Complex Analysis

1. Differentiability: CR necessary, not sufficient

2
𝑥𝑦
a. Counter example: 𝑓(𝑥 + 𝑖𝑦) = 2 2
𝑥 +𝑦
1 1
b. CR in polar form: 𝑢𝑟 = 𝑟
𝑣, 𝑟
𝑢 = − 𝑣𝑟
θ θ
2. Analytic:
a. Infinitely differentiable on the domain ⇔ Analytic/Holomorphic
b. Analytic on the entire finite complex plane ⇔ Entire function
c. Necessary condition: CR equations satisfied (sufficient: partial derivatives are continuous)
d. Analytic function ⇒ 𝑢, 𝑣 are harmonic
e. Milne-Thompson’s method:
𝑓(𝑧) = 𝑢(𝑧, 0) + 𝑖𝑣(𝑧, 0) ⇒ 𝑓'(𝑧) = 𝑢𝑥(𝑧, 0) + 𝑖𝑣𝑥(𝑧, 0) = 𝑢𝑥(𝑧, 0) − 𝑖𝑢𝑦(𝑧, 0) = 𝑣𝑦(𝑧, 0) + 𝑖𝑣𝑥(𝑧
i. (1 + 𝑖)𝑓'(𝑧) = (𝑢𝑥 − 𝑣𝑥) − 𝑖(𝑢𝑦 − 𝑣𝑦)
ii. (1 − 𝑖)𝑓'(𝑧) = (𝑢𝑥 + 𝑣𝑥) − 𝑖(𝑢𝑦 + 𝑣𝑦)
3. Integration:
a. Simply connected: every simple closed curve contains only points of domain

b. Cauchy-Goursat theorem: ∮ 𝑓(𝑧)𝑑𝑧 = 0, for 𝑓 analytic in C

𝐶

𝑓(𝑧)
c. Cauchy’s first integral theorem: ∮ 𝑑𝑧 = 2π𝑖 𝑓(𝑎)
𝑧−𝑎
𝐶
i. Proof: 𝑓(𝑧) = [𝑓(𝑧) − 𝑓(𝑎)] + 𝑓(𝑎) on small circular loop of radius
𝑟 < δ: |𝑓(𝑧) − 𝑓(𝑎)| < ϵ∀|𝑧 − 𝑎| < δ

𝑓(𝑧) 2π𝑖 (𝑛)

d. Cacuhy’s general integral theorem: ∮ 𝑛+1 𝑑𝑧 = 𝑓 (𝑎)
𝑛!
𝐶 (𝑧−𝑎)
i. Proof using Cauchy’s first integral theorem and induction

e. Residue theorem: ∮ 𝑓(𝑧)𝑑𝑧 = 2π𝑖 Σ 𝑅𝑒𝑠(𝑓, 𝑧𝑖)

𝐶
i. Simple pole: 𝑅𝑒𝑠(𝑓, 𝑧0) = 𝑎−1 = lim (𝑧 − 𝑧0)𝑓(𝑧)
𝑧 → 𝑧0
𝑘−1
1 𝑑 𝑘
ii. Pole of order k: 𝑅𝑒𝑠(𝑓, 𝑧0) = 𝑎−1 = lim
(𝑘−1)! 𝑧 → 𝑧 𝑑𝑧𝑘−1
[(𝑧 − 𝑧0) 𝑓(𝑧)]
0

f. Morera theorem (partial converse of Cauchy’s theorem for multiply connected domains):

∮ 𝑓(𝑧)𝑑𝑧 = 0 ∀𝐶 ⊂ 𝐷 ⇒ 𝑓 is analytic in 𝐷, 𝐶 is any simple closed contour contained in domain D

𝐶
4. Expansions:

Utkarsh Kumar
 1 2
a. Taylor’s expansion for analytic function, 𝑓(𝑧) = 𝑓(𝑧0) + (𝑧 − 𝑧0)𝑓'(𝑧0) + 2!
(𝑧 − 𝑧0) 𝑓''(𝑧0) +...
i. Simple proof using Cauchy’s first and general integral formulae
ii. Legitimate proof using binomial expansion
∞
𝑛
b. Laurent’s expansion for function analytic in 𝑅1 ≤ |𝑧 − 𝑧0| ≤ 𝑅2, 𝑓(𝑧) = ∑ 𝑎𝑛(𝑧 − 𝑧0)
−∞

1 (𝑛) 1 𝑓(ξ)
i. 𝑎𝑛 = 𝑛!
𝑓 (𝑧0) = 2π𝑖
∮ 𝑛+1 𝑑ξ
𝐶 (ξ−𝑧0)
ii. Proof using 𝑓(𝑧) = 𝐼𝑐 − 𝐼𝑐 , 𝐶1 is contour around point in question (non-analytic)
1

iii. Principal part: negative powers; analytic part: non-negative powers

5. Singularity: non-analyticity at the point
π −1
a. Isolated: analytic in an annulus around the point; [sin( 𝑧 )] , log(𝑧) have non-isolated
singularities at 𝑧 = 0
i. Can be removable, pole or essential
b. Removable: no negative power terms in Laurent series; function can be redefined to be analytic
at the point
∞
𝑖
c. Pole of order n ⇒ 𝑓(𝑧) = ∑ 𝑎𝑖(𝑧 − 𝑧0) ; 𝑎−𝑛 ≠ 0; lim (𝑧 − 𝑧0)𝑓(𝑧) = 𝐴(≠ 0)
−𝑛 𝑧 → 𝑧0

d. Essential: infinite negative power terms in Laurent series

e. To check: find limit
i. Exists finitely ⇒ removable singularity
ii. Limit goes to infinity ⇒ pole
iii. Else, essential singularity
f. Removable, pole, essential singularity at ∞ ⇔ removable, pole, essential singularity at 0 for
f(1/z)
g. Riemann’s theorem: isolated singularity, bounded in some deleted neighbourhood ⇒ can be
redefined to be analytic

Theorems
1. Rouche’s theorem: 𝑓, 𝑔 analytic on and inside closed domain C, |𝑔| < |𝑓| 𝑜𝑛 𝐶 ⇒ 𝑓, 𝑓 + 𝑔 have
same number of roots inside C
𝑀
2. Cauchy’s inequality: 𝑎𝑛 ≤ 𝑛 for function analytic in domain D of radius r with supremum M
𝑟

𝑓'(𝑧)
3. Argument theorem: ∮ 𝑓(𝑧)
𝑑𝑧 = 2π𝑖(𝑁 − 𝑃); 𝑁:number of roots (with multiplicities), 𝑃:number of poles
𝐶
(with multiplicities)
4. Identity theorem for analytic function: 𝑓(𝑥𝑛) = 0, < 𝑥𝑛 > → 𝑥0 ∈ 𝐷 ⇒ 𝑓(𝑥) = 0 ∀𝑥 ∈ 𝐷
𝑛 𝑛0+𝑖
a. Proof: 𝑓 (𝑥0) = 0 ∀𝑛 ∈ 𝑁; else 𝑓(𝑥) = (𝑥 − 𝑥0) 𝑔(𝑥); 𝑔(𝑥0) ≠ 0 ⇒ 𝑔(𝑥) ≠ 0 in some
neighbourhood of 𝑥0 ⇒ contradiction

