣ = 1 implies that z is unit vector. One of the most important properties

is the triangle inequality which will be very handy for us in the coming sections

∣z1 + z2 ∣ ≤ ∣z1 ∣ + ∣z2 ∣

To prove it note that

∣z1 + z2 ∣2 = (z1 + z2 )(z1 + z2 ) = ∣z1 ∣2 + z1 z¯2 + z2 z¯1 + ∣z2 ∣2

= ∣z1 ∣2 + z1 z¯2 + z1 z¯2 + ∣z2 ∣2

= ∣z1 ∣2 + 2Re(z1 z¯2 ) + ∣z2 ∣2

Note also

Re(z1 z¯2 ) ≤ ∣z1 z¯2 ∣ = ∣z1 ∣∣z2 ∣

Finally we have

∣z1 + z2 ∣2 ≤ ∣z1 ∣2 + 2∣z1 z2 ∣ + ∣z2 ∣2 = (∣z1 ∣ + ∣z2 ∣)2

An easy corollary is the reverse triangle inequality

∣z1 + z2 ∣ ≥ ∣z1 ∣ − ∣z2 ∣

One can also deduce the following by induction on n

n n
∣ ∑ zk ∣ ≤ ∑ ∣zk ∣
k=1 k=1

157
Note that geometrically in terms of vectors the triangle inequality implies that the length of two vectors

is greater than the length of their sum. This also holds in triangles in Euclidean geometry where the sum

of lengths of two sides is greater than the third the length.

21.3 Complex functions

Complex functions are mapping f ∶ C → C. A function that is differentiable on a certain neighborhood of

the plain is called an analytic function. You can see a specialized complex analysis book to understand

these terms in details. Specifically the relation to Cauchy-Riemann equations.

21.3.1 Exponential function

The exponential function f = exp(z) is an entire function (analytic/differentiable in the whole plain). We

see that the function has no discontinuities. Also we have

ez = ex eiy = ex (cos(y) + i sin(y))

Which implies that ∣ez ∣ = ex . The function acts smoothly with the usual derivative (ez )′ = ez and the

expansion

∞
zk
ez = ∑
k=0 k!

One also can note that the function is periodic with period 2kπi.

21.3.2 Sine and Cosine and hyperbolic functions

These functions are also entire in the complex plane. They almost have the same properties as the real

counterparts but one essential difference is boundedness. We know in the real case that ∣ sin(x)∣ ≤ 1 but

it is not the case for ∣ sin(z)∣. You can see that since sin(z) = 1
2i
(eiz − e−iz ) . Now since ∣e−iz ∣ = ey we see

that ey is unbounded near infinity then sin(z) is also unbounded. You can deduce the same for cos(z).

Note that also d

dz
sin(z) = cos(z) also we have d
dz
cos(z) = − sin(z) . The expansions around zero are

∞
(−1)n 2k+1
sin(z) = ∑ z
k=0 (2k + 1)!
∞
(−1)n 2k
cos(z) = ∑ z
k=0 (2k)!

The hyperbolic functions can be defined in terms of the sine and cosine functions using the relations

158
 cosh(z) = cos(iz) , sinh(z) = −i sin(iz)

You can then deduce the derivatives and the series expansions around zero. The exponential represen-

tation is also useful. Note that

ez + e−z ez − e−z
cosh(z) = , sinh(z) =
2 2

21.3.3 Complex logarithm

Usually we think of logarithm as the inverse of the exponential function away from zero. As in real

analysis we have log(ex ) = x but it is completely risky to think the same for complex analysis. The

problem with complex exponential is its periodicity. Note that we have ez+2kπi = ez e2kπi = ez , hence the

function is not one to one. We have infinite values that map to the same value. This makes the complex

logarithm a multi-valued function with the definition

log(z) = log ∣z∣ + iarg(z)

Some authors might use ln(z) interchangeably with log(z). Here we usually use log(z) to denote the

natural complex logarithm. Let us see some examples

Example. Find log(−2) , log(i) and log(1 + i)

solution

The easiest way is to use the polar representation. Note that we have eiπ = −1 which implies 2eiπ+2kπi =

−2. Simply we have

log(−2) = log(2) + i(π + 2kπ)

Similarly we have

π π
log(i) = log(1) + i ( + 2kπ) = i ( + 2kπ)
2 2
√ iπ
Finally since 1 + i = 2e 4 which implies

√ π
log(1 + i) = log( 2) + i ( + 2kπ)
4
Properties

1. log(z1 z2 ) = log(z1 ) + log(z2 )

2. log(z n ) = n log(z)

159
 3. elog(z) = z

proof

1. Suppose that z1 = r1 eiφ1 , z2 = r2 eiφ2 and m = k + n

log(z1 z2 ) + 2nπi = log(r1 r2 ei(φ1 +φ2 ) )

= log(r1 r2 ) + i (φ1 + φ2 + 2mπ)

= log(r1 ) + i(π1 + 2nπ) + log(r2 ) + i(φ2 + 2kπ)

= log(z1 ) + log(z2 )

2. Similarly we have for z = reiφ

log(z n ) = log(rn einφ ) = n log(r) + i(nφ + 2knπ) = n log(z)

3. Finally we have

elog(z) = elog ∣r∣+i(φ+2kπ) = reiφ = z

Some properties don’t necessarily hold for example log(ez ) ≠ z since ez+2kπi = ez and that implies z =

z + 2kπi which is obviously can’t be true unless we choose k = 0. This suggests that we can make nice

properties by choosing proper values of the argument. This raises the concept of Branches of logarithm.

The definition log(z) = log ∣z∣ + i(φ + 2kπ) makes the complex logarithm multi valued as explained earlier.

Each value for k raises a different branch of the logarithm that makes it single valued. One interesting

branch is called Principal logarithm. We define it as the following

Log(z) = log ∣z∣ + iArg(z)

where we define −π < Arg(z) ≤ π. Note that we usually use log(z) to denote the principal value when it

is clear from the context. Clearly this makes the function single valued.

Example. Find Log(−2) and Log(1 + i)

solution

The main trick it to find the value of the argument that falls in the interval (−π, π]. Using that we

conclude

log(−2) = log(2) + iπ

Similarly we have

160
 √ π
log(1 + i) = log( 2) + i
4
The usual properties described in the previous section don’t usually carry on for the principal value. For

instance, Log(z1 z2 ) ≠ Log(z1 ) + Log(z2 ). Let z1 = z2 = −1 and note that

Log(−1 × −1) = log(1) = 0 ≠ Log(−1) + Log(−1) = 2Log(−1) = 2πi

Now let us define the concept of a Branch cut. In order to make the complex logarithm an analytic

function we have to make it differentiable in a certain neighborhood. Note that the principal value of the

logarithm by choosing a certain branch the function became one-to-one but this is not enough because

the function Log(z) doesn’t behave well on the negative real axis.

I{z}

π
R{z}

−π

Figure 2: Behaviour of the Principal Logarithm near the branch cut

We see from the graph in Figure 2 that the function approach different value as we approach the negative

real axis. This causes a discontinuity that prevents the function from being analytic on that line.

Theorem 1 The function with ∣z∣ ≠ 0

Log(z) = log ∣z∣ + iφ , φ ∈ (−π, π)

1
is analytic and the derivative is z
.

Using that we can expand the function along values away from the branch cut. For instance for ∣z∣ < 1

we have

∞
zk
log(1 − z) = − ∑
k=1 k

161
We are not going to go into details of the proofs of such claims because they require a more firm under-

standing of analytic functions.

21.4 Taylor and Laurent expansions

In this section we establish the foundations of contour integration by studying the theory of Laurent

expansion, residues and poles. It is preferable if you have a basic knowledge in real analysis because

many theorems carry on from the real case to the complex plane. We start by the following theorem

Theorem 2 Let f be an analytic function in a domain D and z0 be a point in D Then

∞
f (k) (z0 )
f (z) = ∑ (z − z0 )k
k=0 k!
is a valid representation in the largest circle with radius ∣z − z0 ∣ contained in D.

Example. Find the expansion of 1

z
around z0 = 2.

solution

Since the function is analytic in the whole plane except at the origin. Then we can use the Theorem to

deduce that

(−1)k k!
f (k) (2) =
2k+1
You can verify that by taking multiple derivatives and substitute the value of z0 = 2. Hence we deduce

that

1 ∞ (−1)k k!
=∑ (z − 2)k
z k=0 2k+1
And the radius of convergence is ∣z − 2∣ < 2 which represents a ball centered at z0 = 2 with radius less

than 2. As we see from the Figure 3.

Note that we cannot include zero because the function is not defined there. So the maximum circle is

that of radius 2.

