WTW123 – Proofs, Formulas and Graphs

1. Formulas
Section 1.1 – Graphs and Interval Division
𝑥−𝑦
Error
(If 𝑦 is an approximation for 𝑥)
|𝑥 − 𝑦|
Absolute Error
(If 𝑦 is an approximation for 𝑥)
|𝑥 − 𝑦|
Relative Error |𝑥|
(If 𝑦 is an approximation for 𝑥)

Section 1.2 – Polynomial Equations

|𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |
𝑚=
|𝑎𝑛 |
Theorem 4 (𝒎) for 𝑓(𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + ⋯ + 𝑎0
• If 𝑚 ≥ 1, then all the roots have absolute values less or equal to 𝑚.
• If 𝑚 < 1, all roots will have absolute value less than 1.

Section 1.3 – Fixed Point Iteration

Iterations 𝑥𝑘+1 = 𝑔(𝑥𝑘 )
𝒆𝒏 𝑒𝑛 = 𝑐 − 𝑥𝑛
𝒗𝒏 𝑣𝑛 = |𝑥𝑛+1 − 𝑥𝑛 |
|𝑐 − 𝑥𝑛 | ≤ 𝑚|𝑐 − 𝑥𝑛−1 | = |𝑒𝑛 | ≤ 𝑚|𝑒𝑛−1 |
1 = 𝑣𝑛
Theorem 6 (Error |𝑐 − 𝑥𝑛 | ≤ |𝑥𝑛+1 − 𝑥𝑛 | |𝑒𝑛 | ≤
1−𝑚 1−𝑚
Estimates) 𝑚 𝑛
𝑚𝑛
=
|𝑐 − 𝑥𝑛 | ≤ |𝑥 − 𝑥0 | |𝑒𝑛 | ≤ 𝑣
1−𝑚 1 1−𝑚 0

Section 1.4 – Newton-Raphson Method

Newton-Raphson 𝑓(𝑥𝑘 )
𝑥𝑘+1 = 𝑥𝑘 − , 𝑘 = 0,1,2, …
Iterations 𝑓 ′ (𝑥𝑘 )

Section 1.5 – Convergence

𝑒𝑛+1 𝑛
→𝑘
𝑒𝑛 ∞
Fixed Point Iteration (with 𝑘 constant)
(Convergence) • If 𝑘 ≠ 0, the convergence is called linear
• If 𝑘 = 0 the convergence is called superlinear.
𝑒𝑛+1 𝑛
→𝑘
𝑒𝑛 ∞
Newton-Raphson (with 𝑘 constant)
(Linear Convergence)
• If 𝑘 ≠ 0, the convergence is called linear
• If 𝑘 = 0 the convergence is called superlinear.
Newton-Raphson 𝑒𝑛+1 𝑛
→𝑘
(Quadratic 𝑒𝑛2 ∞
Convergence) (with 𝑘 = 1 a constant)
Section 2.1 – Numerical Integration and Rectangle Rules
𝑛

𝑳𝑹(𝒇, 𝒉) 𝐿𝑅(𝑓, ℎ) = ∑ 𝑓(𝑥𝑘−1 )ℎ = ℎ[𝑓(𝑥0 ) + 𝑓(𝑥1 ) + ⋯ + 𝑓(𝑥𝑛−1 )]

𝑘=1
𝑛

𝑹𝑹(𝒇, 𝒉) 𝑅𝑅(𝑓, ℎ) = ∑ 𝑓(𝑥𝑘 )ℎ = ℎ[𝑓(𝑥1 ) + 𝑓(𝑥2 ) + ⋯ + 𝑓(𝑥𝑛 )]

𝑘=1
𝑛

𝑀𝑅(𝑓, ℎ) = ∑ 𝑓(𝑥𝑘∗ )ℎ = ℎ[𝑓(𝑥1∗ ) + 𝑓(𝑥2∗ ) + ⋯ + 𝑓(𝑥𝑛∗ )]

𝑴𝑹(𝒇, 𝒉) 𝑘=1
𝑥𝑘 +𝑥𝑘−1
(where 𝑥𝑘∗ = is the midpoint of the interval [𝑥𝑘−1 , 𝑥𝑘 ])
2

Section 2.2 – Trapezoidal and Simpson Rules

𝑛
1
𝑇(𝑓, ℎ) = ∑ (𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 ))ℎ
2
𝑘=1
ℎ
𝑻(𝒇, 𝒉) = (𝑓(𝑥0 ) + 2𝑓(𝑥1 ) + 2𝑓(𝑥2 ) + ⋯ + 2𝑓(𝑥𝑛−1 ) + 𝑓(𝑥𝑛 ))
2
𝑛−1
ℎ
= (𝑓(𝑥0 ) + 𝑓(𝑥𝑛 )) + ℎ ∑ 𝑓(𝑥𝑘 )
2
𝑘=1
𝑛
ℎ
𝑆(𝑓, ℎ) = ∑ (𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 ))
3
𝑘=1
ℎ 2ℎ
𝑺(𝒇, 𝒉) 𝑆(𝑓, ℎ) = (𝑓(𝑥0 ) + 𝑓(𝑥2𝑛 )) + (𝑓(𝑥2 ) + 𝑓(𝑥4 ) + ⋯ 𝑓(𝑥2𝑛−2 ))
3 3
4ℎ
+ (𝑓(𝑥1 ) + 𝑓(𝑥3 ) + ⋯ 𝑓(𝑥2𝑛−1 ))
3

Section 2.3 – Piecewise Continuous Functions and Accuracy

𝑥𝑘
Local Error 𝑒𝑘 = ∫ 𝑓(𝑥) 𝑑𝑥 − ℎ𝑓(𝑥𝑘∗ )
𝑥𝑘−1
𝑏 𝑛

𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − ∑ ℎ𝑓(𝑥𝑘∗ )

𝑎 𝑘=1
𝑛 𝑥𝑘 𝑛

= ∑∫ 𝑓(𝑥) 𝑑𝑥 − ∑ ℎ𝑓(𝑥𝑘∗ )
Global Error 𝑘=1 𝑥𝑘−1 𝑘=1
𝑛 𝑥𝑘
= ∑ (∫ 𝑓(𝑥) 𝑑𝑥 − ℎ𝑓(𝑥𝑘∗ ))
𝑘=1 𝑥𝑘−1