Utkarsh Kumar
 5. Liouville’s theorem: 𝑓 is bounded entire function ⇒ 𝑓 is constant
𝑛 1 𝑓(𝑧) 𝑀
a. Proof: 𝑓(𝑧) = Σ𝑎 𝑧 ; 𝑎 = 2π𝑖
∮ 𝑛+1 𝑑𝑧 ≤ 𝑛 → 0 𝑎𝑠 𝑟 → ∞
𝑛 𝑛 𝑧 𝑟
b. Corollary: 𝑓 does not attain a disk of values ⇒ 𝑓 is constant
6. Gauss Lucas theorem: roots of 𝑓'(𝑧) lie within convex hull of roots of 𝑓(𝑧) (somewhat like Rolle’s
theorem)
7. Casorati Weierstrass theorem: in every neighborhood of essential singularity, 𝑓(𝑧) attains all complex
values
8. Fundamental theorem of algebra: every polynomial of degree 𝑛 has exactly 𝑛 roots
1 𝑛−1 −1 −(𝑛−𝑖)
a. On a contour of radius 𝑟 > 1, 𝑛 |𝑎0 +... + 𝑎𝑛−1𝑧 | ≤ |𝑎𝑛| (Σ|𝑎𝑖|𝑟 ) → 0 as 𝑟 → ∞;
|𝑎𝑛𝑧 |

choose large enough 𝑟 such that ratio < 1

Complex integration contours

1. : semicircle indented at 0

2. : sector of angle 45º resting on X-axis

3. : annulus keeping out +X axis and |z|=ε

4. : rectangle +-R, +-R+πi

5. : standard semi-circular contour on f(z)= log(z+i)/(z2+1)

6. : semi-circle indented at z=0

7. : semi-circle indented at z=0

8. : semi-circle indented at z = 1, -1

Utkarsh Kumar
 9. Hint: pizza slice of angle 2π/5

Miscellaneous
𝑖θ
1. log(𝑟𝑒 ) = log(𝑟) + 𝑖(θ + 2𝑛π)

Utkarsh Kumar
 1

1
Images taken from IMS notes

Utkarsh Kumar
 Partial Differential Equations

Solving PDE
1. Lagrangian equations: 𝑃𝑝 + 𝑄𝑞 = 𝑅 ⇔ 𝑑𝑥/𝑃 = 𝑑𝑦/𝑄 = 𝑑𝑧/𝑅
2. Special forms:
a. Function of 𝑝, 𝑞 only⇒ 𝑝 = 𝑎
b. Function of 𝑝, 𝑞, 𝑧 only ⇒ 𝑞 = 𝑝 × 𝑎
c. 𝑓1(𝑥, 𝑝) = 𝑓2(𝑦, 𝑞) ⇒ both equal constant 𝑎
d. Clairaut: 𝑧 = 𝑝𝑥 + 𝑞𝑦 + 𝑓(𝑝, 𝑞) ⇒ 𝑧 = 𝑝𝑎 + 𝑞𝑏 + 𝑓(𝑎, 𝑏)
𝑖 𝑗 𝑖 𝑗 𝑢 𝑣
e. Σ 𝑥 𝑦 𝐷 𝐷' 𝑧 = 𝑉 ⇒ 𝑥 = 𝑒 , 𝑦 = 𝑒 ;
𝑚 𝑛 𝑚 𝑛
(𝑥 𝑦 𝐷 𝐷' )𝑧 = 𝐷(𝐷 − 1)... (𝐷 − 𝑚 + 1)𝐷'(𝐷' − 1)... (𝐷' − 𝑛 + 1)𝑧
𝑑𝑝 𝑑𝑞 𝑑𝑥 𝑑𝑦 𝑑𝑧
3. Charpit eqn:
𝑓𝑥+ 𝑝𝑓𝑧
= 𝑓𝑦+ 𝑞𝑓𝑧
= −𝑓𝑝
= −𝑓𝑞
= −𝑝𝑓𝑝−𝑞𝑓𝑞
a. Pick two, get expressions involving at least 𝑝 or 𝑞
b. Integrate 𝑑𝑧 = 𝑝𝑑𝑥 + 𝑞𝑑𝑦 to get complete solution for 𝑧
𝑑𝑝𝑖 𝑑𝑥𝑖 𝑑𝑧
4. Jacobi’s eqn: 𝑓 =... = −𝑓 =... = −∑𝑝 𝑓 ; 𝑧 must be absent
𝑥 𝑖
𝑝 𝑖 𝑝 𝑖 𝑖

∂φ ∂ψ ∂φ ∂ψ
a. Verify for solutions 𝑢, 𝑣, 𝑤...: ∑( ∂𝑝 ∂𝑥 − ∂𝑝 ∂𝑥 ) = 0
𝑖 𝑖 𝑖𝑖 𝑖

−𝑝𝑖
b. If 𝑧 present, replace: 𝑧 → 𝑥 , 𝑝𝑖 → 𝑝𝑛+1
𝑛+1
5. Finding integral surface:
a. 𝑢(𝑥, 𝑦, 𝑧) = 𝑐1, 𝑣(𝑥, 𝑦, 𝑧) = 𝑐2, 𝑓(𝑥, 𝑦, 𝑧) = 0 ⇒ parametrise 𝑥(𝑡), 𝑦(𝑡), 𝑧(𝑡) using 𝑓; find relation
between 𝑐1, 𝑐2; replace to get surface
b. 𝑢(𝑥, 𝑦, 𝑧) = 𝑐1, 𝑣(𝑥, 𝑦, 𝑧) = 𝑐2, 𝑓(𝑥, 𝑦, 𝑧) = 0, 𝑔(𝑥, 𝑦, 𝑧) = 0 ⇒ eliminate 𝑥, 𝑦, 𝑧 to get relation
between 𝑐1, 𝑐2; replace to get surface
c. φ(𝑥, 𝑦, 𝑧, 𝑎, 𝑏) = 0, 𝑓(𝑥, 𝑦, 𝑧) = 0 ⇒ parametrise 𝑥(𝑡), 𝑦(𝑡), 𝑧(𝑡) using 𝑓 to get
6. Cauchy’s characteristics to solve 𝑓(𝑥, 𝑦, 𝑝, 𝑞, 𝑧) = 0:
a. 𝑝'(𝑡) = − 𝑓𝑥 − 𝑝𝑓𝑧, 𝑞'(𝑡) = − 𝑓𝑦 − 𝑞𝑓𝑧
b. 𝑥'(𝑡) = 𝑓𝑝, 𝑦'(𝑡) = 𝑓𝑞, 𝑧'(𝑡) = 𝑝𝑓𝑝 + 𝑞𝑓𝑞
c. Initial value in terms of λ: 𝑧' = 𝑝0𝑥' + 𝑞0𝑦', 𝑓 = 0, given integral surface etc
d. Eliminate 𝑡, λ
7. CF, PI
a. CF:
−𝑐𝑥/𝑎 −𝑐𝑦/𝑏
i. (𝑎𝐷 + 𝑏𝐷' + 𝑐)𝑧 = 0 ⇒ 𝐶𝐹 = 𝑒 φ(𝑎𝑦 − 𝑏𝑥) = 𝑒 φ(𝑎𝑦 − 𝑏𝑥)