Theorem 3 Let f be an analytic function in the punctured disk r < ∣z −z0 ∣ < R . Then f has a series representation

∞
f (z) = ∑ ak (z − z0 )k
k=−∞

The theorem is actually saying that if a function is not analytic on a point z0 but analytic around it

then we can expand the function using what we call a Laurent expansion. The difference between

162
 0 1 2 3 4

Figure 3: Radius of convergence

Laurent expansion and Taylor is that Laurent allows negative indices of the series. We can think of it as a

generalization of the Taylor theorem which is a special case where coefficients with negative indices are

zero.

Example Expand the function f (z) = 1

z(1−z)
in the domain 0 < ∣z∣ < 1.

solution

We need to rewrite the function as

∞
f (z) = ∑ ak (z − z0 )k
k=−∞

We need to expand both functions in the domain 0 < ∣z∣ < 1. Note that the function f (z) = 1
z
is a

Laurent expansion around zero in the punctured disk ∣z∣ > 0 what is remaining is to expand the function

f (z) = 1
1−z
in the domain ∣z∣ < 1. Hence the intersection of both domains is actually 0 < ∣z∣ < 1. Note that

∞
1
= ∑ zk
1 − z k=0
is a Taylor expansion valid in ∣z∣ < 1 hence we have

1 ∞ k ∞ k−1 ∞
f (z) = ∑z = ∑z = ∑ zk
z k=0 k=0 k=−1

which is exactly what we want.

sin(z)
Example Find the Laurent expansion of the function f (z) = z2
around z = 0.

solution

Note that
∞
(−1)k 2k+1
sin(z) = ∑ z
k=0 (2k + 1)!

163
Which is a valid expansion in the domain z ∈ C. Also 1
z2
is a valid expansion in the domain ∣z∣ > 0. Hence

∞
(−1)k 2k−1
f (z) = ∑ z
k=0 (2k + 1)!

is a Laurent expansion in the domain ∣z∣ > 0 .

ez
Example. Find the Laurent expansion of the function f (z) = z−1
around z = 1.

solution

First note that around z = 0 we have

∞
zk
ez = ∑
k=0 k!

Now let z → z − 1, then

∞
1
ez−1 = ∑ (z − 1)k
k=0 k!
Which implies that

1 ∞ 1
ez = ∑ (z − 1)
k
e k=0 k!
is a valid expansion in the whole domain C. Finally we have

1 ∞ 1
f (z) = ∑ (z − 1)
k−1
e k=0 k!
in the punctured disk ∣z − 1∣ > 0.

21.5 Poles and residues

Here we discuss the of types of points where a complex function is not defined at. Usually they are called

singularities. Mainly we start by discussing the concept of Isolated Singularities.

An isolated singular point is a point where the function is analytic in the punctured disk by removing

that point.

1. Removable singularity A singular point is removable if limz→z0 f (z) exists. Equivalently, if all the

negative indexed terms in the Laurent expansion are zeros,namely a−k = 0 for all k > 0.

2. Poles a function f has a pole of order m > 0 if a−m ≠ 0 and a−k = 0 for all k > m.

3. Essential singularity The Laurent expansion has an infinite number of non-positive indexed terms.

Example. Classify the following singularities

164
 sin(z)
1. f (z) = z
at z = 0

ez
2. f (z) = (z−1)2
at z = 1

3. f (z) = e1/z at z = 0

solution

1. This is a removable singularity

sin(z)
lim =1
z→0 z
2. By expanding around 1 we have

1 ∞ 1
f (z) = ∑ (z − 1)
k−2
e k=0 k!
This implies 1 is a pole of order 2.

3. Note that

∞
1
e1/z = ∑ k
k=0 k!z
Then z = 0 is an essential singularity.

Theorem 4 The function f has a pole of order m at z0 if and only if

g(z)
f (z) =
(z − z0 )m

where g(z0 ) ≠ 0.

The proof is done by writing the Laurent expansion of f and take (z − z0 )−m as a common factor.

165
Residues are closely related to the Laurent expansion of a function. They will become really handy when

we start discussion about contours. We usually use the following notation Res(f, z0 ) as the residue of

the function f (z) at z = z0 .

Definition. Let f (z) be an analytic function in a domain D with the Laurent expansion
∞
f (z) = ∑ ak (z − z0 )k
k=−∞

around z = z0 . Then we define

Res(f, z0 ) = a−1

Where a−1 is the cofficient of (z − z0 )−1

Example. Find the residues of the following functions around z = 0

sin(z)
1. f (z) = z

ez
2. f (z) = z2

3. f (z) = e1/z

solution

1. Note that since f has a removable singularity and since a−k = 0 for all k > 0 then we have Res(f, 0) = 0.

2. It is simple to see that

1 ∞ 1
f (z) = ∑ zk
e k=−2 (k + 2)!
at k = −1 we conclude

1
Res(f, 0) =
e
3. Using the Taylor expansion of ez and letting z → 1/z and rewrite the expansion as

‘
0
1
e1/z = ∑ zk
k=−∞ (−k)!

Hence we have

Res(f, 0) = 1

Usually we don’t have to find the Laurent expansion in order to find the residues at a specific point. Let

us first note that the residues of an entire function is zero at any point. Now let us see theorems that help

us find the residues much more easier

166
Theorem 5 Suppose z0 is a pole of order m of f (z) then

1 dm−1
Res(f, z0 ) = lim (z − z0 )m f (z)
z→z0 (m − 1)! dz m−1

proof

Write the Laurent expansion of f

a−m a−1
f (z) = +⋯+ +⋯
(z − z0 ) m (z − z0 )
Multiply by (z − z0 )m

(z − z0 )m f (z) = a−m + ⋯ + a−1 (z − z0 )m−1 + ⋯

Take the derivative of both sides

dm−1
(z − z0 )m f (z) = (m − 1)!a−1 + O((z − z0 ))
dz m−1
We finish by Letting z → z0

1 dm−1
lim (z − z0 )m f (z) = a−1
(m − 1)! z→z0 dz m−1

Theorem 6 Suppose we have the following function

f (z)
h(z) =
g(z)
where z0 is a simple zero of g(z) and f (z0 ) ≠ 0 then

f (z0 )
Res(h, z0 ) =
g ′ (z0 )

proof

Since z0 is a simple pole of h then

f (z) f (z) f (z0 )

Res(h, z0 ) = lim (z − z0 ) = lim =
z→z0 g(z) z→z0 g(z)−g(z0 ) g ′ (z0 )
z−z0

167
21.6 Integration around paths

Since in the complex plane C every complex number z ∈ C has two components real and complex. We

can not use the same process as in real analysis to evaluate integrals. Now, integration happens around

paths. We will be concerned about a family paths. Mainly those curves that have continuous derivatives.

These are called smoothed curves. Mainly these curves don’t have a peak where the derivative doesn’t

exist, hence named smooth. We can attach a finite number of smooth curves to obtain piece-wise smooth

curves. Also those curves must be simple, that means it doesn’t cross itself.

y y

b
C1 C2 C3 C

a b x a x

(a) Piece-wise smooth curve (b) Smooth curve

Figure 4: Examples of smooth curves

Note that the orientation of a curve is essential to the evaluation of a path integral. Usually we work

with anti-clockwise orientations.

Theorem 7 Let f be continuous on a smooth curve γ given by γ(t) = x(t) + iy(t) , where t ∈ [a, b] then

b
∫ f (z)dz = ∫ f (γ(t))γ ′ (t)dt
γ a

Example

Integrate the function f (z) = z 2 around a circle of radius 1 traced counter-clockwise.

solution

Denote the curve γ(t) = eit where t ∈ [0, 2π). Now since z 2 is continuous on the simple smooth path we

have

2π 1 2πi
∫ z dz = i ∫
2
e2it eit dt = (e − e0 ) = 0
γ 0 3
This seems to imply that if we integrate around a simple closed curve the integration will be always zero.

This is not always the case, consider integrating z −1 around the same curve

2π 2π
−1
∫ z dz = i ∫ e−it eit dt = i ∫ dt = 2πi
γ 0 0

168
One might notice that the function f (z) = z −1 has a simple pole inside the closed path while f (z) = z 2

is analytic in and on the contour. This is indeed the case as the Cauchy’s theorem illustrates. We first

define simply connected domains. A domain is simply connected if every closed contour/path can be

shrunk into one point. That implies the domain can’t contain holes. Let us know take a look at the

Cauchy-Grousat Theorem.

Theorem 8 Let f is an analytic in a simpley connected domain D. Then for every simple closed contour C in D

we have

∮ f (z)dz = 0
C

Example. Choose a closed simple contour γ(t) in the complex plain then find the integration of f (z) =

(z 2 + 3z + 1)n where n ≥ 1 around that contour.

solution

We don’t have to worry about parametrizing the integral since the function is entire. We apply the

theorem directly to conclude that

∫ f (z) = ∫ (z + 3z + 1) = 0
2 n
γ C

One might think that regardless of the path of integration we can deduce that Independence of paths

which is explained in the theorem.

Theorem 9 For any analytic function f in a connected domain D for any simple paths γ1 , γ2 inside D we have

∫ f (z)dz = ∫ f (z)dz
γ1 γ2

Where γ1 , γ2 have the same initial and final points.