= 𝑒1 + 𝑒2 + 𝑒3 + ⋯ + 𝑒𝑛
𝑏
1
Global Error Estimate |∫ 𝑓(𝑥) 𝑑𝑥 − 𝐿𝑅(𝑓, ℎ)| ≤ 𝑀1 (𝑏 − 𝑎)ℎ
𝑎 2
for LR
with 𝑀1 = max|𝑓 ′ |
[𝑎,𝑏]
𝑏
1
Global Error Estimate |∫ 𝑓(𝑥) 𝑑𝑥 − 𝑅𝑅(𝑓, ℎ)| ≤ 𝑀1 (𝑏 − 𝑎)ℎ
𝑎 2
for RR
with 𝑀1 = max|𝑓 ′ |
[𝑎,𝑏]
𝑏
1
Global Error Estimate |∫ 𝑓(𝑥) 𝑑𝑥 − 𝑀𝑅(𝑓, ℎ)| ≤ 𝑀 (𝑏 − 𝑎)ℎ2
𝑎 24 2
for MR
with 𝑀2 = max|𝑓 ′ ′|
[𝑎,𝑏]
 𝑏
1
Global Error Estimate |∫ 𝑓(𝑥) 𝑑𝑥 − 𝑇(𝑓, ℎ)| ≤ 𝑀 (𝑏 − 𝑎)ℎ2
𝑎 12 2
for T
with 𝑀2 = max|𝑓 ′ ′|
[𝑎,𝑏]
𝑏
1
Global Error Estimate |∫ 𝑓(𝑥) 𝑑𝑥 − 𝑆(𝑓, ℎ)| ≤ 𝑀 (𝑏 − 𝑎)ℎ4
𝑎 180 4
for S
with 𝑀4 = max|𝑓 (4) |
[𝑎,𝑏]

Section 2.4 – Local Errors for the Rectangle Rules

𝑥𝑘
1
Local Error for LR |𝑒𝑘 | = |∫ 𝑓 − 𝑓(𝑥𝑘−1 )ℎ𝑘 | ≤ 𝑀ℎ2
𝑥𝑘−1 2
𝑥𝑘
1
Local Error for RR |𝑒𝑘 | = |∫ 𝑓 − 𝑓(𝑥𝑘 )ℎ𝑘 | ≤ 𝑀ℎ2
𝑥𝑘−1 2
𝑥𝑘
1
|𝑒𝑘 | = |∫ 𝑓 − 𝑓(𝑥𝑘∗ )ℎ𝑘 | ≤ 𝑀ℎ3
Local Error for MR 𝑥𝑘−1 24
𝑥𝑘 +𝑥𝑘−1
(where 𝑥𝑘∗ = is the midpoint of the interval [𝑥𝑘−1 , 𝑥𝑘 ])
2

Section 3.1 – First Order Differential Equations: Euler’s Method

𝒕𝒌+𝟏 𝑡𝑘 + ℎ
𝒚𝒌+𝟏
𝑦𝑘 + ℎ(𝑓(𝑦𝑘 ))
(Euler’s Method)

Section 3.2 – First Order Differential Equations: Improved Euler’s Method

𝒕𝒌+𝟏 𝑡𝑘 + ℎ
𝒑𝒌+𝟏 𝑦𝑘 + ℎ(𝑓(𝑦𝑘 ))
𝒚𝒌+𝟏 ℎ
(Improved Euler’s 𝑦𝑘 + (𝑓(𝑦𝑘 ) + 𝑓(𝑝𝑘+1 ))
Method) 2

Section 3.3 – Accuracy

Global Error 𝐸𝑘 = 𝑦(𝑡𝑘 ) − 𝑦𝑘
Local Error 𝑒𝑘 = 𝑦(𝑡𝑘 ) − 𝑦𝑘∗ where 𝑦𝑘∗ = 𝑦(𝑡𝑘−1 ) + ℎ𝑓(𝑦(𝑡𝑘−1 )
Global Error Estimate |𝐸𝑘 | ≤ 𝐶𝑒 𝐷𝑇 ℎ
1
Local Error Estimate |𝑒𝑘 | ≤ 𝑀2 ℎ2
2

Please Note…

𝑬𝒌 Denotes the global error

𝒆𝒌 Denotes the local error
𝑪
Constants that exist if the solution of an initial value problem is approximated on [0. 𝑇]
𝑫
𝑻 The solution of an initial value problem is approximated on an interval [0, 𝑇]
𝑴𝟐 𝑀2 = max|𝑦 ′′ |
𝒉 Denotes the step length
Section 3.4 – Systems of Differential Equations
Euler algorithm
𝑢𝑘+1 = (𝐼 + ℎ𝐴)𝑢𝑘
(for homogeneous linear systems)
Improved Euler algorithm ℎ2 2
𝑢𝑘+1 = (𝐼 + ℎ𝐴 + 𝐴 ) 𝑢𝑘
(for homogeneous linear systems) 2

Section 3.5 – Higher Order Differential Equations

No formulas

Section 4.1 – Gauss Elimination

Gauss Elimination Notation
𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧(𝒌) ← 𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧(𝒌) − 𝒎 × 𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧(𝒊) Equation(𝑘) is substituted by Equation(𝑘) minus
𝑚 times Equation(𝑖)
𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧(𝒌) ↔ 𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧(𝒊) Equation(𝑘) and Equation(𝑖) are interchanged
𝐫𝐨𝐰(𝒌) ← 𝐫𝐨𝐰(𝒌) − 𝒎 × 𝐫𝐨𝐰(𝒊) row(𝑘) is substituted by row(𝑘) minus 𝑚 times
row(𝑖)
𝐫𝐨𝐰(𝒌) ↔ 𝐫𝐨𝐰(𝒊) row(𝑘) and row(𝑖) are interchanged

Determinant
By row operations det 𝐴 = (−1)𝑝 det(𝑈)
𝑛

By Row 𝒊 det 𝐴 = ∑(−1)𝑖+𝑗 𝑎𝑖𝑗 𝑀𝑖𝑗

𝑗=1
𝑛

By Column 𝒋 det 𝐴 = ∑(−1)𝑖+𝑗 𝑎𝑖𝑗 𝑀𝑖𝑗

𝑖=1
Section 4.2 – Iterative Methods for Linear Systems
Strictly diagonally |𝑎𝑖𝑗 | > ∑|𝑎𝑖𝑖 | for 𝑖 = 1,2,3,4 … , 𝑛
dominant matrix 𝑗≠𝑖
𝑛
(𝑟+1) (𝑟)
𝑥1 = 𝑑1 + ∑ 𝑐1𝑘 𝑥𝑘
𝑘=1
𝑛
(𝑟+1) (𝑟)
𝑥2 = 𝑑2 + ∑ 𝑐2𝑘 𝑥𝑘
Jacobi-iteration
𝑘=1
⋮
𝑛
(𝑟+1) (𝑟)
𝑥𝑛 = 𝑑𝑛 + ∑ 𝑐𝑛𝑘 𝑥𝑘
𝑘=1
𝑛
(𝑟+1) (𝑟)
𝑥1 = 𝑑1 + ∑ 𝑐1𝑘 𝑥𝑘
𝑘=2
𝑛
(𝑟+1) (𝑟+1) (𝑟)
𝑥2 = 𝑑2 + 𝑐21 𝑥1 + ∑ 𝑐2𝑘 𝑥𝑘
𝑘=3
⋮
Gauss-Seidel 𝑖−1 𝑛
(𝑟+1) (𝑟+1) (𝑟)
𝑥𝑖 = 𝑑𝑖 + ∑ 𝑐𝑖𝑘 𝑥𝑘 + ∑ 𝑐𝑖𝑘 𝑥𝑘
𝑘=1 𝑘=𝑖+1
⋮
𝑛−1
(𝑟+1) (𝑟+1)
𝑥𝑛 = 𝑑𝑛 + ∑ 𝑐𝑛𝑘 𝑥𝑘
𝑘=1
𝑛
2-norm ‖𝑥̅ ‖2 = √(∑|𝑥𝑖 |2 )
𝑖=1
𝑛

