Numerical Methods — Exam Preparation Exam: 17 March 2026

Numerical Methods — Complete Study Guide

Covering: Root Finding, Linear Systems, Interpolation
Exam: 17 March 2026

Contents

1 UNIT 1: Root Finding Methods 2

1.1 1.1 Bisection (Section) Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 1.2 Fixed Point Iteration Method . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 1.3 Newton–Raphson Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 1.4 Secant Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 1.5 Rates of Convergence — Summary . . . . . . . . . . . . . . . . . . . . . . . 7

2 UNIT 2: Systems of Linear Equations 8

2.1 2.1 Gauss Elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 2.2 Gauss–Seidel Iteration Method . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 2.3 Jacobi Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 UNIT 3: Finite Difference Operators & Interpolation 10

3.1 3.1 Finite Difference Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 3.2 Forward Difference Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 3.3 Newton’s Forward Interpolation Formula . . . . . . . . . . . . . . . . . . . . 11
3.4 3.4 Newton’s Backward Interpolation Formula . . . . . . . . . . . . . . . . . . . 11

4 Quick Revision — All Key Formulas at a Glance 13

1
Numerical Methods — Exam Preparation Exam: 17 March 2026

UNIT 1: Root Finding Methods

1.1 Bisection (Section) Method

Definition & Concept

The Bisection Method finds a root of f (x) = 0 in [a, b] by repeatedly halving the
interval.
Existence Condition (IVT): If f is continuous on [a, b] and f (a) · f (b) < 0, then at
least one root exists in (a, b).

Algorithm and Error Bound

1. Check f (a) · f (b) < 0 (sign change).

a+b
2. Compute midpoint: c =
2
3. If f (c) = 0, c is the root. Stop.
4. If f (a) · f (c) < 0, set b = c; else set a = c.
5. Repeat from step 2.

b−a
Error bound after n iterations: |en | ≤
2n

Q1 (Exam Sheet). Bisection Method — f (x) = x3 − 4x − 9 = 0 on [2, 3], 4

iterations
Step 0 — Verify existence condition:

f (2) = 8 − 8 − 9 = −9 < 0, f (3) = 27 − 12 − 9 = 6 > 0

Since f (2) · f (3) = (−9)(6) = −54 < 0, a root exists in (2, 3). ✓

Iteration 1: a = 2, b = 3
2+3
c1 = = 2.5, f (2.5) = (2.5)3 − 4(2.5) − 9 = 15.625 − 10 − 9 = −3.375 < 0
2
Since f (a) and f (c1 ) are both negative: set a = 2.5. New interval: [2.5, 3].

Iteration 2: a = 2.5, b = 3
2.5 + 3
c2 = = 2.75, f (2.75) = 20.797 − 11 − 9 = 0.797 > 0
2
f (a) < 0, f (c2 ) > 0: set b = 2.75. New interval: [2.5, 2.75].

Iteration 3: a = 2.5, b = 2.75

2.5 + 2.75
c3 = = 2.625, f (2.625) = 18.088 − 10.5 − 9 = −1.412 < 0
2
Set a = 2.625. New interval: [2.625, 2.75].

Iteration 4: a = 2.625, b = 2.75

2.625 + 2.75
c4 = = 2.6875, f (2.6875) = 19.404 − 10.75 − 9 = −0.346 < 0
2

2
Numerical Methods — Exam Preparation Exam: 17 March 2026

New interval: [2.6875, 2.75].

Summary Table:

n a b c = (a + b)/2 f (a) f (c) New Interval

1 2.000 3.000 2.500 −9.000 −3.375 [2.5, 3.0]
2 2.500 3.000 2.750 −3.375 +0.797 [2.5, 2.75]
3 2.500 2.750 2.625 −3.375 −1.412 [2.625, 2.75]
4 2.625 2.750 2.688 −1.412 −0.346 [2.688, 2.75]

Convergence comment: After 4 iterations the root is narrowed to [2.688, 2.750]. Error
bound = (3 − 2)/24 = 0.0625. The interval shrinks by half each step — this is linear
convergence with rate 1/2.