Example Evaluate the function f (z) = z around the following path

a b x

By independence of paths we can use a simple contour (line) that connect the same points a and b. Define

γ(t) = bt + (1 − t)a where t ∈ [0, 1].

169
 1 b2 − a2 b2 − a2
∫ zdz = (b − a) ∫ bt + (1 − t)a dt = b2 − a2 − =
γ 0 2 2
One can also use the anti-derivative

b
z2 b2 − a2
∫ zdz = [ ] =
γ 2 a 2
This is indeed true by the fundamental theorem.

Theorem 10 For any analytic function f in a connected domain D for any simple paths γ(t), t ∈ [a, b] inside D

we have

∫ f (z)dz = F (γ(b)) − F (γ(a))

γ

So if the function is analytic and by the independence of path theorem we can find the anti-derivative

and use the initial and final points.

Now we look at the approach to evaluate functions that have poles inside the contour. First let us look

at the most important theorem , Cauchy’s Integral formula.

Theorem 11 Let f be an analytic function in a simply connected domain D then for any contour γ ∈ D we have

f (z)
∮ dz = 2πi f (z0 )
γ z − z0

proof

First we can deform the contour γ into a circular contour C. Then we have the following

f (z) f (z) dz f (z) − f (z0 )

∫ dz = ∫ = f (z0 ) ∫ +∫ dz
γ z − z0 C z − z0 C z − z0 C z − z0
Note the second function is analytic since it has a removable singularity at z = z0 hence

f (z) − f (z0 )
∫ dz = 0
C z − z0
Hence we conclude

f (z) dz
∫ dz = f (z0 ) ∫ = 2πi f (z0 )
γ z − z0 C z − z0

Using that we can conclude the Cauchy’s Integral formula for Derivatives

f (z) 2πi (n)

∫ dz = f (z0 )
γ (z − z0 )n+1 n!

170
This theorem implies that the evaluation of a closed contour depends on the residue evaluation. Since

every analytic function f can be written as

∞
f (k) (z0 )
f (z) = ∑ (z − z0 )k
k=0 k!
Then we have

f (z) f (n) (z0 ) ∞

f (k) (z0 )
= + ∑ (z − z0 )k
(z − z0 )n+1 (k!)(z − z0 ) k≠n k!
So the only term that contributes to the integral is the coefficient of (z − z0 )−1 which is by definition the

residue. Hence we conclude that if f has a singularity z = z0 we have

∮ f (z)dz = 2πi Res(f, z0 )

γ

We generalize the theorem by the following theorem Cauchy Residue Theorem or the residue theorem.

Theorem 12 Let f be an analytic function on a connected domain D except for some isolated points z1 , ⋯, zn then

for any simple closed contour γ ⊂ D that contains the singularities we have

n
∫ f (z)dz = 2πi ∑ Res(f (z), zk )
γ k=1

proof

Deform the contour γ into a set of circular contours γ1 , ⋯, γn around each singularity

∫ f (z)dz = ∫ f (z)dz + ∫ f (z)dz + ⋯ + ∫ f (z)dz

γ γ1 γ2 γn

Note for any 1 ≤ k ≤ n

∫ f (z)dz = 2πiRes(f, zk )
γk

Hence we deduce the result as sum of the residues. This is very powerful since it explains that the

evaluation of a function with some singularity around a contour containing them is no more than the

sum of residues. This relationship will be essential when we try to evaluate real integrals.

Example. Let P (z) be a polynomial with the leading coefficient as 1. Then find the following

dz
∫
γ P (z)
where γ contains the zeros of P .

171
solution

Factor P (z) = (z − z0 )(z − z1 ). Consider two cases

Case 1. the polynomial has a repeated root z0 then

dz dz 1
∫ =∫ = 2πiRes ( , z0 ) = 0
γ P (z) γ (z − z0 )2 (z − z0 )2

Since the Laurent expansion doesn’t contain a coefficient of (z − z0 )−1 .

Case 2. the polynomial has a two distinctive roots z0 , z1 then

1
dz dz
∫ =∫ = ∑ Res(1/P (z), zk )
γ P (z) γ (z − z0 )(z − z1 ) k=0

Note that

(z − z0 ) 1
Res(1/P (z), z0 ) = lim =
z→z0 (z − z0 )(z − z1 ) z0 − z1

(z − z1 ) 1
Res(1/P (z), z1 ) = lim =
z→z1 (z − z0 )(z − z1 ) z1 − z0
Hence we conclude that also

dz
∫ =0
γ P (z)
Can you generalize that ?

21.7 Bounds on integrals

Some integrals on contours are too difficult to evaluate so we better find a good bound or prove the

integral goes to 0 by proving

∣∮ f (z)dz∣ ≤ 
γ

So the integral can be made arbitrary close hence it goes to 0. We start by an important bound called the

Estimation lemma.

Theorem 13 Let f be a complex-valued, continuous function on the contour c if f is bounded by M for all z on c

then

∫ f (z)dz ≤ M L
c

where L is the length of the curve c.

172
proof

Note that by definition we have

∣∫ f (z) dz∣ ≤ ∫ ∣f (z)∣ ⋅ ∣dz∣

c c

Now assume that ∣f (z)∣ ≤ M that means the function is bounded on the curve

∫ ∣f (z)∣ ⋅ ∣dz∣ ≤ M ∫ ∣dz∣

c c

Now assume that the c = γ(t) is a parameterization of the curve then

b
∫ ∣dz∣ = ∫ ∣γ ′ (t)∣ dt
c a

Now by the definition of the length L of a curve we have

b
∫ ∣γ ′ (t)∣ dt = L
a

Example. Prove that

1
lim ∫ dz = 0
R→∞ CR z2 +1
Where CR is a semi-circle of radius R and centered at the origin.

proof

First note that for any point on the curve we have ∣z∣ = R hence

1 1
∣ ∣≤ 2
z2 +1 R −1
Where we used the Triangle inequality to show ∣z∣ − 1 ≥ R − 1 for large R. Also note that length of the

semi circle contour is πR. Hence we deduce

1 πR
∣∫ dz∣ ≤ 2
CR z2 + 1 R −1
By taking R → ∞ we conclude our result.

The estimation lemma could be used to prove a more generalized form which we call the Jordan’s

lemma.

Theorem 14 Let f be a complex-valued, continuous function on the contour CR which is a semi-circle on the

upper half plane centered at the origin. Let f be defined as the following

173
 f (z) = eiaz g(z)

For some a > 0 and assume that ∣f ∣ ≤ MR on CR then

π
∣∫ f (z)dz∣ ≤ MR
CR a

proof

By using the bound on g we have

∣∫ f (z)dz∣ ≤ MR ∫ ∣eiaz ∣ ⋅ ∣dz∣

CR CR

Now note that on the curve for t ∈ (0, π)

∣eiaz ∣ = ∣eiaR(cos(t)+i sin(t)) ∣ ≤ e−aR sin(t)

Hence we deduce by the parameterization Reiat

π
∣∫ f (z)dz∣ ≤ RMR ∫ e−aR sin(t) dt
CR 0

Since sin(t) is symmetric around π/2 we have

π π/2
∫ e−aR sin(t) dt = 2 ∫ e−aR sin(t) dt
0 0

Now use the fact sin(t) ≥ 2t

π
to deduce

π/2 π/2 π π
2∫ e−aR sin(t) dt ≤ 2 ∫ e−2atR/π dt = (1 − e−aR ) ≤
0 0 aR aR
We finally get the result

π π
∣∫ f (z)dz∣ ≤ × RMR = MR
CR aR a
Example. Prove that

e2iz
lim ∫ dz = 0
R→∞ CR z2 + 1
Where CR is a semi-circle of radius R and centered at the origin.

proof

By Jordan’s lemma

174
 e2iz π 1
∣∫ dz∣ ≤
CR z +1
2 2 R2 − 1
It is left as an exercise to show the same result holds for

P (z)
f (z) = eiaz
Q(z)

where P (z) and Q(z) are polynomials and D(P ) ≤ D(Q) + 1. Another remark is that we can generalize

the result to contours in the lower half-plain but wit the condition a < 0.

21.8 Contours around poles

Now we look at the case where we have a singular point along the path of integration. To define that we

have to look at the concept Cauchy Principal Value. Let use first look at the definition in the real case.

Consider the case where f is continuous on the interval [a, b] except for c in that interval then

b c− b
PV ∫ f (x)dx = lim+ ∫ f (x)dx + ∫ f (x)dx
a →0 a c+

Interestingly we can look at the point x = c as a singular point of the function f (x).