Utkarsh Kumar
 𝑎𝑖𝑥+𝑏𝑖𝑦
ii. If irreducible to linear terms, 𝐶𝐹 = Σ 𝑐𝑖𝑒
b. PI: 1/𝑓(𝐷, 𝐷') φ(𝑥, 𝑦)
𝑛
i. 𝑓 is deg 𝑛 homog, φ = φ(𝑎𝑥 + 𝑏𝑦) ⇒ 𝑃𝐼 = 1/𝑓(𝑎, 𝑏) ∫ φ 𝑑𝑡
1 𝑎𝑥+𝑏𝑦 𝑎𝑥+𝑏𝑦 1
ii. 𝑓(𝐷,𝐷')
𝑒 φ(𝑥, 𝑦) = 𝑒 𝑓(𝐷+𝑎,𝐷'+𝑏)
φ(𝑥, 𝑦)
1 2 2 2 2
iii. 𝑓(𝐷,𝐷')
sin(𝑎𝑥 + 𝑏𝑦): replace: 𝐷 → (− 𝑎 ), 𝐷𝐷' → (− 𝑎𝑏), 𝐷' → (− 𝑏 )
1
iv. 𝑓(𝐷,𝐷')
< 𝑝𝑜𝑙𝑦𝑛𝑜𝑚𝑖𝑎𝑙 >: divide by lowest power of 𝐷(𝐷') and expand
1
v. 𝐷+𝑚𝐷'
φ(𝑥, 𝑦) = ∫ φ(𝑥, 𝑚𝑥 + 𝑐) 𝑑𝑥; 𝑦 = 𝑚𝑥 + 𝑐 ; proof using Lagrange
subsidiary equations
∂𝑢
8. General solution: replace 𝑏 → φ(𝑎), ‘eliminate’ 𝑎 from 𝑢 = 0, ∂𝑎 = 0
∂𝑢 ∂𝑢
a. Singular solution from 𝑢(𝑥, 𝑦, 𝑧, 𝑎, 𝑏) = 0: eliminate 𝑎, 𝑏 from 𝑢 = ∂𝑎
= ∂𝑏
=0
b. Obtain a complete solution φ(𝑥, 𝑦, 𝑧) = 0 from another complete solution ψ(𝑥, 𝑦, 𝑧) = 0: pick
curve 𝑓 on φ, find intersection 𝑔 of ψ, 𝑓 and find cover of g to get φ

Reduction to canonical form for 𝑅𝑟 + 𝑆𝑠 + 𝑇𝑡 = 𝑉

2
1. Solve 𝑅𝑚 + 𝑆𝑚 + 𝑇 = 0 to get λ1, λ2
2. To get characteristics: 𝑑𝑦/𝑑𝑥 + λ𝑖 = 0 ⇒ 𝑢, 𝑣
3. Independent variables: 𝑥, 𝑦 → 𝑢, 𝑣
4. Substitute in original eqn and reduce to canonical form

Method of separation of variables:

2
1. Wave eqn: 𝑢𝑡𝑡 = 𝑐 𝑢 (can be extended to 2D, 3D)
𝑥𝑥
2
2. Steady state in heat transmission: ∇ 𝑢 = 0
2
3. Heat transmission in rod: 𝑢𝑡 = 𝑐 𝑢𝑥𝑥
4. Boundary information non zero ⇒ find steady state eqn
𝑙
2
a. Cosine series ⇒ 𝑢(𝑥, 𝑡) = 𝐸0/2 + Σ𝑋𝑛(𝑥)𝑇𝑛(𝑡); 𝐸0 = 𝑙
∫ 𝑔(𝑥) 𝑑𝑥
0
𝑙
1
b. Sudden insulation ⇒ steady state, 𝑢(𝑥) = 𝑙
∫ 𝑢1(𝑥) 𝑑𝑥; 𝑢1(𝑥) being temperature at insulation
0

Proofs
1. Solutions to a linear PDE of order 1 is given by 𝑓(𝑢, 𝑣) = 0, where 𝑢, 𝑣 satisfy Lagrange subsidiary
equations for given equation

Utkarsh Kumar
 a. Form PDE for 𝑓(𝑢, 𝑣) = 0 by eliminating 𝑓: 𝑃𝑝 + 𝑄𝑞 = 𝑅
𝑑𝑥 𝑑𝑦 𝑑𝑧
b. 𝑢 = 𝑐1, 𝑣 = 𝑐2 ⇒ 𝑃
= 𝑄
= 𝑅

Miscellaneous
∂𝑓 ∂𝑓 ∂𝑓
1. Given 𝑓(𝑥, 𝑦, 𝑧) = 𝑐, to find orthogonal trajectories: ∂𝑥
𝑝+ ∂𝑦
𝑞= ∂𝑧
2. Given differential equation 𝑃𝑝 + 𝑄𝑞 = 𝑅, to find orthogonal trajectories: 𝑃𝑑𝑥 + 𝑞𝑑𝑦 + 𝑅𝑑𝑧 = 0
a. Proof: eliminate independent variable 𝑧;(𝑃𝑝 + 𝑄𝑞 = 𝑅) → (𝑃ϕ𝑥 + 𝑞ϕ𝑦 + 𝑅ϕ𝑧 = 0) →
orthogonal trajectories are given by 𝑃𝑑𝑥 + 𝑄𝑑𝑦 + 𝑅𝑑𝑧 = 0

Utkarsh Kumar
 Real Analysis
Terms with sets:
1. Interior points (neighbourhood) and interior of a set, 𝑖𝑛𝑡(𝑆)
2. Derived set, 𝐷(𝑆): set of limit points; closure of S, 𝑐𝑙𝑠(𝑆) = 𝑆 ∪ 𝐷(𝑆)
3. Dense subset, 𝑆 ⊂ ℝ: 𝑐𝑙𝑠(𝑆) = ℝ
a. Dense in itself: each member is a limit point
b. Perfect set: contains all its limit points; every member is a limit point
4. Adherent point: 𝑥 ∈ 𝑐𝑙𝑠(𝑆)(contained or is limit point)
5. Compact set, S: 𝑆 ⊆ ℝ, S is closed and bounded; alternatively, every open cover of 𝑆 has a finite open
cover
a. {𝐺α}is cover of set S iff 𝑆 ⊆ ∪ 𝐺α
b. Cover is open if it contains open sets only
6. Bolzano Weierstrass theorem: infinite bounded set (or bounded sequence) ⇒ limit point exists
7. Intersection of infinite open sets need not be open (contrapositive: union of infinite closed sets need not
be closed)

ℝ properties:
1. Field properties, well ordered
2. Completeness axiom (inf, sup contained in ℝ)
3. Archimedean property,
4. Dedekind’s completeness axiom
5. Trichotomy law
6. order complete