1-norm ‖𝑥̅ ‖1 = ∑|𝑥𝑖 |

𝑖=1
Maximum norm ‖𝑥̅ ‖∞ = max{|𝑥𝑖 | ∶ 𝑖 = 1,2, … , 𝑛}
2. Proofs
Section 1.1 – Graphs and Interval Division
No proofs

Section 1.2 – Polynomials Equations

Theorem 2
Each polynomial equation of odd degree has at least one real root.

Proof
Considerations:
• Consider 𝑓(𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + ⋯ + 𝑎0 = 0
• Consider the case when 𝑎𝑛 = 1 (both sides of the equation can be divided by 𝑎𝑛 )
Note the following:
• If 𝑛 is odd, 𝑥 𝑛 has the same sign as 𝑥
Choose 𝑥 such that |𝑥| > 1 and |𝑥| > |𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |, then
|𝑎0 + 𝑎1 𝑥 + ⋯ + 𝑎𝑛−1 𝑥 𝑛−1 |
≤ |𝑎0 | + |𝑎1 ||𝑥| + ⋯ + |𝑎𝑛−1 ||𝑥|𝑛−1 (triangle inequality and absolute value properties)
< (|𝑎0 | + |𝑎1 | + |𝑎𝑛−1 |)|𝑥|𝑛−1 (multiply with applicable 𝑥 to factorise out 𝑥 𝑛−1 )
< |𝑥|𝑛 (since |𝑥| > |𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |)
Case 1: If 𝑥 = 𝑏 > 0,
then 𝑏 𝑛 > 0 and
−𝑏 𝑛 < (𝑎0 + 𝑎1 𝑏 + ⋯ + 𝑎𝑛−1 𝑏 𝑛−1 )
∴ 𝑓(𝑏) = 𝑎0 + 𝑎1 𝑏 + ⋯ + 𝑎𝑛−1 𝑏𝑛−1 + 𝑏 𝑛 > 0
Case 2: If 𝑥 = 𝑐 < 0,
then 𝑐 𝑛 < 0 and |𝑐 𝑛 | = −𝑐 𝑛 and
𝑎0 + 𝑎1 𝑐 + ⋯ + 𝑎𝑛−1 𝑐 𝑛−1 < −𝑐 𝑛
∴ 𝑓(𝑐) = 𝑎0 + 𝑎1 𝑐 + ⋯ + 𝑎𝑛−1 𝑐 𝑛−1 + 𝑐 𝑛 < 0
The result follows from the intermediate value theorem.
Theorem 4
|𝑎0 |+|𝑎1 |+⋯+|𝑎𝑛−1 |
Consider the equation 𝑓(𝑥) = 0 with 𝑓(𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + ⋯ + 𝑎0 and let 𝑚 = .
|𝑎𝑛 |

• If 𝑚 ≥ 1, then all the roots have absolute values less or equal to 𝑚.

• If 𝑚 < 1, all roots will have absolute value less than 1.

Proof
Considerations:
• 𝑓(𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + ⋯ + 𝑎0 = 0
|𝑎 |+|𝑎 |+⋯+|𝑎 |
• 𝑚 = 0 1|𝑎 | 𝑛−1 .
𝑛
If 𝑥 is a nonzero root of the equation then,
|𝑎𝑛 ||𝑥|𝑛
= |𝑎𝑛 𝑥 𝑛 |
= |−(𝑎0 + 𝑎1 𝑥 + ⋯ + 𝑎𝑛−1 𝑥 𝑛−1 )|
≤ |𝑎0 | + |𝑎1 ||𝑥| + ⋯ + |𝑎𝑛−1 ||𝑥|𝑛−1 (triangle inequality and properties of absolute value)
If |𝑥| ≥ 1 then
|𝑎𝑛 | ≤ (|𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |)|𝑥|𝑛−1
Dividing |𝑎𝑛 ||𝑥|𝑛−1 , it follows that |𝑥| ≤ 𝑚
|𝑎𝑛 | (|𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |)|𝑥|𝑛−1
≤
|𝑎𝑛 ||𝑥|𝑛−1 |𝑎𝑛 ||𝑥|𝑛−1
|𝑎0 | + |𝑎1 | + ⋯ + |𝑎𝑛−1 |
|𝑥| ≤ =𝑚
|𝑎𝑛 |
As a consequence, we have 𝑚 ≥ 1.
Hence if 𝑚 ≥ 1, and 𝑥 is a root of the equation, then |𝑥| ≤ 𝑚
(we can also conclude that a root 𝑥 with |𝑥| > 1 D.N.E. if 𝑚 < 1).

Section 1.3 – Fixed Point Iteration

Theorem 1
If |𝑔′ (𝑥)| < 1 for every 𝑥 ∈ (𝑎, 𝑏), then the equation 𝑥 = 𝑔(𝑥) has at most one solution in [𝑎, 𝑏].

Proof
Let ℎ(𝑥) = 𝑥 − 𝑔(𝑥)
then ℎ(𝑥) = 0 if and only if 𝑥 = 𝑔(𝑥).
Now, ℎ′ (𝑥) = 1 − 𝑔′ (𝑥) > 0 (since −1 > 𝑔′ (𝑥) > 1) for every 𝑥 ∈ (𝑎, 𝑏).
Thus, ℎ(𝑥) is strictly increasing on (𝑎, 𝑏)
Therefore
• This implies that ℎ has at most one zero.
• 𝑥 = 𝑔(𝑥) can have at most 1 solution

Section 1.4 – Newton-Raphson Method

No Proofs

Section 1.5 – Convergence

No Proofs
Section 2.1 – Numerical Integration and Rectangle Rules
𝑳𝑹(𝒇 + 𝒈, 𝒉) = 𝑳𝑹(𝒇, 𝒉) + 𝑳𝑹(𝒈, 𝒉)
𝑛