Example. Find the principal value of the function f (x) = 1

x
on the interval x ∈ (−1, 1).

solution

First note that the integral in the usual definition of Riemann doesn’t exist because the function is not

continuous in the interval. Now let us look at the principal value

1 1 − 1 1 1 1 1 1 1
PV ∫ dx = lim+ ∫ dx + ∫ dx == lim+ − ∫ dx + ∫ dx = 0
−1 x →0 −1 x  x →0  x  x

Hence the principal value exists while the integral is infinite on the interval. But if the Riemann integral

exists then the principal value integral exists and they are equal.

We can generalize the case to complex functions. Suppose that the function f around the along the

contour C has a pole at the point z = z0 then we can make a detour (semi-circle) around the pole and

take the radius goes to 0 as in the Figure 5.

Suppose that f (z) has a singularity at z = 0 and we need to find the integral along the contour C that is

a line connecting a and b then

− b
P V ∫ f (z)dz = lim ∫ f (x)dx + ∫ f (z)dx + ∫ f (x)dx
C →0 a C 

Let us now look at a simpler way to evaluate the contour around a simple pole instead of taking the

limit.

175
 y

Cε

a −ε ε b x

Figure 5: Avoiding singular points

Theorem 15 Let f (z) have a simple pole at z = z0 . Suppose that Cr is an arc starting at an angle θi and ending

at an angle θf around z0 then

f (z)
lim ∫ dz = i(θf − θi )f (z0 )
r→0 Cr z − z0

This theorem is powerful and gives away a simple way to evaluate integrals around arbitrary arcs.
eiz
Example . Find the integral of f (z) = z
around z = 0 traversed contour-clockwise.

solution

Consider Cr to be a semi-circle of radius r centered at the origin then

eiz
lim ∫ dz = (0 − π)iei0 = −πi
r→0 Cr z−0
Note the negative sign because we are starting at an angle π and up to 0.

176
22 Real integrals using contour integration

22.1 Trigonometric functions

In this section we consider solving integrals of functions of the form f (cos θ, sin θ) over the interval

(0, 2π). This can be done by realizing that for any point satisfying ∣z∣ = 1 then we have

z + z −1 = 2 cos θ

z − z −1 = 2i sin θ

Hence one can verify that

z + z −1 z − z −1 dz 2π
∮ f( , ) =∫ f (cos θ, sin θ)dθ
∣z∣=1 2 2i iz 0

This can be done by parameterizing ∣z∣ = 1 as eiθ on the interval θ ∈ (0, 2π). Hence one can use the residue

theorem to compute the complex integral.

22.1.1 Example

Evaluate the following integral

2π sin θ
∫ dθ
0 sin θ + 2
solution

Hence we can show

z−z −1
2π sin θ dz z2 − 1
∫ dθ = ∮ 2i
= −i ∮ dz
0 sin θ + 2 ∣z∣=1 2 + z−z
2i
−1
iz ∣z∣=1 z(z + 4iz − 1)
2

√ √
The function has three poles 0, −2i ± 3i with the only poles inside the contour 0, (−2 + 3)i

z2 − 1
Res(f, 0) = lim =1
z→0 z 2 + 4iz − 1

√
√ √ z2 − 1 −(−2 + 3)2 − 1 2
Res(f, (−2+ 3)i) = lim√ (z−(−2+ 3)i) √ √ = √ √ = −√
z→(−2+ 3)i z(z − (−2 + 3)i)(z − (−2 − 3)i) −2 3(−2 + 3) 3

Finally we get the value by summing the residues

z2 − 1 2
−i ∮ dz = 2π (1 − √ )
∣z∣=1 z(z 2 + 4iz − 1) 3

177
22.2 Integrating around an ellipse

22.2.1 Example

Prove for a, b > 0

2π dt 2π
∫ =
0 a2 cos2 t + b2 sin2 t ab
solution

Let us integrate the following function

1
f (z) =
z
Around the ellipse

b γ(t)

Figure 6: Integration around ellipse

By the residue theorem

∮ f (z) dz = 2πi Res(f, 0)

γ

The parameterization of the ellipse γ(t) = a cos(t) + ib sin(t)

2π −a sin t + ib cos t
∮ f (z) dz = ∫ dt
γ 0 a cos t + ib sin t
The intgrand simplifies to the following

−a sin t + ib cos t a cos t − ib sin t (b2 − a2 ) sin t cos t + iab

× =
a cos t + ib sin t a cos t − ib sin t a2 cos2 t + b2 sin2 t

Hence

2π (b2 − a2 ) sin t cos t + iab

∫ dt = 2πi Res(f, 0) = 2πi
0 a2 cos2 t + b2 sin2 t

178
By equating the imaginary parts

2π dt
iab ∫ = 2πi
0 a2 cos2 t + b2 sin2 t
Which implies the result

2π dt 2π
∫ =
0 a2 cos2 t + b2 sin2 t ab

22.3 Creating crazy integrals

Theorem. Let f be analytic function in the unit circle ∣z∣ ≤ 1 such that f ≠ 0 . Then

2π
∫ f (eit ) dt = 2π f (0)
0

proof

Since the function f is analytic in and on the contour we have by the Cauchy integral theorem

f (z)
∮ dz = 2πi Res(f /z, 0)
∣z∣=1 z
Use z = eit whenere 0 ≤ t ≤ 2π

2π f (eit ) it
i∫ e dt = 2πi Res(f /z, 0)
0 eit
Note that

f (z) f (z)
Res ( , 0) = lim z = f (0)
z z→0 z
Hence

2π
∫ f (eit ) dt = 2π f (0)
0

22.3.1 Example

Consider the function

f (z) = exp(exp(z))

It then follows from the theorem

179
 2π
∫ exp(exp(exp(it))) dt = 2π e
0

Note that

cos t+i sin t

exp(exp(exp(it))) = ee
cos t
(cos(sin t)+i sin(sin t))
= ee
cos t
= ee cos(sin t)
(cos(ecos t sin(sin t)) + i sin(ecos t sin(sin t)))

By taking the real part

2π cos t
∫ ee cos(sin t)
cos(ecos t sin(sin t)) dt = 2π e
0

22.3.2 Example

f (z) = log(2 + z)

Where we take the principal logarithm with the branch cut located at y = 0, x ≤ −2

2π
∫ log(2 + eit ) = 2π log(2)
0

Note that

sin t
log(2 + eit ) = log ∣(2 + cos t)2 + sin2 t∣ + i arctan ( )
1 + cos t
1 sin t
= log(5 + 4 cos t) + i arctan ( )
2 2 + cos t

It follows by taking the real parts

2π
∫ log(5 + 4 cos t) dt = 4π log(2)
0

22.3.3 Example

By combining the two functions

f (z) = exp(exp(z)) log(2 + z)

We deduce that

180
 2π sin t
∫ exp(ecos t cos(sin t))(cos(ecos t sin(sin t)) log(5+4 cos t)−2 arctan ( ) sin(ecos t sin(sin t))) dt = 4πe log(2)
0 2 + cos t

22.4 Trigonometric functions with rationals of polynomials

Here we are interested in integrals of the form

∞ cos(ax)
∫ dx
−∞ P (x)
where P (x) is continuous in the real line and of degree 2 or more. You may notice that it is enough to

look at the following function

eiaz
f (z) =
P (z)
and then take the real part. Generally speaking to solve that we take a semi-circle in the upper half plane

where the contour will enclose some or all of the zeros of P (z). Notice that we may use the Jordan’s

lemma to prove that the circular part goes to zero and what is left is the real integral along the real axis.

22.4.1 Example

Evaluate the integral

∞ cos(ax)
∫ dx
−∞ x2 + 1
proof

Let use consider the function with a > 0

eiaz
f (z) =
z2 + 1
and integrate around the contour in Figure 7

We can write the integral using the residue theorem as

R eiax
∫ dx + ∫ f (z)dz = 2πi Res(f, i)
−R x2 + 1 CR

By taking the limit R → ∞ note that by the Jordan’s lemma the integral along the circular part vanishes

(prove it). Hence we have

∞ eiax
∫ dx = 2πi Res(f, i)
−∞ x2 + 1

181
 CR
i
R

−R R

Figure 7: Circular contour

The evaluation of the residues

eiaz e−a
Res(f, i) = lim(z − i) =
z→i (z − i)(z + i) 2i
Finally we get

∞ eiax
∫ dx = πe−a
−∞ x2 + 1
and eventually

∞ cos(ax)
∫ dx = πe−a
−∞ x2 + 1
Advice. One might ask how to choose the complex function and the contour. There is no general formula

for that and you can obtain that by experience and trail and error. It is actually a good idea to try different

contours and see why they work!

22.5 Integration along contours with detours

Assume that we need to integrate a function and we have a pole at the path of integration then we have

to make a small detour around the pole with some angle.