Q1 (Assignment). Bisection Method — f (x) = x3 − 4x − 9 = 0 correct to 3

decimal places

Continuing from above, we need error < 0.0005, so n such that (b − a)/2n < 0.0005, i.e.
2n > 2000, giving n ≥ 11 iterations.
Continuing the table:

n a b c
5 2.688 2.750 2.719
6 2.719 2.750 2.734
7 2.734 2.750 2.742
8 2.734 2.742 2.738
9 2.738 2.742 2.740
10 2.094 2.096 2.095
11 Root ≈ 2.706

The root to 3 decimal places is x∗ ≈ 2.706.

Q2 (Exam Sheet). Bisection — f (x) = x − cos x = 0 on [0, 1], 3 decimal places

Verify: f (0) = 0 − cos(0) = 0 − 1 = −1 < 0; f (1) = 1 − cos(1) = 1 − 0.5403 = 0.4597 >

0. ✓

n a b c = (a + b)/2 f (c) = c − cos c

1 0.000 1.000 0.5000 0.500 − 0.878 = −0.378 (neg) ⇒ a = 0.5
2 0.500 1.000 0.7500 0.750 − 0.732 = +0.018 (pos) ⇒ b = 0.75
3 0.500 0.750 0.6250 0.625 − 0.812 = −0.187 (neg) ⇒ a = 0.625
4 0.625 0.750 0.6875 0.688 − 0.773 = −0.085 (neg) ⇒ a = 0.6875
5 0.688 0.750 0.7188 0.719 − 0.752 = −0.033 (neg) ⇒ a = 0.7188
6 0.719 0.750 0.7344 0.734 − 0.741 = −0.007 (neg) ⇒ a = 0.7344
7 0.734 0.750 0.7422 0.742 − 0.737 = +0.005 (pos) ⇒ b = 0.7422
8 0.734 0.742 0.7383 0.738 − 0.739 = −0.001 (neg) ⇒ a = 0.7383
9 0.738 0.742 0.7402 0.740 − 0.738 ≈ 0

Root ≈ 0.739 correct to 3 decimal places.

3
Numerical Methods — Exam Preparation Exam: 17 March 2026

1.2 Fixed Point Iteration Method

Concept

Rewrite f (x) = 0 as x = g(x). Starting from x0 , iterate:

xn+1 = g(xn )

until |xn+1 − xn | < ε (tolerance).

Convergence Condition and Order

• Condition: Converges if |g ′ (x)| < 1 in a neighbourhood of the root.

• Order: 1 (linear). Error en+1 ≈ g ′ (x∗ ) en .
• If |g ′ (x∗ )| > 1: diverges.

Assignment Q2. Fixed Point — Solve x = cos x, x0 = 0.5, 4 iterations

Here g(x) = cos x, so g ′ (x) = − sin x.

Convergence check at root x∗ ≈ 0.739: |g ′ (0.739)| = | sin(0.739)| ≈ 0.674 < 1.
Converges.

n xn xn+1 = cos(xn )
0 0.5000 cos(0.5000) = 0.8776
1 0.8776 cos(0.8776) = 0.6390
2 0.6390 cos(0.6390) = 0.8026
3 0.8026 cos(0.8026) = 0.6947
4 0.6947 cos(0.6947) = 0.7659

The iterations oscillate and converge toward x∗ ≈ 0.7391.

Order of Convergence: Linear (first-order): |en+1 | ≈ |g ′ (x∗ )| |en | ≈ 0.674 |en |

1.3 Newton–Raphson Method

Formula
f (xn )
xn+1 = xn −
f ′ (xn )
Starting guess: x0 . Derived from Taylor series truncated at first-order.

Rate of Convergence — Quadratic

For a simple root x∗ with f ′ (x∗ ) ̸= 0:

|f ′′ (x∗ )|
|en+1 | ≈ |en |2
2|f ′ (x∗ )|

Each iteration approximately doubles the number of correct digits.