𝐿𝑅(𝑓 + 𝑔, ℎ) = ∑(𝑓 + 𝑔)(𝑥𝑘−1 )ℎ

𝑘=1
𝑛

= ∑[𝑓(𝑥𝑘−1 ) + 𝑔(𝑥𝑘−1 )] ℎ
𝑘=1
𝑛 𝑛

= ∑ 𝑓(𝑥𝑘−1 )ℎ + ∑ 𝑔(𝑥𝑘−1 )ℎ
𝑘=1 𝑘=1
= 𝐿𝑅(𝑓, ℎ) + 𝐿𝑅(𝑔, ℎ)

𝑹𝑹(𝒇 + 𝒈, 𝒉) = 𝑹𝑹(𝒇, 𝒉) + 𝑹𝑹(𝒈, 𝒉)

𝑛

𝑅𝑅(𝑓 + 𝑔, ℎ) = ∑(𝑓 + 𝑔)(𝑥𝑘 )ℎ

𝑘=1
𝑛

= ∑[𝑓(𝑥𝑘 ) + 𝑔(𝑥𝑘 )] ℎ
𝑘=1
𝑛 𝑛

= ∑ 𝑓(𝑥𝑘 )ℎ + ∑ 𝑔(𝑥𝑘 )ℎ
𝑘=1 𝑘=1
= 𝑅𝑅(𝑓, ℎ) + 𝑅𝑅(𝑔, ℎ)

𝑴𝑹(𝒇 + 𝒈, 𝒉) = 𝑴𝑹(𝒇, 𝒉) + 𝑴𝑹(𝒈, 𝒉)

𝑛

𝑀𝑅(𝑓 + 𝑔, ℎ) = ∑(𝑓 + 𝑔)(𝑥𝑘∗ )ℎ

𝑘=1
𝑛

= ∑[𝑓(𝑥𝑘∗ ) + 𝑔(𝑥𝑘∗ )] ℎ
𝑘=1
𝑛 𝑛

= ∑ 𝑓(𝑥𝑘∗ )ℎ + ∑ 𝑔(𝑥𝑘∗ )ℎ
𝑘=1 𝑘=1
= 𝑀𝑅(𝑓, ℎ) + 𝑀𝑅(𝑔, ℎ)
1
where 𝑥𝑘∗ = (𝑥𝑘 + 𝑥𝑘−1 )
2

𝑳𝑹(𝒄𝒇, 𝒉) = 𝒄𝑳𝑹(𝒇, 𝒉)
𝑛

𝐿𝑅(𝑐𝑓, ℎ) = ∑ 𝑐𝑓(𝑥𝑘−1 ) ℎ
𝑘=1
𝑛

= 𝑐 ∑ 𝑓(𝑥𝑘−1 )ℎ
𝑘=1
= 𝑐𝐿𝑅(𝑓, ℎ)

𝑹𝑹(𝒄𝒇, 𝒉) = 𝒄𝑹𝑹(𝒇, 𝒉)
𝑛

𝑅𝑅(𝑐𝑓, ℎ) = ∑ 𝑐𝑓(𝑥𝑘 ) ℎ
𝑘=1
𝑛

= 𝑐 ∑ 𝑓(𝑥𝑘 )ℎ
𝑘=1
= 𝑐𝑅𝑅(𝑓, ℎ)
 𝑴𝑹(𝒄𝒇, 𝒉) = 𝒄𝑴𝑹(𝒇, 𝒉)
𝑛

𝑀𝑅(𝑐𝑓, ℎ) = ∑ 𝑐𝑓(𝑥𝑘∗ ) ℎ
𝑘=1
𝑛

= 𝑐 ∑ 𝑓(𝑥𝑘∗ )ℎ
𝑘=1
= 𝑐𝑀𝑅(𝑓, ℎ)

The degree of precision for the left-hand rectangle rule is 0

Let 𝑓(𝑥) = 𝑘
Result 1 𝑏
∫ 𝑓(𝑥) 𝑑𝑥
𝑎
= 𝑘𝑥|𝑏𝑎
= 𝑘(𝑏 − 𝑎)
Result 2 𝐿𝑅(𝑓, ℎ)
𝑛

= ∑ 𝑓(𝑥𝑘−1 )ℎ
𝑘=1
= ℎ[𝑓(𝑥0 ) + 𝑓(𝑥1 ) + ⋯ + 𝑓(𝑥𝑛−1 )]
= ℎ(𝑘 + 𝑘 … + 𝑘)
= ℎ(𝑛𝑘)
𝑏−𝑎
=( ) 𝑛𝑘
𝑛
= 𝑘(𝑏 − 𝑎)
We calculate 𝐸(𝑓, ℎ) by combining result 1 and 2
𝑏
𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − 𝐿𝑅(𝑓, ℎ)
𝑎
= 𝑘(𝑏 − 𝑎) − 𝑘(𝑏 − 𝑎)
=0
Therefore, the degree of precision is at least 0

The degree of precision for the right-hand rectangle rule is 0

Let 𝑓(𝑥) = 𝑘
Result 1 𝑏
∫ 𝑓(𝑥) 𝑑𝑥
𝑎
= 𝑘𝑥|𝑏𝑎
= 𝑘(𝑏 − 𝑎)
Result 2 𝑅𝑅(𝑓, ℎ)
𝑛

= ∑ 𝑓(𝑥𝑘 )ℎ
𝑘=1
= ℎ[𝑓(𝑥1 ) + 𝑓(𝑥2 ) + ⋯ + 𝑓(𝑥𝑛 )]
= ℎ(𝑘 + 𝑘 … + 𝑘)
= ℎ(𝑛𝑘)
𝑏−𝑎
=( ) 𝑛𝑘
𝑛
= 𝑘(𝑏 − 𝑎)
We calculate 𝐸(𝑓, ℎ) by combining result 1 and 2
𝑏
𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − 𝑅𝑅(𝑓, ℎ)
𝑎
= 𝑘(𝑏 − 𝑎) − 𝑘(𝑏 − 𝑎)
=0
Therefore, the degree of precision is at least 0
The degree of precision for the midpoint rectangle rule is 1
1
Let 𝑓(𝑥) = 𝑥 and let 𝑥𝑘∗ = 2 (𝑥𝑘−1 + 𝑥𝑘 )

Result 1 𝑏
∫ 𝑓(𝑥) 𝑑𝑥
𝑎
𝑛 𝑥𝑘
= ∑∫ 𝑓(𝑥) 𝑑𝑥
𝑘=1 𝑥𝑘−1
𝑛
1 2 𝑥𝑘
=∑ 𝑥 |
2 𝑥𝑘−1
𝑘=1
𝑛
1
= ∑ (𝑥𝑘2 − 𝑥𝑘−1
2
)
2
𝑘=1
𝑛
1
= ∑ (𝑥𝑘 − 𝑥𝑘−1 )(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
𝑛
1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
Result 2 𝑀𝑅(𝑓, ℎ)
𝑛