22.5.1 Example

Prove the following

∞ sin(x) π
∫ dx =
0 x 2
proof

Consider the following function

182
 eiz
f (z) =
z
Note that in order to integrate a round a semi-circle in the upper half plane we have to make a detour

around the pole at z = 0

CR

cr

−R −r r R x

Figure 8: Integral with detour

Hence by the residue theorem

R eix −r eix
∫ dx + ∫ dx + ∫ f (z)dz + ∫ f (z)dz = 0
r x −R x CR Cr

Note first by taking R → ∞ we have by the Jordan’s lemma

lim ∫ f (z)dz = 0
R→∞ Cr

Also we know how to deal with detours around poles

eiz
lim ∫ dz = −πie0i = −πi
r→0 Cr z−0
Hence we finally get

∞ eix 0 eix
∫ dx + ∫ dx = πi
0 x −∞ x

By inverting the sign in the second integral

∞ eix − e−ix ∞ sin(x)

∫ dx = 2i ∫ dx = πi
0 x 0 x
By rearranging we have our result.

183
22.6 Integrals of functions with branch cuts

We have a serious problem when dealing with functions like f (z) = log(z) or z a where we have a branch

point at 0. In order to evaluate the integral we have to avoid both the branch point and the branch cut.

Generally speaking the integral around a branch point will most probably vanish but this is not taken

for granted. You have actually to prove it first.

22.6.1 Example

Prove the following

∞ log(x) π log(2)
∫ = √
0 x2 + 2 4 2
proof

Let us consider the usual integral of a semi-circle with a detour around the branch point at 0. But first

we have to choose an analytic branch of the logarithm.

Consider the following branch

log(z) = log ∣z∣ + iarg(z)

where arg(z) ∈ (−π/2, 3π/2) i.e we are taking the branch cut on the negative imaginary axis. By this

construction we have an analytic function along the path of integration. Now consider the function

log(z)
f (z) =
z2 + 2
√
The zero z = 2i lies inside the semi-circle

CR

√
2i
Cr

−R −r r R x
Branch

Figure 9: Integral of function with branch cut

By the residue theorem

184
 R log ∣x∣ −r log ∣x∣ + iπ √
∫ dx + ∫ dx + ∫ f (z)dz + ∫ f (z)dz = 2πiRes(f, 2i)
r x +2
2 −R x +2
2 CR Cr

Note the ∣ log(z)∣ = ∣ log ∣z∣ + iarg(z)∣ ≤ ∣ log R∣ + ∣arg(z)∣. Since the argument is bounded we can bound the

function using

log(z) ∣ log ∣z∣∣ + Const

∣ ∣≤
z2 + 2 ∣z∣2 − 2
Which vanishes on the larger curve since

log R + Const
lim R =0
R→∞ R2 − 2
Similarly we have

∣ log r∣ + Const
lim r =0
r→0 r2 − 2
Hence it follows by the Estimation lemma that both integrals vanish and

∞ log ∣x∣ 0 log ∣x∣ + iπ √

∫ dx + ∫ dx = 2πiRes(f, 2i)
0 x +2
2 −∞ x +2
2

By rearranging

∞ log ∣x∣ ∞ dx √
2∫ dx + πi ∫ = 2πiRes(f, 2i)
0 x2 + 2 0 x2 + 2
The residue evaluation

√ √
√ log( 2i) log 2 + 2 i
π
Res(f, 2i) = √ = √
2 2i 2 2i
Hence

√
∞ log ∣x∣ ∞ dx log 2 + π2 i
2∫ dx + πi ∫ =π √
0 x2 + 2 0 x2 + 2 2
By equating the real parts we get our result.

22.6.2 Example

Prove the following

∞ log (x2 + 1) arctan2 (x) π3

∫ dx = + π log2 (2)
0 x2 12

185
solution

Lemma

∞ log3 (1 + x2 )
∫ dx = π 3 + 3π log2 (4)
0 x2
proof

Start by the following

∞ Γ(1 − p)Γ(p − s)
∫ x−p (1 + x)s−1 dx =
0 Γ(1 − s)
Let x → x 2

∞ Γ(1 − p)Γ(p − s)
∫ x−2p+1 (1 + x2 )s−1 dx =
0 2Γ(1 − s)
Let p = 3/2
∞ 1 Γ(−1/2)Γ(3/2 − s)
∫ dx =
0 x2 (1 + x2 )1−s 2Γ(1 − s)
By taking the third derivative and s → 1

∞ log3 (1 + x2 ) Γ(−1/2) d3 Γ(3/2 − s)

∫ dx = [ 3 ] = π 3 + 3π log2 (4)
0 x2 2 ds 2Γ(1 − s) s=1
Now, Consider the function

log3 (1 − iz)
f (z) =
z2
Define the principle logarithm as follows

log z = log ∣z∣ + iArg(z)

Note that for x > 0 the argument can be evaluated as

√
log z = log x2 + y 2 + i arctan(y/x)

For the principle logarithm the branch cut is defined as I(1 − iz) = 0, R(1 − iz) ≤ 0 which then reduces

for z = x + iy to x = 0, 1 + y ≤ 0 . Hence the branch cut on the imaginary axis where x = 0, y < −1 . Also

note that f (z) is analytic on the punctured plane since around z = 0

log3 (1 − iz) ((iz) + (iz)2 /2 + (iz)3 /3 + O(z 4 ))3

= − = O(z 2 )
z2 z2

186
Hence at z = 0 we have a removable singularity. Define the following contour were we avoid the branch

cut

CR

−R R

−i

Branch

Figure 10: Integral with branch cut

√ 3 √
0 (log( 1 + x2 ) + i arctan(x)) R (log( 1 + x2 ) + i arctan(x))3
∫ f (z) dz + ∫ dx + ∫ dx = 0
CR −R x2 0 x2
Apparently

√ 3 √
R (log( 1 + x2 ) − i arctan(x)) R (log( 1 + x2 ) + i arctan(x))3
∫ f (z) dz + ∫ dx + ∫ dx = 0
CR 0 x2 0 x2
Note that

(x + y)3 = x3 + y 3 + 3x2 y + 3yx2

This simplifies to

R −3 arctan2 (x) log(1 + x2 ) + 1/4 log3 (1 + x2 )

∫ f (z) dz + ∫ dx = 0
CR 0 x2
Taking the limit

log3 (1 − iz) (log ∣1 + R∣ + 2π)3

∣∫ f (z) dz∣ = ∣∫ dz∣ ≤ πR ∼∞ 0
CR CR z2 R2
So this simplifies to

∞ arctan2 (x) log(1 + x2 ) 1 ∞ log3 (1 + x2 )

∫ dx = ∫ dx
0 x2 12 0 x2
Using the Lemma we reach our result.

187
22.6.3 Example

Prove a, b, c, d > 0

∞ log(a2 + b2 x2 ) π ad + bc
∫ dx = log
0 c2 + d2 x2 cd d
proof

Consider the function

log(a − ibz)
f (z) =
c2 + d 2 z 2
We need the logarithm with the branch cut y < − ab , x = 0 . Note that this corresponds to

√ π 3π
log(a + ibz) = log (a + y)2 + b2 x2 + iθ , θ ∈ [− , )
2 2
Consider the contour that avoids the branch-cut on the negative imaginary part

Cρ
c
d
ρ

−ρ ρ

− ab

Branch

Figure 11: Contour with branch cut on the imaginary axis

ρ log(a − ibx) 0 log(a − ibx) c

∫ f (z) dz + ∫ dx + ∫ dx = 2πi Res (f, i)
Cρ 0 c + dx
2 2 −ρ c + dx
2 2 d
In the second integral let x → −x

ρ log(a − ibx) ρ log(a + ibx) c

∫ f (z) dz + ∫ dx + ∫ dx = 2πi Res (f, i)
Cρ 0 c + dx
2 2 0 c2 + dx2 d
Note that in the x-axis we have

√ bx
log(a ± bix) = log( a2 + b2 x2 ) ± i arctan ( )
a

188
Using that we deduce

log(a − bix) + log(a + ibx) = log(a2 + b2 x2 )

This implies to

ρ log(a2 + b2 x2 ) c
∫ f (z) dz + ∫ dx = 2πi Res (f, i)
Cρ 0 c2 + dx2 d
For the circular part

log(a − ibz) log(a + bρ) + 2π

∣∫ dz∣ ≤ πρ ∼∞ 0
Cρ c2 + d2 z 2 ∣c2 − d2 ρ2 ∣
Let us look at the residue

c log(a − ibz) π ad + bc
2πiRes (f, i) = 2πi limc 2
= log
d z→ d i 2d z cd d
Hence this simplifies to

∞ log(a2 + b2 x2 ) π ad + bc
∫ dx = log
0 c + dx
2 2 cd d

22.6.4 Example

Prove the following

∞ log(x) cos(x) πEi(1) π

∫ dx = − −
0 (x2 + 1)2 4e 4e
proof

Consider the following function

log(z) iz
f (z) = e
(z 2 + 1)2
Now consider the the branch of the logarithm

log(z) = log ∣r∣ + iθ , θ ∈ (−π/2, 3π/2]