Quadratic means: if error is 0.1 at step n, it becomes ≈ 0.01 at step n + 1.

4
Numerical Methods — Exam Preparation Exam: 17 March 2026

Assignment Q3. Newton–Raphson — f (x) = x3 − 2x − 5 = 0, x0 = 2, 3 iterations

f (x) = x3 − 2x − 5, f ′ (x) = 3x2 − 2.

Iteration 1:
f (2) = 8 − 4 − 5 = −1, f ′ (2) = 3(4) − 2 = 10
−1
x1 = 2 − = 2 + 0.1 = 2.1000
10
Iteration 2:

f (2.1) = (2.1)3 − 2(2.1) − 5 = 9.261 − 4.2 − 5 = 0.061

f ′ (2.1) = 3(2.1)2 − 2 = 3(4.41) − 2 = 11.23

0.061
x2 = 2.1 − = 2.1 − 0.00543 = 2.0946
11.23
Iteration 3:

f (2.0946) = (2.0946)3 − 2(2.0946) − 5 ≈ 9.194 − 4.189 − 5 = 0.005

f ′ (2.0946) = 3(2.0946)2 − 2 ≈ 3(4.387) − 2 = 11.162

0.005
x3 = 2.0946 − ≈ 2.09455
11.162
Summary Table:

n xn f (xn ) f ′ (xn ) xn+1 = xn − f /f ′

0 2.0000 −1.0000 10.000 2.1000
1 2.1000 +0.0610 11.230 2.0946
2 2.0946 +0.0050 11.162 2.09455
3 2.09455 ≈0 — —

Root ≈ 2.0946.
Convergence comment: Errors: 0.1 → 0.005 → 0.00005. Each step the error squares
(approximately). This is quadratic convergence.

1.4 Secant Method

Formula
Avoids computing f ′ (x) by approximating the derivative with a finite difference:
xn − xn−1
xn+1 = xn − f (xn ) ·
f (xn ) − f (xn−1 )

Requires two initial guesses x0 and x1 .

5
Numerical Methods — Exam Preparation Exam: 17 March 2026

Order of Convergence

Superlinear convergence with order p ≈ 1.618 (the golden ratio ϕ):

|en+1 | ≈ C|en |1.618

Faster than bisection (linear), but slower than Newton–Raphson (quadratic).

Advantage: No derivative f ′ needed.

Q3 (Exam Sheet). Secant Method — f (x) = x3 − 2x − 5 = 0, x0 = 2, x1 = 3, 3

iterations
Iteration 1 (using x0 = 2, x1 = 3):

f (2) = 8 − 4 − 5 = −1, f (3) = 27 − 6 − 5 = 16

3−2 16
x2 = 3 − 16 · =3− = 3 − 0.9412 = 2.0588
16 − (−1) 17
Iteration 2 (using x1 = 3, x2 = 2.0588):

f (2.0588) = (2.0588)3 − 2(2.0588) − 5 ≈ 8.730 − 4.118 − 5 = −0.387

2.0588 − 3 −0.9412
x3 = 2.0588 − (−0.387) · = 2.0588 − (−0.387) ·
−0.387 − 16 −16.387
= 2.0588 + 0.0222 = 2.0810
Iteration 3 (using x2 = 2.0588, x3 = 2.0810):

f (2.0810) = (2.0810)3 − 2(2.0810) − 5 ≈ 9.014 − 4.162 − 5 = −0.148

2.0810 − 2.0588
x4 = 2.0810 − (−0.148) · = 2.0810 + 0.0137 = 2.0947
−0.148 − (−0.387)
Root ≈ 2.0946.
Comparison with Bisection:

Method Iterations (to 4 d.p.) Order Needs f ′ ?