= ∑ 𝑓(𝑥𝑘∗ )ℎ
𝑘=1
𝑛
1
= ∑ ℎ ( (𝑥𝑘 + 𝑥𝑘−1 ))
2
𝑘=1
𝑛
1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
We calculate 𝐸(𝑓, ℎ) by combining result 1 and 2
𝑏
𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − 𝑀𝑅(𝑓, ℎ)
𝑎
𝑛 𝑛
1 1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 ) − ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2 2
𝑘=1 𝑘=1
=0
Therefore, the degree of precision is at least 1

Section 2.2 – Trapezoidal and Simpson Rules

𝑻(𝒇 + 𝒈, 𝒉) = 𝑻(𝒇, 𝒉) + 𝑻(𝒈, 𝒉)
𝑛
1
𝑇(𝑓 + 𝑔, ℎ) = ∑ [ ℎ((𝑓 + 𝑔)(𝑥𝑘−1 ) + (𝑓 + 𝑔)(𝑥𝑘 ))]
2
𝑘=1
𝑛
1
= ∑ [ ℎ[𝑓(𝑥𝑘−1 ) + 𝑔(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 ) + 𝑔(𝑥𝑘 ))]
2
𝑘=1
𝑛
1 1
= ∑ [ ℎ(𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 )) + ℎ(𝑔(𝑥𝑘−1 ) + 𝑔(𝑥𝑘 ))]
2 2
𝑘=1
𝑛 𝑛
1 1
= ∑ ℎ(𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 )) + ∑ ℎ (𝑔(𝑥𝑘−1 ) + 𝑔(𝑥𝑘 ))
2 2
𝑘=1 𝑘=1
= 𝑇(𝑓, ℎ) + 𝑇(𝑔, ℎ)
 𝑺(𝒇 + 𝒈, 𝒉) = 𝑺(𝒇, 𝒉) + 𝑺(𝒈, 𝒉)
𝑛
1
𝑆(𝑓 + 𝑔, ℎ) = ∑ ℎ[(𝑓 + 𝑔)(𝑥2𝑘−2 ) + 4(𝑓 + 𝑔)(𝑥2𝑘−1 ) + (𝑓 + 𝑔)(𝑥2𝑘 )]
3
𝑘=1
𝑛
1
= ∑ ℎ[𝑓(𝑥2𝑘−2 ) + 𝑔(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 4𝑔(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 ) + 𝑔(𝑥2𝑘 )]
3
𝑘=1
𝑛
1 1
= ∑ [ ℎ(𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 ))] + [ ℎ(𝑔(𝑥2𝑘−2 ) + 4𝑔(𝑥2𝑘−1 ) + 𝑔(𝑥2𝑘 ))]
3 3
𝑘=1
𝑛 𝑛
1 1
= ∑ ℎ(𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 )) + ∑ ℎ(𝑔(𝑥2𝑘−2 ) + 4𝑔(𝑥2𝑘−1 ) + 𝑔(𝑥2𝑘 ))
3 3
𝑘=1 𝑘=1
= 𝑆(𝑓, ℎ) + 𝑆(𝑔, ℎ)

𝑻(𝒄𝒇, 𝒉) = 𝒄𝑻(𝒇, 𝒉)
𝑛
1
𝑇(𝑐𝑓, ℎ) = ∑ 𝑐 [ ℎ(𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 ))]
2
𝑘=1
𝑛
1
= 𝑐 ∑ [ ℎ(𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 ))]
2
𝑘=1
= 𝑐𝑇(𝑓, ℎ)

𝑺(𝒄𝒇, 𝒉) = 𝒄𝑺(𝒇, 𝒉)
𝑛
1
𝑆(𝑐𝑓, ℎ) = ∑ 𝑐 [ ℎ(𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 ))]
3
𝑘=1
𝑛
1
= 𝑐 ∑ [ ℎ(𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 ))]
3
𝑘=1
= 𝑐𝑆(𝑓, ℎ)
The degree of precision for the trapezoidal rule is 1
Let 𝑓(𝑥) = 𝑥
Result 1 𝑏
∫ 𝑓(𝑥) 𝑑𝑥
𝑎
𝑛 𝑥𝑘
= ∑∫ 𝑓(𝑥) 𝑑𝑥
𝑘=1 𝑥𝑘−1
𝑛 𝑥
1 𝑘
= ∑ 𝑥2|
2 𝑥𝑘−1
𝑘=1
𝑛
1
= ∑ (𝑥𝑘2 − 𝑥𝑘−1
2
)
2
𝑘=1
𝑛
1
= ∑ (𝑥𝑘 − 𝑥𝑘−1 )(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
𝑛
1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
Result 2 𝑇(𝑓, ℎ)
𝑛
1
= ∑ ℎ(𝑓(𝑥𝑘−1 ) + 𝑓(𝑥𝑘 ))
2
𝑘=1
𝑛
1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2
𝑘=1
We calculate 𝐸(𝑓, ℎ) by combining result 1 and 2
𝑏
𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − 𝑇(𝑓, ℎ)
𝑎
𝑛 𝑛
1 1
= ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 ) − ∑ ℎ(𝑥𝑘 + 𝑥𝑘−1 )
2 2
𝑘=1 𝑘=1
=0
Therefore, the degree of precision is at least 1
The degree of precision for Simpson’s rule is 3
Let 𝑓(𝑥) = 𝑥
Result 1 𝑏
∫ 𝑓(𝑥) 𝑑𝑥
𝑎
𝑛 𝑥2𝑘
= ∑∫ 𝑓(𝑥) 𝑑𝑥
𝑘=1 𝑥2𝑘−2
𝑛
1 2 𝑥2𝑘
=∑ 𝑥 |
2 𝑥2𝑘−2
𝑘=1
𝑛
1 2 2
= ∑ (𝑥2𝑘 − 𝑥2𝑘−2 )
2
𝑘=1
𝑛
1
= ∑ (𝑥2𝑘 − 𝑥2𝑘−2 )(𝑥2𝑘 + 𝑥2𝑘−2 )
2
𝑘=1
𝑛
1
= ∑ 2ℎ(𝑥2𝑘 + 𝑥2𝑘−2 )
2
𝑘=1
𝑛

= ∑ ℎ(𝑥2𝑘 + 𝑥2𝑘−2 )
𝑘=1
Result 2 𝑆(𝑓, ℎ)
𝑛
1
= ∑ ( ℎ(𝑓(𝑥2𝑘−2 ) + 4𝑓(𝑥2𝑘−1 ) + 𝑓(𝑥2𝑘 )))
3
𝑘=1
𝑛
1
= ∑ ℎ(𝑥2𝑘−2 + 4𝑥2𝑘−1 + 𝑥2𝑘 )
3
𝑘=1
𝑛
1 𝑥2𝑘−2 + 𝑥2𝑘
= ∑ ℎ (𝑥2𝑘−2 + 4 ( ) + 𝑥2𝑘 )
3 2
𝑘=1
𝑛
1
= ∑ ℎ(𝑥2𝑘−2 + 2𝑥2𝑘−2 + 2𝑥2𝑘 + 𝑥2𝑘 )
3
𝑘=1
𝑛
1
= ∑ ℎ(3𝑥2𝑘−2 + 3𝑥2𝑘 )
3
𝑘=1
𝑛