Consider the contour in Figure 12

Then by the residue theorem

R log(x) ix −r (log ∣x∣ + πi)
∫ f (z) dz + ∫ f (z) dz + ∫ e dx + ∫ eix dx = 2πi Res(f, i)
CR cr r (x2 + 1)2 −R (x2 + 1)2

189
 CR
i

−R −r r R
Cr

Figure 12: Contour with detour around 0

Which simplifies to

R log(x) ix R (log(x) + πi)

∫ f (z) dz + ∫ f (z) dz + ∫ e dx + ∫ e−ix dx = 2πiRes(f, i)
CR cr r (x + 1)
2 2 r (x2 + 1)2
Or

R cos(x) log(x) R e−ix

∫ f (z) dz + ∫ f (z) dz + 2 ∫ dx + πi ∫ dx = 2πiRes(f, i)
CR cr r (x2 + 1)2 r (x2 + 1)2
Note that for the semi-circle Reiθ

log(R) + iθ log R + 2π
MR = maxθ∈[0,π] ∣ ∣≤
(R e + 1)
2 2iθ 2 (R2 − 1)2
It follows by the Jordan’s lemma

lim ∫ f (z) dz = 0
R→∞ CR

For the smaller semi-circle

π log r + iπ reiθ iθ
∫ f (z) dz = ir ∫ e e dθ
cr 0 (r2 e2iθ + 1)2
Note that by a similar argument to the Jordan’s lemma we have

lim ∣∫ f (z) dz∣ ≤ lim πrMr → 0

r→0 cr r→0

It follows then

∞ cos(x) log(x) ∞ e−ix

2∫ dx + πi ∫ dx = 2πiResi f (z)
0 (x2 + 1)2 0 (x2 + 1)2
Evaluating the residue

190
 d log(z) iz π+i
Resi f (z) = lim ((z − i)2 2 e )=
z→i dz (z + 1) 2 4e
Note that

∞ eix π Ei(1)
∫ dx = −i
0 (x2 + 1)2 2e 2e
Since

∞ cos(x) −i
2∫ dx = 2πi ( )
0 (x2 + 1)2 2e
By integrating around a semi-circle in the upper half plane

∞ cos(x) π
∫ dx =
0 (x + 1)
2 2 2e
For the other integral, It is easy to see that

∞ sin(x) 1 −a ′
∫ dx = [e Ei (a) − ea Ei(−a)]
0 x +a
2 2 2a
By differentiation and letting a → 1 we have

∞ sin(x) Ei(1)
∫ dx =
0 (x + 1)
2 2 2e

We deduce that

∞ log(x) cos(x) π Ei(1) π i

2∫ dx + πi ( − i ) = 2πi ( + )
0 (x2 + 1)2 2e 2e 4e 4e
This simplifies to

∞ log(x) cos(x) πEi(1) π

∫ dx = − −
0 (x2 + 1)2 4e 4e

22.6.5 Example

Prove the following

∞ xα
∫ dx = −π csc(πα)
0 x+1
proof

Consider the following function

191
 zα eα log(z)
f (z) = =
1+z 1+z
We consider the branch of the logarithm

log(z) = ln ∣z∣ + iθ , θ ∈ [0, 2π)

This function is single-valued and on the integration path we have

⎧
⎪
⎪
⎪
⎪ α
θ→0
⎪x
eα log(z)
=x e
α iθα
=⎨
⎪
⎪
⎪
⎪
⎪xα e2παi θ → 2π
⎩
Integrating around contour (called key-hole contour) as in Figure 13

Cρ

C

−1 O x

Figure 13: Key hole contour

By the residue theorem we have

∮ f (z) dz = 2πi Res(f, −1)

Which becomes

ρ ρ
∫ f (z) dz + ∫ f (z) dz + ∫ f (z)dz − ∫ f (z)dz = 2πi Res(f, −1)
Cρ C  

By the chosen branch of the algorithm we have

ρ xα ρ xα
∫ f (z) dz + ∫ f (z) dz + ∫ dx − e2πiα ∫ dx = 2πi Res(f, −1)
Cρ C  x+1  x+1

Note for the circle ∣z∣ = ρ we have

192
 zα ∣z∣α ρα+1
∣∮ f (z) dz∣ ≤ 2πρ max ∣ ∣ ≤ 2πρ = 2π
∣z∣=ρ 1+z ∣∣z∣ − 1∣ ∣ρ − 1∣
Note the triangle inequality

∣ω + 1∣ ≥ ∣∣ω∣ − ∣1∣∣

Note that if −1 < α < 0 we get

ρα+1
∣ lim ∮ f (z) dz∣ ≤ 2π →0
ρ→∞ ∣z∣=ρ ∣ρ − 1∣
Similarly we have

α+1
∣lim ∮ f (z) dz∣ ≤ 2π →0
→0 ∣z∣= ∣ − 1∣
We deduce that as  → 0 and ρ → ∞

∞ xα
(1 − e2πiα ) ∫ dx = 2πi Res(f, −1) , −1 < α < 0
0 x+1
Notice that

Res(f, −1) = lim eα log(z) = eα(ln ∣−1∣+πi) = eαπi

z→−1

Hence we have

∞ xα eαπi 2πi
∫ dx = 2πi = = −π csc(πα)
0 x+1 1 − e2απi e−πiα − eπiα
Note we can deduce the Euler reflection formula

Γ(α)Γ(1 − α) = π csc(πα)

22.6.6 Example

Prove the following

π/2 πΓ(m + 1)
∫ cos(nt) cosm (t) dt =
0 2m+1 Γ ( n+m+2
2
) Γ ( 2−n+m
2
)
proof

Let us integrate the following function

193
 m
f (z) = z n−m−1 (1 + z 2 )

We choose the principle logarithm where

log(z) = log ∣z∣ + Arg(z)

Note that the function z n−m−1 = e(n−m−1) log(z) will have a branch cut on the negative x axis. Also we

have

m 2
)
(1 + z 2 ) = em log(1+z

The principle branch will be on the imaginary y axis where ∣y∣ ≥ 1 which implies that y ≥ 1 or y ≤ −1. We

integrate around the semi-circle of radius 1 as in Figure 14

i
γ 1

γ 2

γ 3
−i

Figure 14: Semi-circle indented at the branch points

By the residue theorem

i2 −i+i3
∫ f (z) dz + ∫ f (z) dz + ∫ f (z) dz + ∫ f (z) dz + ∫ f (z) dz + ∫ f (z) dz = 0
γ1 γ2 γ3 i−i1 −i2 C

I’ll prove the second integral goes to 0

Use the parameterization γ2 (t) = 2 eit , −π/2 ≤ t ≤ π/2

π/2
∣in−m
2 ∫ eint log(1 + (2 eit )2 ) dt∣ ≤ π log(2)n−m
2
−π/2

By taking the limit we deduce the integral goes to 0. Similarly we have

lim ∫ f (z) dz = lim ∫ f (z) dz = 0

1 →0 γ1  →0 2 γ3

194
Note on on the circular part of ∣z∣ = 1

π/2 m
∫ f (z) dz = i ∫ eit eint−imt−it (1 + e2it ) dt
C −π/2
π/2
=i∫ eint (e−it + eit )m dt
−π/2
π/2
= i2m ∫ eint cosm (t) dt
−π/2

The integrals reduce to

0 m −i m π/2
∫ xn−m−1 (1 + x2 ) dx + ∫ xn−m−1 (1 + x2 ) dx = −i2m ∫ eint cosm (t) dt
i 0 −π/2

By taking x = it and x = −it respectively

1 m 1 m π/2
−i ∫ (it)n−m−1 (1 − t2 ) dt − i ∫ (−it)n−m−1 (1 − t2 ) dt = −i2m ∫ eint cosm (t) dt
0 0 −π/2

We then can combine the integrals

1 m π/2
−i((i)n−m−1 + (−i)n−m−1 ) ∫ tn−m−1 (1 − t2 ) dt = −i2m ∫ eint cosm (t) dt
0 −π/2

Note that

nπ − mπ
−(in−m − (−i)n−m ) = − (ei(n−m)π/2 − e−i(n−m)π/2 ) = −2i sin ( )
2
We deduce that

π/2 nπ − mπ 1
∫ eint cosm (t) dt = 21−m sin ( ) ∫ tn−m−1 (1 − t2 )m dt
−π/2 2 0

Using the Euler formula we have

π/2 nπ − mπ 1
∫ cos(nt) cosm (t) dt = 2−m sin ( ) ∫ tn−m−1 (1 − t2 )m dt
0 2 0

Note the beta integral

1 Γ(m + 1)Γ ( n−m )

∫ tn−m−1 (1 − t2 )m dt = 2
0 2Γ ( n+m+2
2
)
By the reflection formula we have

n−m n−m π
Γ( ) Γ (1 − )=
2 2 sin ( nπ−mπ
2
)

195
We deduce then that

π/2 πΓ(m + 1)
∫ cos(nt) cosm (t) dt =
0 2m+1 Γ ( n+m+2
2
) Γ ( 2−n+m
2
)

22.7 Rectangular contours

In this section we consider the case of a rectangular contour. Usually considering such contours we have

to separate the contour into four integrals. Note that such contours are useful for evaluating integrals of

hyperbolic functions and infinite series in general.