Bisection ∼ 14 1 (linear) No
Secant ∼6 1.618 (superlinear) No
Newton–Raphson ∼4 2 (quadratic) Yes

Assignment Q4. Secant Method — f (x) = e−x − x = 0, x0 = 0, x1 = 1, 3

iterations
Iteration 1:

f (0) = e0 − 0 = 1, f (1) = e−1 − 1 ≈ 0.3679 − 1 = −0.6321

1−0 1
x2 = 1 − (−0.6321) · = 1 − (−0.6321) · = 1 − 0.3874 = 0.6126
−0.6321 − 1 −1.6321
Iteration 2:

f (0.6126) = e−0.6126 − 0.6126 ≈ 0.5421 − 0.6126 = −0.0705

6
Numerical Methods — Exam Preparation Exam: 17 March 2026

0.6126 − 1 −0.3874
x3 = 0.6126 − (−0.0705) · = 0.6126 − (−0.0705) ·
−0.0705 − (−0.6321) 0.5616
= 0.6126 − 0.04864 = 0.5640
Iteration 3:

f (0.5640) = e−0.564 − 0.5640 ≈ 0.5686 − 0.5640 = +0.0046

x4 ≈ 0.5671
True root: x∗ = 0.56714 . . . ✓
Comparison with NR: For f (x) = e−x − x, f ′ (x) = −e−x − 1. NR converges in ∼ 3
iterations (quadratic); Secant needs ∼ 4–5 but does NOT need f ′ .

1.5 Rates of Convergence — Summary

Definition of Order of Convergence

An iterative sequence {xn } converges to x∗ with order p if:

|en+1 |
lim =C (C > 0, C < ∞)
n→∞ |en |p

where en = xn − x∗ is the error. If p = 1: linear; p = 2: quadratic.

Bisection is Linearly Convergent — Proof Sketch

b−a
After n steps: |en | ≤ . Then:
2n
|en+1 | (b − a)/2n+1 1
= n
=
|en | (b − a)/2 2

|en+1 | 1
So lim = — linear convergence with asymptotic constant C = 1/2.
|en | 2

Method Order p Asymptotic constant C Remarks

1
Bisection 1 2
Always converges; guaranteed
Fixed-Point 1 |g ′ (x∗ )| Need |g ′ | < 1 near root
Secant 1.618 Depends on f ′′ No derivative needed
Newton–Raphson 2 |f ′′ (x∗ )|/(2|f ′ (x∗ )|) Fastest; needs f ′ (x)

7
Numerical Methods — Exam Preparation Exam: 17 March 2026

UNIT 2: Systems of Linear Equations

2.1 Gauss Elimination

Method
Transform the augmented matrix [A | b] to upper triangular form via row operations,
then solve by back substitution.

Example — Gauss Elimination

Solve: 2x + y − z = 8;
 −3x − y + 2z = −11; −2x + y + 2z = −3.
2 1 −1 | 8
Augmented matrix: −3 −1 2 | −11
−2 1 2 | −3  
2 1 −1 | 8
Step 1: R2 ← R2 + 23 R1 ; R3 ← R3 + R1 : 0 1/2 1/2 | 1
 0 2 1 | 5
2 1 −1 | 8
Step 2: R3 ← R3 − 4R2 : 0 1/2 1/2 | 1
0 0 −1 | 1
Back substitution: z = −1; y = 2(1 − 12 (−1)) = 3; x = 12 (8 − 3 + (−1)) = 2.
Solution: x = 2, y = 3, z = −1.

2.2 Gauss–Seidel Iteration Method

Method
For system Ax = b, isolate each variable and update using latest values:
!
(k+1) 1 X (k+1)
X (k)
xi = bi − aij xj − aij xj
aii j<i j>i

Uses updated values immediately (unlike Jacobi which uses only old values).

Diagonal Dominance — Convergence Condition

Strictly diagonally dominant means:
X
|aii | > |aij | for all i
j̸=i

If this holds, Gauss–Seidel is guaranteed to converge.

Gauss–Seidel converges faster than Jacobi because it uses updated values immediately.