= ∑ ℎ(𝑥2𝑘−2 + 𝑥2𝑘 )
𝑘=1
We calculate 𝐸(𝑓, ℎ) by combining result 1 and 2
𝑏
𝐸(𝑓, ℎ) = ∫ 𝑓(𝑥) 𝑑𝑥 − 𝑆(𝑓, ℎ)
𝑎
𝑛 𝑛

= ∑ ℎ(𝑥2𝑘−2 + 𝑥2𝑘 ) − ∑ ℎ(𝑥2𝑘−2 + 𝑥2𝑘 )

𝑘=1 𝑘=1
=0
Therefore, the degree of precision is at least 1
Section 2.3 – Piecewise Continuous Functions and Accuracy
𝒃
𝟏
|∫ 𝒇(𝒙) 𝒅𝒙 − 𝑳𝑹(𝒇, 𝒉)| ≤ 𝑴 (𝒃 − 𝒂)𝒉
𝒂 𝟐 𝟏
Consider the following
𝑛
|𝐸(𝑓, ℎ)| ≤ |𝑒1 | + |𝑒2 | + ⋯ + |𝑒𝑛 | = ∑|𝑒𝑘 |
𝑘=1

Choose 𝑀1 = max|𝑓 ′ (𝑥)| then

[𝑎,𝑏]
max |𝑓 ′ | ≤ max|𝑓 ′ | = 𝑀1
[𝑥𝑘−1 ,𝑥𝑘 ] [𝑎,𝑏]
2
ℎ ℎ2 ℎ2
⇒ |𝑒𝑘 | ≤ max |𝑓 ′ | ≤ max|𝑓 ′ | = 𝑀1
2 [𝑥𝑘−1 ,𝑥𝑘 ] 2 [𝑎,𝑏] 2
Combine these results we get
𝑛
|𝐸(𝑓, ℎ)| ≤ ∑|𝑒𝑘 |
𝑘=1
𝑛
ℎ2
≤∑ 𝑀
2 1
𝑘=1
2
ℎ
=𝑛 𝑀
2 1
1
= 𝑀1 (𝑏 − 𝑎)ℎ
2
𝒃
𝟏
|∫ 𝒇(𝒙) 𝒅𝒙 − 𝑹𝑹(𝒇, 𝒉)| ≤ 𝑴 (𝒃 − 𝒂)𝒉
𝒂 𝟐 𝟏
Consider the following
𝑛
|𝐸(𝑓, ℎ)| ≤ |𝑒1 | + |𝑒2 | + ⋯ + |𝑒𝑛 | = ∑|𝑒𝑘 |
𝑘=1

Choose 𝑀1 = max|𝑓 ′ (𝑥)| then

[𝑎,𝑏]
max |𝑓 ′ | ≤ max|𝑓 ′ | = 𝑀1
[𝑥𝑘−1 ,𝑥𝑘 ] [𝑎,𝑏]
2
ℎ ℎ2 ℎ2
⇒ |𝑒𝑘 | ≤ max |𝑓 ′ | ≤ max|𝑓 ′ | = 𝑀1
2 [𝑥𝑘−1 ,𝑥𝑘 ] 2 [𝑎,𝑏] 2
Combine these results we get
𝑛

𝐸(𝑓, ℎ) ≤ ∑|𝑒𝑘 |
𝑘=1
𝑛
ℎ2
≤∑ 𝑀
2 1
𝑘=1
2
ℎ
=𝑛 𝑀
2 1
1
= 𝑀1 (𝑏 − 𝑎)ℎ
2
 𝒃
𝟏
|∫ 𝒇(𝒙) 𝒅𝒙 − 𝑴𝑹(𝒇, 𝒉)| ≤ 𝑴 (𝒃 − 𝒂)𝒉𝟐
𝒂 𝟐𝟒 𝟐
Consider the following
𝑛
|𝐸(𝑓, ℎ)| ≤ |𝑒1 | + |𝑒2 | + ⋯ + |𝑒𝑛 | = ∑|𝑒𝑘 |
𝑘=1

Choose 𝑀2 = max|𝑓 ′′ (𝑥)| then

[𝑎,𝑏]
max |𝑓 | ≤ max|𝑓 ′ ′| = 𝑀2
′′
[𝑥𝑘−1 ,𝑥𝑘 ] [𝑎,𝑏]
3
ℎ ℎ3 ℎ3
⇒ |𝑒𝑘 | ≤ max |𝑓 ′′ | ≤ max|𝑓 ′′ | = 𝑀
24 [𝑥𝑘−1 ,𝑥𝑘 ] 24 [𝑎,𝑏] 24 2
Combine these results we get
𝑛
|𝐸(𝑓, ℎ)| ≤ ∑|𝑒𝑘 |
𝑘=1
𝑛
ℎ3
≤∑ 𝑀
24 2
𝑘=1
3
ℎ
=𝑛 𝑀
24 2
1
= 𝑀 (𝑏 − 𝑎)ℎ2
24 2
𝒃
𝟏
|∫ 𝒇(𝒙) 𝒅𝒙 − 𝑻(𝒇, 𝒉)| ≤ 𝑴 (𝒃 − 𝒂)𝒉𝟐
𝒂 𝟏𝟐 𝟐
Consider the following
𝑛
|𝐸(𝑓, ℎ)| ≤ |𝑒1 | + |𝑒2 | + ⋯ + |𝑒𝑛 | = ∑|𝑒𝑘 |
𝑘=1

Choose 𝑀2 = max|𝑓 ′′ (𝑥)| then

[𝑎,𝑏]
max |𝑓 ′′ | ≤ max|𝑓 ′ ′| = 𝑀2
[𝑥𝑘−1 ,𝑥𝑘 ] [𝑎,𝑏]
3
ℎ ℎ3 ℎ3
⇒ |𝑒𝑘 | ≤ max |𝑓 ′′ | ≤ max|𝑓 ′′ | = 𝑀
12 [𝑥𝑘−1 ,𝑥𝑘 ] 12 [𝑎,𝑏] 12 2
Combine these results we get
𝑛
|𝐸(𝑓, ℎ)| ≤ ∑|𝑒𝑘 |
𝑘=1
𝑛
ℎ3
≤ ∑ 𝑀2
12
𝑘=1
3
ℎ
=𝑛 𝑀
12 2
1
= 𝑀 (𝑏 − 𝑎)ℎ2
12 2
 𝒃
𝟏
|∫ 𝒇(𝒙) 𝒅𝒙 − 𝑺(𝒇, 𝒉)| ≤ 𝑴 (𝒃 − 𝒂)𝒉𝟒
𝒂 𝟏𝟖𝟎 𝟒
Consider the following
𝑛
|𝐸(𝑓, ℎ)| ≤ |𝑒1 | + |𝑒2 | + ⋯ + |𝑒𝑛 | = ∑|𝑒𝑘 |
𝑘=1