22.7.1 Example
∞ cos(ax) πa
∫ dx = π sech ( )
−∞ cosh(x) 2
proof

Consider
eiaz
f (z) =
sinh(z)
If we integrate around a contour of height π and stretch it to infinity we get

πi/2 − T πi/2 + T

−πi/2 − T −πi/2 + T

Figure 15: Rectangular contour

By taking T → ∞

iπ/2+∞ iπ/2−∞ −iπ/2−∞ −iπ/2+∞

∫ f (x) dx + ∫ f (x) dx + ∫ f (x) dx + ∫ f (x) dx = 2πiRes(f, 0)
−iπ/2+∞ iπ/2+∞ iπ/2−∞ −iπ/2−∞

Consider

−iπ/2+∞ eiax
∫ dx
−iπ/2−∞ sinh(x)

196
Let x = −πi/2 + y

πa ∞ eiay
ie 2
∫ dy
−∞ cosh(y)
Similarly we have for

iπ/2−∞ eiax
∫ dx
iπ/2+∞ sinh(x)
By letting x = πi/2 + y

∞ eiay
ie−
πa
2
∫ dy
−∞ cosh(y)
Now consider the sum of the two remaining integrals

iπ/2+∞ eiax −iπ/2−∞ eiax

∫ dx + ∫ dx
−iπ/2+∞ sinh(x) iπ/2−∞ sinh(x)
Let x → −x in the second integral and summing

iπ/2+∞ eiax − e−iax iπ/2+∞ sin(ax)

∫ dx = 2i ∫ dx
−iπ/2+∞ sinh(x) −iπ/2+∞ sinh(x)

Let consider for simplicity

iπ/2+T sin(x)
∫ dx
−iπ/2+T sinh(x)
Let y = −i(x − T )

π/2 sin(i(y + T )) π/2 sinh(y + T ) π/2+T sinh(y)

−i ∫ dy = −i ∫ dy = −i ∫ dy
−iπ/2 sinh(i(y + T ) −π/2 sin(y + T ) −π/2+T sin(y)

Hence by taking T → ∞ we are integrating around an infinitely small area which goes to 0

∞ eiay
+ e−
πa πa
i (e 2 2 )∫ dy = 2πiRes(f, 0)
−∞ cosh(y)
Calculating the residue we have

eiaz eiaz
Res(f, 0) = lim z = lim =1
z→0 sinh(z) z→0 cosh(z)
Using that we get

∞ eiay 2π
∫ dy = πa
e 2 + e− 2
πa
−∞ cosh(y)

197
By taking the real part

∞ cos(ay) π
∫ dy = π sech ( a)
−∞ cosh(y) 2

22.7.2 Example
∞ 1 cosh (y + 4 ) π
2 π π
∫ dy = 3 (π cosh ( ) − 4 sinh ( ))
−∞ (5π 2 + 8πy + 16y 2 ) cosh3 (y) π 4 4
proof

Consider
sinh(z)
f (z) =
z sinh3 (z − π/4)
If we integrate around a contour of height π and stretch it to infinity we get

πi/2 − T πi/2 + T

π/4

−πi/2 − T −πi/2 + T

By taking T → ∞

iπ/2+∞ iπ/2−∞ −iπ/2−∞ −iπ/2+∞ π

∫ f (x) dx + ∫ f (x) dx + ∫ f (x) dx + ∫ f (x) dx = 2πiRes(f, )
−iπ/2+∞ iπ/2+∞ iπ/2−∞ −iπ/2−∞ 4
Consider

−iπ/2+∞ sinh(x)
∫ dx
−iπ/2−∞ x sinh3 (x − π/4)
Let x = −π/2i + π/4 + y

∞ 1 cosh(π/4 + y)
−∫ dy
−∞ −iπ/2 + π/4 + y cosh3 (y)
Similarly we have for

iπ/2−∞ sinh(x)
∫ dx
iπ/2+∞ x sinh3 (x − π/4)
By letting x = iπ/2 + π/4 + y

198
 ∞ 1 cosh(π/4 + y)
∫ dy
−∞ iπ/2 + π/4 + y cosh3 (y)
The other integrals go to 0 hence

∞ 1 cosh (y + 4 ) π
π
−16πi ∫ dy = 2πiRes(f, )
−∞ (5π 2 + 8πy + 16y 2 ) cosh3 (y) 4
Calculating the residue we have

∞ 1 cosh (y + 4 ) π
−(16(π cosh(π/4) − 4 sinh(π/4))
−16πi ∫ dy = 2πi
−∞ (5π + 8πy + 16y ) cosh3 (y)
2 2 π3
Which reduces to our result

∞ 1 cosh (y + 4 ) 2
π
π π
∫ dy = 3 (π cosh ( ) − 4 sinh ( ))
−∞ (5π + 8πy + 16y ) cosh (y)
2 2 3
π 4 4

22.7.3 Example

Prove the following

∞ sin(ax) 1 a 1
∫ dx = coth ( ) −
0 e2πx −1 4 2 2a
solution

By integrating the following function

eiaz
f (z) =
e2πz − 1
The function is analytic in and on the contour, indented at the poles of the function

R+i
i
i − i1 γ1

i2 γ 2

2 R

Figure 16: Indented rectangular contour

Hence by the residue theorem

199
 R R+i i+2 i−i1
∫ f (z) dz + ∫ f (x) dx + ∫ f (x) dx + ∫ f (x) dx + ∫ f (z) dz − ∫ f (x) dx = 0
γ2 2 R R+i γ1 i2

Let us first look at R → ∞

The vertical line

R+i eiax 1 eia(R+ixR)

∣∫ dx∣ = ∣iR ∫ dx∣
R e2πx − 1 0 e2π(R+ixR) − 1
1 ∣eia(R+ixR) ∣
≤ R∫ dx
0 ∣e2π(R+ixR) − 1∣
R 1
≤ e−axR dx
∫
∣e2πR − 1∣ 0
1
= (1 − e−aR ) ∼∞ 0
a∣e2πR − 1∣

The other vertical line

i(1−1 ) eiax (1−1 ) e−ax

∫ dx = i ∫ dx
i2 e2πx − 1 2 e2πix − 1
1 (1−1 ) e−ax
= ∫ dx
2 2 sin(πx)eiπx
1 (1−1 ) e−ax
= ∫ dx
2 2 sin(πx)eiπx
1 (1−1 ) cos(πx) −ax i (1−1 )
= ∫ e dx − ∫ e−ax dx
2 2 sin(πx) 2 2

Since the first integral diverges when 1 , 2 → 0

i eiax 1 cos(πx) 1 − e−a

PV ∫ dx = P V ∫ e−ax dx − i
0 e2πx −1 0 2 sin(πx) 2a
The remaining integrals can be evaluated using residues

πi i
lim ∫ f (z) dz = − Res(f, 0) = −
2 →0 γ 2 4 2

πi i
lim ∫ f (z) dz = − Res(f, i) = − e−a
1 →0 γ1 4 2
By combining the results together

∞ eiax ∞ eia(x+i) 1 cos(πx) 1 − e−a e−a + 1

PV ∫ dx − P V ∫ dx − P V ∫ e−ax dx + i =i
0 e2πx−1 0 e2π(x+i) −1 0 2 sin(πx) 2a 4
Which reduces to

200
 ∞ eiax 1 cos(πx) e−a + 1 1 − e−a
(1 − e−a )P V ∫ dx − P V ∫ e −ax
dx = i − i
0 e2πx − 1 0 2 sin(πx) 4 2a
By equating the imaginary parts

∞ sin(ax) 1 a 1
∫ dx = coth ( ) −
0 e2πx − 1 4 2 2a

22.8 Triangular contours

22.8.1 Example
∞ ∞ sin(x) √
cos(x) π
∫ √ dx = ∫ √ dx =
0 x 0 x 2
solution

Choose the function with the principal logarithm

f (z) = z −1/2 eiz = e−1/2 log(z)+iz

By integrating around the following contour

iR

ir Cr

r R

Figure 17: Triangular contour

By the residue theorem

R iR
∫ f (z) dz + ∫ f (x) dx + ∫ f (z) dz + ∫ f (x) dx = 0
Cr r γ ir

Taking the integral around the small quarter circle with r → 0

√ π/2 √ π/2
e−it/2 erie dt∣ ≤ ∣e−r sin(t) ∣ dt ∼ 0
it
∣∫ f (z) dz∣ ≤ ∣ r ∫ r∫
Cr 0 0