Example — Gauss–Seidel, 3 iterations

Solve: 10x + 2y + z = 9; 2x + 20y − 2z = −44; −2x + 3y + 10z = 22.
Diagonal dominance check: Row 1: |10| > |2|+|1| = 3 ✓; Row 2: |20| > |2|+|2| = 4
✓; Row 3: |10| > |2| + |3| = 5 ✓. Convergence guaranteed.

8
Numerical Methods — Exam Preparation Exam: 17 March 2026

Iterative formulae:
1 1 1
x= 10
(9 − 2y − z), y= 20
(−44 − 2x + 2z), z= 10
(22 + 2x − 3y)

Start: x(0) = y (0) = z (0) = 0.

Iteration 1:
x(1) = 9
10
= 0.9
−45.8
y (1) = 1
20
(−44 − 2(0.9) + 0) = 20
= −2.29
z (1) = 1
10
(22 + 2(0.9) − 3(−2.29)) = 22+1.8+6.87
10
= 3.067
Iteration 2:

x(2) = 1
10
(9 − 2(−2.29) − 3.067) = 9+4.58−3.067
10
= 1.051

y (2) = 1
20
(−44 − 2(1.051) + 2(3.067)) = −1.999
z (2) = 1
10
(22 + 2(1.051) − 3(−1.999)) = 3.010
Iteration 3: x(3) ≈ 1.000, y (3) ≈ −2.000, z (3) ≈ 3.000
Solution: x = 1, y = −2, z = 3.

2.3 Jacobi Method

Formula and Difference from Gauss–Seidel

All unknowns are updated using only the previous iteration’s values:
!
(k+1) 1 X (k)
xi = bi − aij xj
aii j̸=i

Gauss–Seidel > Jacobi in speed because G-S uses updated values immediately. Both
require diagonal dominance for guaranteed convergence.

9
Numerical Methods — Exam Preparation Exam: 17 March 2026

UNIT 3: Finite Difference Operators & Interpolation

3.1 Finite Difference Operators
Let yi = f (x0 + ih), h = step size (uniform spacing).

Operator Symbol Definition Example

Forward Difference ∆ ∆yi = yi+1 − yi ∆y0 = y1 − y0

Backward Difference ∇ ∇yi = yi − yi−1 ∇y1 = y1 − y0

Central Difference δ δyi = yi+1/2 − yi−1/2 δy1 = y3/2 − y1/2

Average (Mean) µ µyi = 12 (yi+1/2 + yi−1/2 )

Shift Operator E Eyi = yi+1 E 2 y0 = y2

dy
Differential D Dy = dx

Relations Between Difference Operators

∆=E−1 ⇒ E =1+∆
∇ = 1 − E −1 ⇒ E −1 = 1 − ∇
δ = E 1/2 − E −1/2
µ = 21 (E 1/2 + E −1/2 )
1
∆ = ehD − 1 ⇔ D= ln(1 + ∆)
h
∇ = 1 − e−hD

3.2 Forward Difference Table

Example — Construct Forward Difference Table for y = x3

x y ∆y ∆2 y ∆3 y ∆4 y
1 1 — — — —
7
2 8 12
19 6
3 27 18 0
37 6
4 64 24
61
5 125

Note: For a polynomial of degree n, the nth differences are constant and all higher

10
Numerical Methods — Exam Preparation Exam: 17 March 2026

differences are zero. Here ∆3 y = 6 = 3!, ∆4 y = 0.

3.3 Newton’s Forward Interpolation Formula

When to Use
Use when estimating f (x) near the beginning of the table.