Choose 𝑀4 = max|𝑓 (4) (𝑥)| then

[𝑎,𝑏]
max |𝑓 (4) | ≤ max|𝑓 (4) | = 𝑀4
[𝑥𝑘−1 ,𝑥𝑘 ] [𝑎,𝑏]
ℎ5 (4) ℎ5 (4) ℎ5
|𝑒 |
⇒ 𝑘 ≤ max |𝑓 | ≤ max|𝑓 | = 𝑀
180 [𝑥𝑘−1 ,𝑥𝑘] 180 [𝑎,𝑏] 180 4
Combine these results we get
𝑛
|𝐸(𝑓, ℎ)| ≤ ∑|𝑒𝑘 |
𝑘=1
𝑛
ℎ5
≤∑ 𝑀
180 4
𝑘=1
ℎ5
=𝑛 𝑀
180 4
1
= 𝑀 (𝑏 − 𝑎)ℎ4
180 4
Section 2.4 – Local Errors for Rectangle Rules
𝒙 𝟏
|𝒆𝒌 | = |∫𝒙 𝒌 𝒇 − 𝒇(𝒙∗𝒌 )𝒉𝒌 | ≤ 𝑴𝒉𝟐 (Local Error for LR)
𝒌−𝟏 𝟐

First note that

𝑥𝑘
𝑒𝑘 = ∫ 𝑓(𝑥) 𝑑𝑥 − ℎ(𝑥𝑘−1 ) (from definition)
𝑥𝑘−1
𝑥𝑘 𝑥𝑘
=∫ 𝑓(𝑥) 𝑑𝑥 − ∫ 𝑓(𝑥𝑘−1 ) 𝑑𝑥 (∵ ℎ = 𝑥𝑘 − 𝑥𝑘−1 and 𝑓(𝑥𝑘−1 ) is a constant)
𝑥𝑘−1 𝑥𝑘−1
𝑥𝑘
=∫ 𝑓(𝑥) − 𝑓(𝑥𝑘−1 )𝑑𝑥
𝑥𝑘−1
Therefore,
𝑥𝑘
|𝑒𝑘 | = |∫ 𝑓(𝑥) − 𝑓(𝑥𝑘−1 )𝑑𝑥|
𝑥𝑘−1
𝑥𝑘
≤∫ |𝑓(𝑥) − 𝑓(𝑥𝑘−1 )|𝑑𝑥 
𝑥𝑘−1

Now,
Let 𝑀 be an upper bound for |𝑓 ′ | on [𝑥𝑘−1 , 𝑥𝑘 ]
From the Mean Value Theorem, 𝑓(𝑏) − 𝑓(𝑎) = 𝑓 ′ (𝑐)(𝑏 − 𝑎)
Therefore 𝑓(𝑥) − 𝑓(𝑥𝑘−1 ) = 𝑓 ′ (𝑐)(𝑥 − 𝑥𝑘−1 ) where 𝑐 ∈ [𝑥𝑘−1 , 𝑥𝑘 ]
∴ |𝑓(𝑥) − 𝑓(𝑥𝑘−1 )| = |𝑓 ′ (𝑐)||𝑥 − 𝑥𝑘−1 | ≤ 𝑀|𝑥 − 𝑥𝑘−1 | 
𝑥𝑘
|𝑒𝑘 | ≤ ∫ |𝑓(𝑥) − 𝑓(𝑥𝑘−1 )| 𝑑𝑥 ()
𝑥𝑘−1
𝑥
≤ ∫𝑥 𝑘 𝑀|𝑥 − 𝑥𝑘−1 | 𝑑x ()
𝑘−1
𝑥𝑘
1
= 𝑀 ( 𝑥 2 − 𝑥. 𝑥𝑘−1 | )
2 𝑥𝑘−1 (evaluating the integral)
1 (algebra)
= 𝑀 [ (𝑥𝑘2 − 𝑥𝑘−1
2
) − 𝑥𝑘 (𝑥𝑘 − 𝑥𝑘−1 )]
2
1 (algebra)
= 𝑀 [ (𝑥𝑘 − 𝑥𝑘−1 )(𝑥𝑘 + 𝑥𝑘−1 ) − 𝑥𝑘 (𝑥𝑘 − 𝑥𝑘−1 )]
2
1 (algebra)
= 𝑀 [ ℎ(𝑥𝑘 − 𝑥𝑘−1 )]
2
1
= 𝑀ℎ2 (ℎ = (𝑥𝑘−1 − 𝑥𝑘 ))
2
 𝒙 𝟏
|𝒆𝒌 | = |∫𝒙 𝒌 𝒇 − 𝒇(𝒙∗𝒌 )𝒉𝒌 | ≤ 𝑴𝒉𝟐 (Local Error for RR)
𝒌−𝟏 𝟐

First note that

𝑥𝑘
𝑒𝑘 = ∫ 𝑓(𝑥) 𝑑𝑥 − ℎ(𝑥𝑘 ) (from definition)
𝑥𝑘−1
𝑥𝑘 𝑥𝑘
=∫ 𝑓(𝑥) 𝑑𝑥 − ∫ 𝑓(𝑥𝑘 ) 𝑑𝑥 (because ℎ = 𝑥𝑘 − 𝑥𝑘−1 and 𝑓(𝑥𝑘∗ ) is a constant)
𝑥𝑘−1 𝑥𝑘−1
𝑥𝑘
=∫ 𝑓(𝑥) − 𝑓(𝑥𝑘 )𝑑𝑥
𝑥𝑘−1
Therefore,
𝑥𝑘
|𝑒𝑘 | = |∫ 𝑓(𝑥) − 𝑓(𝑥𝑘 )𝑑𝑥|
𝑥𝑘−1
𝑥𝑘
≤∫ |𝑓(𝑥) − 𝑓(𝑥𝑘 )|𝑑𝑥 
𝑥𝑘−1