On γ(t) = (1 − t)R + iRt and 0 ≤ t ≤ 1

1 √ 1 e−Rt
∣∫ f (z) dz∣ = ∣R(i − 1) ∫ e−1/2 log(R(1−t)+iRt) ei(1−t)R−Rt dt∣ ≤ 2R ∫ √ dt
γ 0 0 4
(1 − t)2 + t2

201
 √ √
Note that on the interval 4
(1 − t)2 + t2 ≥ 4
2

√ 1 21/4
∣∫ f (z) dz∣ ≤ 21/4 R ∫ e−Rt dt = √ (1 − e−R ) ∼∞ 0
γ 0 R
Finally what is remaining when r → 0 and R → ∞

∞ eix ∞
∫ √ dx = i ∫ (ix)−1/2 e−x dx
0 x 0

Note that i−1/2 = e−iπ/4

∞ eix ∞ √
∫ √ dx = ie−iπ/4 ∫ x−1/2 e−x dx = ie−iπ/4 π
0 x 0

Using that we have

∞ ∞ sin(x) √
cos(x) π
∫ √ dx = ∫ √ dx =
0 x 0 x 2

22.9 Residue at infinity

Define the residue at infinite as the following

1 1
Res(f, ∞) = −Res ( 2
f ( ) , 0)
z z
This is useful especially considering integrals where we wrap the contour around the whole complex

plain by adding the point at infinity. Hence if we exclude a certain region of the complex plain and

integrate around the whole complex plain (think about it as a sphere) then we can apply the residue

theorem using the residue at infinity.

22.9.1 Example

Prove that

1√ √ π
∫ x 1 − x dx =
0 8
proof

Consider the function

√ 1 2
)
f (z) = z − z 2 = e 2 log(z−z

202
Consider the branch cut on the x-axis

x(1 − x) ≥ 0 Ô⇒ 0 ≤ x ≤ 1

Let w = z − z 2 then

log(w) = log ∣w∣ + iθ, θ ∈ [0, 2π)

Consider the contour where we avoid the two branch points

c0 c1

Figure 18: Dumbbell contour

By the residue theorem we have

1− 1−
e 2 log ∣x−x ∣ dx − ∫ e 2 log ∣x−x ∣+πi
1 2 1 2
∫ f (z) dz + ∫ f (z) dz + ∫ dx = 2πiRes(f, ∞)
c0 c1  

Consider the Laurent expansion of

√
√ √ 1 ∞
1/2 1 k
z − z2 =i z2 1− = iz ∑ ( ) (− )
z k=0 k z
Hence we deuce that

i
Res(f, ∞) = −
8
That implies

1− √ √ π
∫ f (z) dz + ∫ f (z) dz + 2 ∫ x 1 − x dx =
c0 c1  4
The contours around the branch points go to zero. Finally we get

1√ √ π
∫ x 1 − x dx =
0 8

203
22.10 Inverse of Laplace transform

The inverse Laplace transform is defined in the complex plain as

1 γ+iT
f (t) = L−1 {F }(t) = lim ∫ est F (s)ds
2πi T →∞ γ−iT

22.10.1 Example

1 c+i∞ Γ(a + b) a
∫ Γ(a + t)Γ(b − t)s−t dt = s
2πi c−i∞ (1 + s)a+b
proof

Consider the following function

f (z) = Γ(z + a)Γ(b − z)s−z

Suppose that a, b ∈ R and a < b. Note that the Gamma function has a pole of order 1 at each non-positive

integer where we have

(−1)n
Res(Γ, −n) =
n!
The function f has poles at the following points

−n − a, −(n − 1) − a, ⋯, −a, b, b + 1, ⋯, b + n

Notice that the function f is analytic on the region −a < Re(z) < b , hence consider the contour in Figure

19.

c + iT
CR

R
−a − 1 −a b b+1

c − iT

Figure 19: Bromwich contour

By the residue theorem

204
 c+iT n
∫ f (z) dz + ∫ f (z) dz = 2πi ∑ Res f (z)
CR c−iT k=0 z=−a−k

By taking the limit → ∞ we have

c+i∞ ∞
lim ∫ f (z) dz + ∫ f (t) dt = 2πi ∑ Res f (z)
R→∞ CR c−i∞ k=0 z=−a−k

Note that

(−1)k Γ(a + b + k) a+k

Res Γ(z + a)Γ(b − z)s−z = Γ(b + a + k) lim Γ(−k)sa+k = s
z=−a−k z→−a−k Γ(k + 1)
Note that

∞ ∞
Γ(a + b + k) (−s)k
sa ∑ (−s)k = sa Γ(a + b) ∑ (a + b)k = 2 F1 (1; 1, a + b, −s)
k=0 Γ(k + 1) k=0 k!
By definition of the Hypergeometric funciton

∞
(−s)k Γ(a + b) a
sa Γ(a + b) ∑ (a + b)k = sa Γ(a + b)2 F1 (a + b, 1; 1, −s) = s
k=0 k! (1 + s)a+b
Also notice that

lim ∫ f (z) dz = 0
R→∞ CR

So we deduce that

1 c+i∞ Γ(a + b) a
∫ Γ(a + t)Γ(b − t)s−t dt = s
2πi c−i∞ (1 + s)a+b

22.11 Infinite sums

If we have an infinite number of poles then by taking a contour that covers all of them we have an infinite

number of residues that create an infinite series.

22.11.1 Example

Prove that

∞
Hn
∑ 2
= 2ζ(3)
n=1 n

proof

Consider the function

205
 (ψ(−z) + γ)2
f (z) =
z2
Note that f has poles at non-negative integers

By integration around a large circle ∣z∣ = ρ

Cρ

ρ
0 1 2

Figure 20: Contour ∣z∣ = ρ

Note that

∞
∮ f (z) dz = 2πi(Res(f, 0) + ∑ Res(f, n))
n=1

The integration around the circle

(ψ(−z) + γ)2 (∣ψ(−z)∣ + ∣γ∣)2

∣∫ dz∣ ≤ 2π
∣z∣=ρ z2 ρ
As ∣z∣ → ∞ we have ψ(−z) ∼ log(−z) and the principle logarithm

(ψ(−z) + γ)2 (∣ log(−z)∣ + ∣γ∣)2 (∣ log(ρ)∣ + 2π + ∣γ∣)2

∣∫ dz∣ ≤ 2π ≤ 2π ∼∞ 0
∣z∣=ρ z2 ρ ρ
We deduce that

∞
2πi(Res(f, 0) + ∑ Res(f, n)) = 0
n=1

By expansion near z = 0

(ψ(−z) + γ)2 1 ζ(2) ζ(3)

2
≈ 2 −2 2 −2
z z z z
Which implies that

206
 Res(f, 0) = −2ζ(3)

For the residues at non-negative integers n, the expansions

∞
j j + 1 (z − n)
j
1
= ∑ (−1) ( )
z 2 j=0 1 n2+j

∞
1
ψ(−z) + γ = + Hn + ∑ ((−1)k Hn(k+1) − ζ(k + 1))(z − n)k
z−n k=1

This implies that

1 Hn
(ψ(−z) + γ)2 ≈ +2
(z − n) 2 (z − n)
This implies the residue

Hn 2
Res(f, n) = 2 −
n2 n3
We then deduce that

∞
Hn 1
2∑[ − ] − 2ζ(3) = 0
n=1 n2 n3
Finally we get

∞
Hn
∑ 2
= 2ζ(3)
n=1 n

207
References

[1] Leonard Lewin. Polylogarithms and associated functions. New York, 1981.

[2] Marko Petkovsek, Herbert Wilf, Dorom Zeilbegreger. A=B. Philadelphia, 1997.

[3] Pedro Freitas. Integrals of polylogarithmic functions, recurrence relations and associated Euler sums. 2004.

[4] Gradshteyn, Ryhik. Edited by Alan Jeffrey, Daniel Zillinger. Table of integrals, series and products. 2007.

[5] Habib Muzaffiar, Kenneth Williams. Evaluation of complete elleptic integrals of the first kind at singular

moduli. 2006.

[6] Stefan Boettner, Victor Moll. The integrals in Gradshteyn and Ryzhik, Part 16: Complete elliptic integrals.

2010.

[7] Factorial, Gamma and Beta function http://mhtlab.uwaterloo.ca/courses/me755/web_chap1.pdf

[8] Keith Conrad. Differentiation under the integral sign.

[9] http://www.mymathforum.com

[10] http://www.mathhelpboards.com

[11] http://www.mathstackexchange.com

[12] http://www.integralsandseries.prophpbb.com

[13] http://www.wikipedia.org

[14] http://mathworld.wolfram.com

[15] http://dlmf.nist.gov/

[16] Dennis G. Zill, Patrick D. Shanahan. A First Course in Complex Analysis with Applications. 2003.

[17] G. N Watson, M.A. Complex integration and Cauchy’s theorem. 1914.

[18] Ravi P. Agarwal, Kanishka Perera, Sandra Pinelas. An Introduction to Complex Analysis. 2011.

208