Convergence of sequences:
1. Cauchy’s theorems: AM, GM converge to same limit
○ Proof for both
2. Monotone convergence test
3. PMI for recursively defined sequences (MCT)
4. Sandwich theorem
𝑎𝑛
5. Ratio test: 𝑎𝑛+1
→ 𝐿 > 1 ⇒ sequence converges

6. Cauchy sequence: subsequent terms come arbitrarily close to each other

○ Cauchy sequence ⇔ convergence (proof)

Convergence of series
1. Basic Comparison Test, Limit Comparison Test (useful in conjunction with p-series)

Utkarsh Kumar
 2. Sandwich theorem
3. Necessary condition: 𝑎𝑛 → 0 (not sufficient)
𝑢𝑛
4. D'alembert’s ratio test: 𝑢𝑛+1
→ 𝐿; 𝐿 > 1 ⇒ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠; 𝐿 < 1 ⇒ 𝑑𝑖𝑣𝑒𝑟𝑔𝑒𝑠
𝑢𝑛
5. Raabe’s test (if ratio test fails): 𝑛 ( 𝑢 − 1) → 𝐿; 𝐿 > 1 ⇒ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠; 𝐿 < 1 ⇒ 𝑑𝑖𝑣𝑒𝑟𝑔𝑒𝑠
𝑛+1
𝑢𝑛 α𝑛
6. Gauss’ test (if Raabe fails): 𝑢𝑛+1
= 1 + λ/𝑛 + Σ 1+δ ; λ > 1 ⇒ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠; λ ≤ 1 ⇒ 𝑑𝑖𝑣𝑒𝑟𝑔𝑒𝑠
𝑛
𝑢𝑛
7. Logarithmic test (if ratio test fails, e present): 𝑛 log( 𝑢 ) = 𝐿; 𝐿 > 1 ⇒ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠; 𝐿 < 1 ⇒ 𝑑𝑖𝑣𝑒𝑟𝑔𝑒𝑠
𝑛+1
1/𝑛
8. Cauchy’s root test: 𝑢𝑛 → 𝐿; 𝐿 < 1 ⇒ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠; 𝐿 > 1 ⇒ 𝑑𝑖𝑣𝑒𝑟𝑔𝑒𝑠
𝑛 𝑛
9. Cauchy’s condensation test (decreasing series of positive terms): 2 𝑢𝑛(2 ) 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠 ⇔ 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠
10. Abel’s test: Σ 𝑢𝑛 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑛𝑡, {𝑣𝑛} 𝑚𝑜𝑛𝑜𝑡𝑜𝑛𝑖𝑐 𝑎𝑛𝑑 𝑏𝑜𝑢𝑛𝑑𝑒𝑑 ⇒ Σ 𝑢𝑛𝑣𝑛 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠
𝑢𝑛
11. Leibniz test for alternating series: 𝑢𝑛+1
≽1; 𝑢𝑛 → 0

12. Dirichlet test: Σ𝑛𝑢𝑛 is bounded ∀𝑛, 𝑣𝑛 is monotonic sequence: 𝑣𝑛 → 0 ⇒ Σ𝑢𝑛𝑣𝑛 is convergent
13. Cauchy’s product of two series
○ converges if both series converge, and one converges absolutely
𝑛+1 1
i. Convergent series but product doesn’t converge: 𝑎𝑛 = 𝑏𝑛 = (− 1)
𝑛
○ if product is absolutely convergent, limit is product of limits of both series
𝑛
1
○ Cesaro’s theorem: seq of nth term of product series, {𝑐𝑛 = 𝑛
∑ 𝑢𝑖𝑣 }→ 𝑢 × 𝑣
𝑖 𝑛−𝑖
1 1 1 1 1 1 1 1
14. P-series convergence proof: 𝑝 + 𝑝 ≤ 𝑝−1 , 𝑝 + 𝑝 + 𝑝 + 𝑝 ≤ 𝑝−1
2 3 2 4 5 6 7 4

Derangement of series
1. Dirichlet’s theorem: absolutely convergent series ⇒ nature and sum unchanged on derangement
2. Conditionally convergent series ⇒ nature and sum can be changed as per will
○ Can be made to converge to any real number (Riemann’s method of rearrangement)
○ Can be made to diverge to +∞ and -∞
○ Can be made to oscillate finitely or infinitely
3. Pringsheim method: Sum of deranged series (from monotonically decreasing series with alternating
1
terms): 𝑆 − 𝑆0 = 2
𝑔 log(α/β); 𝑔 = 𝑚 𝑓(𝑚), α (β) is number of positive (negative) terms taken at a
time

Infinite products
1. Limit of SOPP is 0 ⇒ infinite product diverges to 0
2. Finitely many negative terms and/or 0s ⇒ consider the terms thereafter for convergence
○ Infinitely many 0s ⇒ diverges to 0

Utkarsh Kumar
 3. (∀𝑛, 𝑎𝑛 > 0) 𝑜𝑟 (∀𝑛, − 1 < 𝑎𝑛 < 0) ⇒ ∏ (1 + 𝑎𝑛) 𝑎𝑛𝑑 ∑𝑎𝑛 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒 𝑎𝑛𝑑 𝑑𝑖𝑣𝑒𝑟𝑔𝑒 𝑡𝑜𝑔𝑒𝑡ℎ𝑒𝑟
4. Absolute convergence: ∏ (1 + |𝑎𝑛|) converges, ie ∑|𝑎𝑛| converges

Continuity of functions
1. Types of discontinuities:
a. Removable type: limit exists, 𝑓(𝑐) not defined; or 𝑓(𝑐) ≠ limit
b. Non-removable type 1: both limits exist but unequal
c. Non-removable type 2: both limits don’t exist (oscillatory)
d. Mixed non-removable: one limit type 1, other type 2
e. Infinite type: one or both limits unbounded
2. Sequential criterion for continuity: ∀ < 𝑎𝑛 > → 𝑐 ⇒ < 𝑓(𝑎𝑛) > → 𝑓(𝑐)
1 1
3. To show continuity, 𝑚𝑎𝑥(𝑓(𝑥), 𝑔(𝑥)) = 2
[𝑓(𝑥) + 𝑔(𝑥)] + 2
|𝑓(𝑥) − 𝑔(𝑥)|
4. 𝑓 continuous on (𝑎, 𝑏) ⇒ 𝑓 is bounded and attains its bounds in (𝑎, 𝑏)