Now,
Let 𝑀 be an upper bound for |𝑓 ′ | on [𝑥𝑘−1 , 𝑥𝑘 ]
From the Mean Value Theorem, 𝑓(𝑏) − 𝑓(𝑎) = 𝑓 ′ (𝑐)(𝑏 − 𝑎)
Therefore 𝑓(𝑥) − 𝑓(𝑥𝑘 ) = 𝑓 ′ (𝑐)(𝑥 − 𝑥𝑘 ) where 𝑐 ∈ [𝑥, 𝑥𝑘∗ ]
∴ |𝑓(𝑥) − 𝑓(𝑥𝑘 )| = |𝑓 ′ (𝑐)||𝑥 − 𝑥𝑘 | ≤ 𝑀|𝑥 − 𝑥𝑘 | 
𝑥𝑘
|𝑒𝑘 | ≤ ∫ |𝑓(𝑥) − 𝑓(𝑥𝑘 )| 𝑑𝑥 ()
𝑥𝑘−1
𝑥
≤ ∫𝑥 𝑘 𝑀|𝑥 − 𝑥𝑘 | 𝑑x ()
𝑘−1
𝑥𝑘
1 (evaluating the integral)
= 𝑀 (𝑥. 𝑥𝑘 − 𝑥 2 | )
2 𝑥𝑘−1
1 (algebra)
= 𝑀 [𝑥𝑘 (𝑥𝑘 − 𝑥𝑘−1 ) − (𝑥𝑘2 − 𝑥𝑘−1
2
)]
2
1
= 𝑀 [𝑥𝑘 (𝑥𝑘 − 𝑥𝑘−1 ) − (𝑥𝑘 − 𝑥𝑘−1 )(𝑥𝑘 + 𝑥𝑘−1 )] (algebra)
2
1
= 𝑀 [ℎ𝑥𝑘 − ℎ(𝑥𝑘 + 𝑥𝑘−1 )]
2 (algebra)
1 2
= 𝑀ℎ (ℎ = (𝑥𝑘−1 − 𝑥𝑘 ))
2
Section 3.1 – First Order Differential Equations: Euler’s Method
Derivation of Euler’s Method
Consider a general initial value problem 𝑦 ′ (𝑡) = 𝑓(𝑦(𝑡)) with 𝑦(𝑡0 ) = 𝑦0 .

The tangent to the graph of the solution 𝑦 through the point (𝑡0 , 𝑦0 ) has slope:
𝑦 ′ (𝑡0 ) = 𝑓(𝑦(𝑡0 )) = 𝑓(𝑦0 )
𝑦1 −𝑦0
If (𝑡1 , 𝑦1 ) is the point on the tangent at 𝑡1 = 𝑡0 + ℎ, the slope of the tangent is also given by . From this it
ℎ
follows that 𝑦1 = 𝑦0 + ℎ𝑓(𝑦0 )

Now to calculate 𝑦2 : The tangent to the graph of the solution 𝑦 at the point (𝑡1 , 𝑦(𝑡1 )) has a slope 𝑦 ′ (𝑡1 ) =
𝑓(𝑦(𝑡1 )).

As only an estimate 𝑦1 for 𝑦(𝑡1 ) is known, we only have an estimate for this slope: 𝑦 ′ (𝑡1 ) = 𝑓(𝑦1 )
Also the point (𝑡1 , 𝑦(𝑡1 )) is unknown and only the point (𝑡1 , 𝑦1 ) close to the graph of 𝑦 is known.

For the time interval [𝑡1 , 𝑡2 ] the function 𝑦 can be approximated by a line (not a tangent line) through the point with
𝑦 −𝑦
(𝑡1 , 𝑦1 ) with slope 𝑓(𝑦1 ). It then follows that 2 1 = 𝑓(𝑦1 ) and 𝑦2 = 𝑦1 + ℎ𝑓(𝑦1 )
ℎ

Section 3.2 – First Order Differential Equations: Improved Euler’s Method

Derivation of Improved Euler’s Method
For an initial value problem 𝑦 ′ (𝑡) = 𝑓(𝑦(𝑡)), 𝑦(𝜏) = 𝑏
We choose evenly spaced times 𝜏 = 𝑡0 < 𝑡1 < 𝑡2 < ⋯ < 𝑡𝑛 < 𝑇 with step length ℎ.

For the interval [𝑡𝑘 , 𝑡𝑘+1 ] it follows from the differential equations and the second fundamental theorem of calculus
that
𝑡𝑘+1
𝑦(𝑡𝑘+1 ) − 𝑦(𝑡𝑘 ) = ∫ 𝑓(𝑦(𝑡)) 𝑑𝑡
𝑡𝑘

Using the trapezoidal rule for numerical integration it follows that

ℎ
𝑦𝑘+1 ≈ 𝑦𝑘 + (𝑓(𝑦(𝑡𝑘 )) + 𝑓(𝑦(𝑡𝑘 + 1)))
2
ℎ
≈ 𝑦𝑘 + (𝑓(𝑦𝑘 ) + 𝑓(𝑦𝑘+1 ))
2
If 𝑦𝑘 is known, the value of 𝑦𝑘+1 cannot be calculated directly from this equation due to the term 𝑓(𝑦𝑘+1 ). An
estimate (the predictor) 𝑝𝑘 for 𝑦𝑘+1 is obtained using Euler’s method:
𝑝𝑘 = 𝑦𝑘 + ℎ(𝑓(𝑦𝑘 )) ≈ 𝑦𝑘+1

And therefore, we have

ℎ
𝑦𝑘+1 = 𝑦𝑘 + (𝑓(𝑦(𝑡𝑘 )) + 𝑓(𝑦(𝑡𝑘 + 1)))
2

Section 3.3 – Accuracy

No Proofs

Section 3.4 – Systems of Differential Equations

No Proofs

Section 3.5 – Higher Order Differential Equations

No Proofs

Section 4.1 – Gauss Elimination

No Proofs

Section 4.2 – Iterative Methods for Linear Systems

No Proofs
3. Graphs
Section 1.1 – Graphs and Interval Division
No graphs

Section 1.2 – Polynomial Equations

No graphs

Section 1.3 – Fixed Point Iteration

Graphical Illustration of Convergence and Divergence

Section 1.4 – Newton-Raphson Method

Graphical Illustration of the Newton-Raphson Method
Section 1.5 – Convergence
No Graphs

Section 2.1 – Numerical Integration and Rectangle Rules

No Graphs

Section 2.2 – Trapezoidal and Simpson Rules

No Graphs

Section 2.3 – Piecewise Continuous Functions and Accuracy

No Graphs

Section 2.4 – Local Errors for Rectangle Rules

No Graphs

Section 3.1 – First Order Differential Equations: Euler’s Method

Graphical Illustration of Euler’s Method
Section 3.2 – First Order Differential Equations: Improved Euler’s Method
Graphical Illustration of Improved Euler’s Method

Section 3.3 – Accuracy

No graphs

Section 3.4 – Systems of Differential Equations

No graphs

Section 3.5 – Higher Order Differential Equations

No graphs

Section 4.1 – Gauss Elimination

No graphs

Section 4.2 – Iterative Methods for Linear Systems

No graphs