Newton’s Forward Interpolation

x − x0
Let s = (note: s > 0 for forward interpolation).
h
s(s − 1) 2 s(s − 1)(s − 2) 3
f (x) = y0 + s∆y0 + ∆ y0 + ∆ y0 + · · ·
2! 3!
n  
X s
Using binomial coefficient notation: f (x) = ∆k y0
k=0
k

Example — Newton’s Forward Interpolation, find f (1.1)

1.1 − 1.0
Given h = 0.2, x0 = 1.0, so s = = 0.5.
0.2

x y ∆y ∆2 y ∆3 y ∆4 y
1.0 1.000
0.728
1.2 1.728 0.288
1.016 0.048
1.4 2.744 0.336 0
1.352 0.048
1.6 4.096 0.384
1.736
1.8 5.832

Applying the formula with s = 0.5:

(0.5)(−0.5) (0.5)(−0.5)(−1.5)
f (1.1) = 1.000 + (0.5)(0.728) + (0.288) + (0.048)
2 6
= 1.000 + 0.364 − 0.036 + 0.003
= 1.331

Exact: (1.1)3 = 1.331 ✓

3.4 Newton’s Backward Interpolation Formula

When to Use
Use when estimating f (x) near the end of the table.

11
Numerical Methods — Exam Preparation Exam: 17 March 2026

Newton’s Backward Interpolation

x − xn
Let s = where xn is the last tabulated point (so s ≤ 0).
h
s(s + 1) 2 s(s + 1)(s + 2) 3
f (x) = yn + s∇yn + ∇ yn + ∇ yn + · · ·
2! 3!
Differences are read from the bottom row of the table.

Example — Newton’s Backward Interpolation, find f (1.7)

1.7 − 1.8
Using the same data, xn = 1.8, h = 0.2, s = = −0.5.
0.2
Backward differences from bottom: ∇yn = 1.736, ∇2 yn = 0.384, ∇3 yn = 0.048,
∇4 yn = 0.

(−0.5)(0.5) (−0.5)(0.5)(1.5)
f (1.7) = 5.832 + (−0.5)(1.736) + (0.384) + (0.048)
2 6
= 5.832 − 0.868 − 0.048 − 0.003
= 4.913

Exact: (1.7)3 = 4.913 ✓

12
Numerical Methods — Exam Preparation Exam: 17 March 2026

Quick Revision — All Key Formulas at a Glance

Method Formula Order / Note

a+b b−a
Bisection c= ; error ≤ n Order 1, linear
2 2

Fixed Point xn+1 = g(xn ); converge if |g ′ | < 1 Order 1, linear

f (xn )
Newton–Raphson xn+1 = xn − Order 2, quadratic
f ′ (xn )
xn − xn−1
Secant xn+1 = xn − f (xn ) · Order 1.618, super-
f (xn ) − f (xn−1 )
linear

(k+1) 1 P 
Gauss–Seidel xi = bi − j̸=i aij xj Converges if diag.
aii
dominant

s(s − 1) 2 x − x0
Newton Fwd f (x) = y0 + s∆y0 + ∆ y0 + · · · s = , near
2! h
start

s(s + 1) 2 x − xn
Newton Bwd f (x) = yn + s∇yn + ∇ yn + · · · s = , near
2! h
end

Operators E = 1+∆; ∇ = 1−E −1 ; ∆ = ehD − Relations

1

Key Points for Exam

Exam Tips:

1. Bisection: Always verify f (a) · f (b) < 0 first. Show complete table.
2. Fixed Point: State |g ′ (x)| < 1 condition before iterating.
3. Newton–Raphson: Show f (xn ) and f ′ (xn ) clearly in a table each step.
4. Secant: Two initial guesses needed. Do not mix up formula direction.
5. Gauss–Seidel: Check diagonal dominance before starting (easy marks).
6. Interpolation: Forward = use from top (near beginning); Backward = use from
bottom (near end).
7. Difference table: Fill carefully column by column. Easy marks if done correctly.
8. Convergence order: Bisection=1, P FPI=1, Secant=1.618, Newton=2.
9. Diagonal dominance: |aii | > j̸=i |aij | for each row.
10. Operator relations: E = 1 + ∆; ∆ = E − 1; ∇ = 1 − E −1 — memorise these.

13