Differentiability of functions
1. Rolle’s theorem: 𝑓 continuous on [𝑎, 𝑏], differentiable on (𝑎, 𝑏);
𝑓(𝑎) = 𝑓(𝑏) ⇒ 𝑓'(𝑐) = 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑐 ∈ (𝑎, 𝑏)
a. Proof: show that 𝑓 attains its maxima/minima/constant function
𝑓(𝑏)−𝑓(𝑎)
2. LMVT: 𝑓 continuous on [𝑎, 𝑏], differentiable on (𝑎, 𝑏) ⇒ 𝑓'(𝑐) = 𝑏−𝑎
𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑐 ∈ (𝑎, 𝑏)
3. Differentiability of bivariate function:
2 2
𝑓(𝑥 + ℎ, 𝑦 + 𝑘) − 𝑓(𝑥, 𝑦) = 𝐴ℎ + 𝐵𝑘 + ℎ + 𝑘 φ(ℎ, 𝑘); φ → 0 𝑎𝑠 (ℎ, 𝑘) → (0, 0)
4. Equality of cross derivatives: 𝑓𝑥𝑦 continuous, 𝑓𝑦 exists
2 2 2 2
○ Exception: 𝑓(𝑥, 𝑦) = 𝑥𝑦(𝑥 − 𝑦 )/(𝑥 + 𝑦 ) (𝑛𝑜𝑛 − 𝑜𝑟𝑖𝑔𝑖𝑛); 0 at origin

Utkarsh Kumar
 2

2
Image taken from IMS notes

Utkarsh Kumar
 Riemann Integration
1. Darboux sums:
a. Lower sum, 𝐿𝑃 = Σ 𝑚𝑟(𝑥𝑟 − 𝑥𝑟−1)
b. Upper sum, 𝑈𝑃 = Σ 𝑀𝑟(𝑥𝑟 − 𝑥𝑟−1)
2. Riemann sums: Upper sum = 𝑖𝑛𝑓 (𝑈𝑝); Lower sum = 𝑠𝑢𝑝 (𝐿𝑃)
a. Riemann integral = Σ 𝑓(ξ𝑟) (𝑥𝑟 − 𝑥𝑟−1)
3. Some sufficient but not necessary conditions for Riemann integrability in closed interval:
a. Continuous functions
b. Non-continuous functions with finite points of discontinuity
c. Non-continuous functions with finite sized derived set
d. Monotonic functions
𝑛 1
𝑛 1
4. Integration can be limiting sum of infinite series: ∑ 2 2 =∫ 2 𝑑𝑥
𝑖=1 𝑛 +𝑖 0 1+𝑥
a. Integrals can be evaluated by differentiating wrt parameter
5. Mean theorems for integrals:
a. First MVT: 𝑔(𝑥) ≥ 0 ∀𝑥 ∈ [𝑎, 𝑏] ⇒ ∫ 𝑓𝑔 𝑑𝑥 = µ∫ 𝑔 𝑑𝑥; 𝑚𝑓 ≤ µ ≤ 𝑀𝑓
b. Second MVT:
𝑏 ξ 𝑏
𝑓(𝑥) 𝑏𝑜𝑢𝑛𝑑𝑒𝑑 𝑎𝑛𝑑 𝑚𝑜𝑛𝑜𝑡𝑜𝑛𝑖𝑐 ⇒ ∫ 𝑓𝑔 𝑑𝑥 = 𝑓(𝑎) ∫ 𝑔 𝑑𝑥 + 𝑓(𝑏) ∫ 𝑔 𝑑𝑥 𝑓𝑜𝑟 ξ ∈ [𝑎, 𝑏]
𝑎 𝑎 ξ
𝑏 ξ
c. Bonnet’s mean value theorem: 𝑓 𝑚𝑜𝑛𝑜𝑡𝑜𝑛𝑖𝑐𝑎𝑙𝑙𝑦 𝑑𝑒𝑐𝑟𝑒𝑎𝑠𝑖𝑛𝑔 ⇒ ∫ 𝑓(𝑥)𝑔(𝑥)𝑑𝑥 = 𝑓(𝑎) ∫ 𝑔(𝑥)𝑑𝑥
𝑎 𝑎
𝑏 𝑏
i. 𝑓 𝑚𝑜𝑛𝑜𝑡𝑜𝑛𝑖𝑐𝑎𝑙𝑙𝑦 𝑖𝑛𝑐𝑟𝑒𝑎𝑠𝑖𝑛𝑔 ⇒ ∫ 𝑓(𝑥)𝑔(𝑥)𝑑𝑥 = 𝑓(𝑏) ∫ 𝑔(𝑥)𝑑𝑥 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 ξ ∈ [𝑎, 𝑏]
𝑎 ξ
1 1
𝑛
6. Integrals that converge: ∫ 𝑥 𝑑𝑥 𝑓𝑜𝑟 𝑛 > − 1; ∫ log(𝑥) 𝑑𝑥
0 0
7. Other tests: limit comparison and basic comparison tests
a. Abel’s test: ∫ 𝑓𝑔 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠 𝑖𝑓 ∫ 𝑓 𝑑𝑜𝑒𝑠, 𝑔 is monotonic, bounded
∞ 𝑡
b. Dirichlet’s test: ∫ 𝑓𝑔 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠 𝑖𝑓 ∫ 𝑓 𝑖𝑠 𝑏𝑜𝑢𝑛𝑑𝑒𝑑 ∀𝑡 ≥ 𝑎, 𝑔 is bounded monotonic, lim 𝑔 = 0
𝑎 𝑎 𝑥→∞

Utkarsh Kumar
 Modern Algebra

Definitions
1. Groups: closure, associativity, identity, inverses
2. Semi-group: closure, associativity
3. Monoid: semi-group with identity
4. Binary operator: closure
5. Ring: Commutative group for addition, semi-group for multiplication, distributivity
○ Characteristic of ring: 𝑚𝑖𝑛 𝑛 ∈ ℕ: 𝑛𝑥 = 0 ∀𝑥 ∈ 𝑅; 0 if no such 𝑛 exists
○ Prime element: 𝑝 | 𝑎𝑏 ⇒ 𝑝 | 𝑎 𝑜𝑟 𝑝 | 𝑏; irreducible: 𝑝 = 𝑎𝑏 ⇒ 𝑎 𝑖𝑠 𝑢𝑛𝑖𝑡, 𝑜𝑟 𝑏 𝑖𝑠 𝑢𝑛𝑖𝑡
6. Integral domain: commutativity, unity, no zero divisors
7. Field: integral domain, all non-zero elements have inverses
8. Ideal: ∀𝑥 ∈ 𝑅, ∀𝑎 ∈ 𝐼, 𝑎𝑥 ∈ 𝐼, 𝑥𝑎 ∈ 𝐼; 𝐼 is subring of 𝑅
○ < 𝑎 > = {𝑎𝑥 | 𝑥 ∈ 𝑅} is smallest ideal containing 𝑎 (is ideal if 𝑅 is commutative)
○ Product of ideals, 𝐻𝐾 = {∑ℎ𝑖𝑘𝑖} is an ideal
○ Unlike groups, 𝐻𝐾 ⊆ 𝐻⋂𝐾
9. Prime ideal: ℎ𝑘 ∈ 𝐼 ⇒ ℎ ∈ 𝐼 𝑜𝑟 𝑘 ∈ 𝐼; maximal ideal: 𝑀 ⊂ 𝐼 ⊂ 𝑅 ⇒ 𝑀 = 𝐼 𝑜𝑟 𝐼 = 𝑅
2
○ Semi-prime ideal: 𝑎 ∈ 𝐼 ⇒ 𝑎 ∈ 𝐼
10. Euclidean domain: ID; ∃𝑑: 𝑅 → ℝ, 𝑑(𝑥) ≥ 0 ∀𝑥 ≠ 0, 𝑑(𝑥𝑦) ≥ 𝑑(𝑥); division algorithm
○ gcd(𝑚, 𝑛) = 𝑑 ⇒ 𝑑'|𝑑 ∀𝑑𝑖𝑣𝑖𝑠𝑜𝑟𝑠 𝑑'
11. Principal ideal domain: integral domain, all ideals are principal ideals
12. UFD: integral domain, ∀𝑥 ∈ 𝑅, 𝑥 = ∏ 𝑝𝑖, factorisation unique to associates, 𝑝𝑖'𝑠 being irreducible
𝑖 𝑖 𝑖 −𝑖
13. Dihedral group, 𝐷𝑛 = {𝑟 , 𝑠𝑟 }; 𝑠𝑟 = 𝑟 𝑠; 𝑢𝑠𝑒𝑓𝑢𝑙 𝑤ℎ𝑒𝑛 |𝑎| = 2, |𝑏| = 11, |𝑎𝑏| = 11
14. F is algebraically closed ⇒ any polynomial on F has all roots within F
○ Thus R, Q are not algebraically closed fields
○ F cannot be finite: 𝑝(𝑥) = 1 + Π(𝑥 − 𝑓𝑖)
15. F is finite field extension of E ⇒ F is a FDVS over E
○ α ∈ 𝐹 is algebraic over E ⇒ ∃𝑝(𝑥) ∈ 𝐸[𝑥]: 𝑝(α) = 0
16. Normalizer of group: 𝑔𝐻 = 𝐻𝑔 ∀𝑔 ∈ 𝐺; centralizer of group: 𝑔ℎ = ℎ𝑔 ∀ℎ ∈ 𝐻, 𝑔 ∈ 𝐺

Theorems
1. Lagrange’s theorem: 𝐻 ≤ 𝐺 ⇒ 𝑂(𝐻) | 𝑂(𝐺) (proof using cosets)
○ Cauchy’s theorem for prime 𝑝: 𝑝 | 𝑂(𝐺) ⇒ ∃𝑥 ∈ 𝐺: 𝑂(𝑥) = 𝑝; proof: consider p-tuples
𝑝−1
(𝑎1,.., 𝑎𝑝): Π𝑎𝑖 = 𝑒; no elements of order 𝑝 ⇒ 𝑝| 𝑛 −1
−1
2. Cosets: 𝑎𝐻 = 𝑏𝐻 ⇔ 𝑏 𝑎 ∈ 𝐻; 𝑎𝐻 = 𝑏𝐻 𝑜𝑟 𝑎𝐻∩𝑏𝐻 = φ
𝑂(𝐻)𝑂(𝐾)
3. 𝑂(𝐻𝐾) = 𝑂(𝐻∩𝐾)
(to disprove existence of multiple subgroups of given order)

Utkarsh Kumar
 4. Prime order ⇒ cyclic group
5. Homomorphism theorems:
𝐻
○ Homomorphism φ: 𝐺 → 𝐻 ⇒ 𝐺 ≌ 𝑘𝑒𝑟 φ
𝐺 𝐺/𝐾
○ 𝐻
≌ 𝐻/𝐾

𝑂(𝐺)
○ Class equation: 𝑂(𝐺) = ∑ 𝑂(𝑁(𝑎))
; sum runs over one element per conjugacy class, (
𝑎∉𝑍(𝐺)
−1
𝑥 ∼ 𝑦 ⇔ 𝑦 = 𝑎 𝑥𝑎); proof: 𝐶(𝑎) ≈ 𝐺/𝑁(𝑎); {𝐶(𝑎𝑖)} is partition of 𝐺
𝑂(𝐺)
○ Class equation can also be written as: 𝑂(𝐺) = 𝑂(𝑍) + Σ 𝑂(𝑁(𝑎))
(since 𝑁(𝑧) = 𝑂(𝐺), and each
element of 𝑍(𝐺) is in its own conjugacy class)
6. finite ID ⇒ field
7. Characteristic of field is 0 or prime
𝑛
8. Every finite field has order 𝑝 for prime 𝑝, 𝑛 ∈ ℕ
○ Proof: 𝑐ℎ𝑎𝑟 𝐹 = 𝑝; 𝑞 | 𝑂(𝐹) ⇒ ∃𝑥: 𝑂(𝑥) = 𝑞 ⇒ 𝑞. 1 = 0 ⇒ 𝑝 | 𝑞
𝑅 𝑅
9. 𝐼 is prime ideal in CR ⇔ 𝐼 is ID; 𝐼 is maximal ideal in CRU ⇔ 𝐼 is field
○ 𝑎 is prime ⇔ < 𝑎 > is prime ideal; 𝑎 is irreducible in PID ⇔ < 𝑎 > is maximal ideal
○ prime ideals need not be maximal; maximal ideals are prime in CRU (not in CR: 4ℤ ⊂ 2ℤ)
10. In ID, primes are irreducibles; in PID and UFD, irreducibles are prime
○ PID, UFD: prime element ⇔ irreducible element
11. ED ⇒ PID: proof: division algorithm
12. PID ⇒ UFD: proof: Π𝑝𝑖 = Π𝑞𝑖 ⇒ 𝑛 = 𝑚; 𝑝𝑖|𝑞𝑖
○ Alternate proof using ACC lemma for PID
○ ACC lemma: given ideal 𝐼, there cannot be infinite chain of ideals: 𝐼1 ⊂ 𝐼2 ⊂ 𝐼3... (take union as 𝐼,
which is an ideal ⇒ 𝐼 = < 𝑐 > ⇒ 𝐼𝑛 = < 𝑐 >)
○ For 𝑎 ∈ 𝑅, decompose <a> into irreducibles satisfying ACC
13. Ring embedding:
○ Ring R can be embedded in ring with unity:
𝑓: 𝑅 → 𝑅 × ℤ, 𝑓(𝑥) = (𝑥, 0), (𝑎, 𝑏) + (𝑐, 𝑑) = (𝑎 + 𝑐, 𝑏 + 𝑑), (𝑎, 𝑏). (𝑐, 𝑑) = (𝑎𝑐 + 𝑏𝑐 + 𝑎𝑑, 𝑏𝑑),
○ ID 𝑅 can be embedded in field 𝐹of fractions of 𝑅:
(𝑎, 𝑏)〜(𝑐, 𝑑) ⇔ 𝑎𝑑 = 𝑏𝑐; (𝑎, 𝑏) + (𝑐, 𝑑) = (𝑎𝑑 + 𝑏𝑐, 𝑏𝑑), (𝑎, 𝑏). (𝑐, 𝑑) = (𝑎𝑐, 𝑏𝑑)
14. For two fields, G is algebraic over F ⇔ G is a finite field extension of F
15. Sylow: |𝐺| = 𝑝𝑞 (𝑝 > 𝑞); 𝑞∤(𝑝 − 1) ⇒ 𝐺 is cyclic

Polynomials
1. 𝐹 𝑖𝑠 𝐼𝐷 ⇒ 𝐹[𝑥] 𝑖𝑠 𝐼𝐷, 𝑑𝑒𝑔(𝑓𝑔) = 𝑑𝑒𝑔(𝑓) + 𝑑𝑒𝑔(𝑔); 𝐹 𝑖𝑠 𝑓𝑖𝑒𝑙𝑑 ⇒ 𝐹[𝑥] is ED
2
2. Eisenstein’s criterion for irreducibility: ∃ 𝑝𝑟𝑖𝑚𝑒 𝑝: 𝑝 | 𝑎𝑖 𝑓𝑜𝑟 𝑖 = 0, 1,.., 𝑛 − 1, 𝑝∤𝑎𝑛, 𝑝 ∤𝑎0
𝑅 𝑅[𝑥]
3. 𝐼
[𝑥] ≌ 𝐼[𝑥]
4. Primitive polynomial: ideal generated by coefficients equals ring

Utkarsh Kumar
Counter-examples
1. All proper subgroups Abelian ⇏G Abelian: quaternion group
2. All subgroups cyclic ⇏G cyclic: quaternion group, Klein-4/D4
3. Converse of Lagrange’s theorem not true: A4 has no subgroup of order 6
2 2 2
4. Ring with unity, (𝑥𝑦) = 𝑥 𝑦 ∀𝑥, 𝑦 ∈ 𝑅 ⇒ Abelian; doesn’t hold without unity
○ 𝑅 = { 𝑚𝑎𝑡(𝑎, 𝑏; 0, 0) }
5. Two non-normal subgroups with product being sub-group: 𝐴4, < (1, 2, 3, 4, 5) > = 𝐴5 ▹ 𝑆5

6. Irreducible, but not prime:(1 + 𝑖 5) ∈ ℤ[𝑖]

7. Normal subgroup is not transitive relation: A4
2 2
8. 𝐻 = {𝑔 ∀𝑔 ∈ 𝐺} need not be a subgroup: (1 2 3)(2 3 4) ∉ 𝐴4

Utkarsh Kumar
 Numerical Analysis

Roots
1. Convergence criterion: | 𝑓'(𝑥)| ≤ 1
2. Bisection method: 𝑝 = 1
3. Secant method: 𝑝 = ( 5 + 1)/2; Regula falsi (false position)
a. Regula Falsi: root bracketing; no bracketing in secant method
𝑓''(α)
b. Prove convergence order: 𝑥𝑘 = α + ϵ𝑘; expand using Taylor to get ϵ𝑘+1 = ϵ𝑘ϵ𝑘−1 2𝑓'(α)
; use
𝑝
ϵ𝑘 = λϵ𝑘−1
𝑓''(α) 𝑝
c. Asymptotic constant obtained using ϵ𝑘+1 = ϵ𝑘ϵ𝑘−1 2𝑓'(α)
, ϵ𝑘 = λϵ𝑘−1
4. Newton Raphson: 𝑝 = 2; multiply last term by n in case of nth order root
𝑓''(α) 2
a. ϵ𝑘 = 2𝑓'(α)
ϵ𝑘−1

Solution to system of eqns

1. Gaussian elimination: upper triangulation; back substitution
* *
2. Gauss Jordan: [𝐴|𝐵] → [𝐴 |𝐵']; 𝐴 is reduced row echelon form: zero rows at bottom, unit pivots,
columns with pivot zeroed out (except pivot)
−1
a. primarily used to find inverse of matrix, [𝐼|𝐴]→[𝐴 |𝐼]
b. Gauss Jacobi: pivot if small pivot elements
𝑘+1 𝑘
i. 𝑋 = 𝐻𝑋 + 𝐶
ii. Partial pivoting: pivoting before every column elimination
c. Gauss Seidel: pivot if small pivot elements

Inter/extra-polation
𝑖
1. Lagrange polynomial: derive by finding 𝑎𝑖: 𝑓(𝑥) = Σ𝑎𝑖𝑥 ; 𝑓(𝑥𝑖) = 𝑦𝑖; n equations, n variables ⇒unique
polynomial obtained
(𝑛+1)
a. Error = 𝑓 (ξ)(𝑥 − 𝑥0)... (𝑥 − 𝑥𝑛)/(𝑛 + 1)!
b. Partial fraction decomposition using Lagrange interpolation on numerator
2. Newton’s divided differences:
ℎ𝐷
a. Operators, ∆, ∇, δ, µ, 𝐸 = 𝑒
𝑛
b. Polynomial of degree 𝑛 ⇒ ∆ constant

Utkarsh Kumar
 𝑛
∆ 𝑓(𝑥0) 𝑓
(𝑛+1)
(ξ)
c. Error = 𝑛+1 Π(𝑥 − 𝑥𝑖) = Π(𝑥 − 𝑥𝑖)
ℎ (𝑛+1)! (𝑛+1)!
Π(𝑡−𝑥𝑖)
i. Proof: consider 𝐹(𝑡) = 𝑓(𝑡) − 𝑃(𝑡) + (𝑓(𝑥) − 𝑃(𝑥)) Π(𝑥−𝑥 ) ;
𝑖
𝑛+1
𝐹 (η) = 0, η ∈ (𝑥0, 𝑥𝑛)
(𝑥−𝑥𝑖)(𝑥−𝑥𝑖+1) 2
ii. Linear interpolation: error = 2!
𝑓''(𝑥) ≤ (𝑥𝑖−1 − 𝑥𝑖)𝑓''/8 ≤ ℎ 𝑓''(𝑥)/8 (to
choose ℎ)

Integration
1. Newton Cotes quadrature
ℎ 3
a. Trapezoid rule: 2
(𝑓(𝑥0) + 𝑓(𝑥0 + ℎ)); error: ℎ 𝑓"(ξ)/12
ℎ 5 𝑖𝑣
b. Simpson’s ⅓ rule: 3
(𝑓(𝑥0) + 4𝑓(𝑥0 + ℎ) + 𝑓(𝑥0 + 2ℎ)); error: ℎ 𝑓 (ξ)/90

3 5 𝑖𝑣
3ℎ
c. Simpson’s ⅜ rule: (𝑓(𝑥0) + 3𝑓(𝑥0 + ℎ) + 3𝑓(𝑥0 + 2ℎ) + 𝑓(𝑥0 + 3ℎ)); error: 80 ℎ 𝑓 (ξ)
8

2. Gauss quadrature (Legendre polynomial roots for integral on interval (− 1, 1))

a. Order 1: 𝐼 = 2𝑓(0)
b. Order 2: 𝐼 = 𝑓(− 1/ 3) + 𝑓(1/ 3)
c. Order 3: 𝐼 = 5/9 𝑓(− 3/5) + 8/9 𝑓(0) + 5/9 𝑓( 3/5)

ODE solution for 𝑑𝑦/𝑑𝑥 = 𝑓(𝑥, 𝑦)

1. RK methods: Taylor’s series with 𝑛 terms (error (𝑛 + 1)th term)
a. RK 1 (Euler’s method): 𝑦(𝑥0 + ℎ) = 𝑦0 + ℎ𝑓(𝑥0, 𝑦0)
b. RK 2 (1 iteration of modified Euler’s): 𝑦(𝑥0 + ℎ) = 𝑦0 + ℎ(𝑘1 + 𝑘2)/2
i. 𝑘1 = 𝑓(𝑥0, 𝑦0)
ii. 𝑘2 = 𝑓(𝑥0 + ℎ, 𝑦0 + ℎ𝑘1)
(0)
iii. Modified euler: 𝑦 (𝑥0 + ℎ) = 𝑦0 + ℎ𝑓(𝑥0, 𝑦0);
(𝑛) ℎ (𝑛−1)
𝑦 (𝑥0 + ℎ) = 𝑦0 + 2
[𝑓(𝑥0, 𝑦0) + 𝑓(𝑥0 + ℎ, 𝑦 )
c. RK 3: 𝑦(𝑥0 + ℎ) = 𝑦0 + (𝑘1 + 4𝑘2 + 𝑘3)/6
i. 𝑘1 = ℎ𝑓(𝑥0, 𝑦0)
ii. 𝑘2 = ℎ𝑓(𝑥0 + ℎ/2, 𝑦0 + 𝑘1/2)
iii. 𝑘3 = ℎ𝑓(𝑥0 + ℎ, 𝑦0 + 𝑘'), where 𝑘' = ℎ𝑓(𝑥0 + ℎ, 𝑦0 + 𝑘2)
d. RK 4: 𝑦(𝑥0 + ℎ) = 𝑦0 + (𝑘1 + 2𝑘2 + 2𝑘3 + 𝑘4)/6
i. 𝑘1 = ℎ𝑓(𝑥0, 𝑦0)

Utkarsh Kumar
ii. 𝑘2 = ℎ𝑓(𝑥0 + ℎ/2, 𝑦0 + 𝑘1/2)
iii. 𝑘3 = ℎ𝑓(𝑥0 + ℎ/2, 𝑦0 + 𝑘2/2)
iv. 𝑘4 = ℎ𝑓(𝑥0 + ℎ, 𝑦0 + 𝑘3)

Utkarsh Kumar
 Linear Programming
1. Graphical method (simplest)
○ Corner method
○ Iso-profit method
2. Simplex methods
○ Write in standard form
i. Maximisation problem
ii. Slack, surplus, artificial variables as per constraint (RHS non-negative)
iii. Artificial variables introduced for =, ≥ constraints
○ Dealing with artificial variables:
i. big-M method (Charne’s penalty): add high 𝑀 cost to artificial variables
*
ii. dual phase method: maximise 𝑧 = Σ(− 𝐴𝑖) in phase I
*
1. 𝑧 𝑚𝑎𝑥
< 0 with non-zero 𝐴𝑖 in basis ⇒ no feasible solution space
*
2. 𝑧 𝑚𝑎𝑥
= 0 with 𝐴𝑖 not in basis ⇒ proceed with non-artificial variables for phase II
*
3. 𝑧 𝑚𝑎𝑥
= 0 with 𝐴𝑖 in basis at zero level ⇒proceed with non-artificial variables and
basic artificial variables for phase II
4. Phase II: proceed with basic solution and tableau obtained in phase I, switching
to original optimisation function 𝑧
○ Basic feasible solution: only slack and artificial variable possibly non-zero
○ Construct simplex tableau
○ Optimality check: compute 𝑞𝑗 = 𝑧𝑗 − 𝑐𝑗; 𝑞𝑗 ≥ 0 ∀𝑖 ⇒ optimality
○ Iteration:
i. Choose variable with most negative 𝑞𝑗 to enter basis, say 𝑥𝑗
ii. Compute 𝑏𝑖/𝑥𝑖 for each row, choosing row with least non-negative ratio, to kick out from
basis
iii. Adjust the tableau with elementary row operations to make 𝐶𝑗 = δ𝑗𝑘
○ Optimality with artificial variable in basis with non-zero value ⇒ no feasible solution space
○ Non-optimality but no variable willing to enter basis (𝑞𝑗≤ 0 ∀𝑗) ⇒ unbounded solution
○ Duality: optimal value of variable in primal = − 𝐶𝑗 for corresponding slack/artificial variable (and
vice-versa)
i. Unbounded solution ⇒dual has infeasible solution space
○ Degeneracy: θ𝑖 = θ𝑗 = 0 ⇒ degeneracy resolution
i. Bland’s method: choose variable with least index to leave basis, and least index to enter
basis
○ Dual simplex method: begin by changing to maximisation problem subject to ≤ constraints
i. 𝐶𝑗 ≥ 0, 𝑏𝑗 ≥ 0 ⇒ optimality

Utkarsh Kumar
 ii. 𝐶𝑗 ≥ 0, 𝑠𝑜𝑚𝑒 𝑏𝑗 < 0 ⇒ non-optimal; iterate
iii. 𝐶𝑗 < 0 method fails
iv. Iteration: choose row with most negative 𝑏𝑗 to exit basis; and choose element with least
non-negative ratio to enter
3. Transportation problem: Vogel’s approximation, iteration
4. Assignment problem: Hungarian method
5. Feasible solution → basic feasible solution: find λ𝑖'𝑠 such that columns vanish
𝑥 λ𝑟
○ Choose 𝑘 = 𝑎𝑟𝑔𝑚𝑖𝑛𝑖 { λ𝑖 | λ𝑖 > 0}; 𝑥𝑟 = 𝑥𝑟 − λ𝑘
𝑥𝑘
𝑖

Utkarsh Kumar
 Fluid Dynamics
1. Continuity (possible flow for incompressible fluid): ∂𝑢/∂𝑥 + ∂𝑣/∂𝑦 + ∂𝑤/∂𝑧 = 0
→
∂𝑞
2. Steady state: ∂𝑡 =0
→ → →
3. Irrotational flow: ∇ × 𝑞 = 0; 𝑞 = 𝑢 𝑖 + 𝑣 𝑗 + 𝑤 𝑘
→
4. Velocity potential: ∇𝑉 = − 𝑞; Stream function: conjugate of velocity potential in 2D
→ →
5. Vorticity: Ω = ∇ × 𝑞
6. Lines:
→ → →
a. Streamlines: 𝑑𝑥/𝑢 = 𝑑𝑦/𝑣 = 𝑑𝑧/𝑤 (tangents represent 𝑞, ie 𝑞 × 𝑑𝑟 = 0)
b. Streak lines: put (𝑥0, 𝑦0, 𝑧0) in path lines to get constants; put back in pathline with 𝑡 = 𝑇 to get
streaklines for 𝑡 < 𝑇 (position at time 𝑇 for particles that were once at (𝑥0, 𝑦0, 𝑧0))
c. Pathlines: 𝑥'(𝑡) = 𝑢, 𝑦'(𝑡) = 𝑣, 𝑧'(𝑡) = 𝑤 (path of a given particle over time)
→
d. Vortex lines: 𝑑𝑥/𝑝1 = 𝑑𝑦/𝑝2 = 𝑑𝑧/𝑝3; (𝑝1, 𝑝2, 𝑝3) = Ω (tangents represent Ω, ie Ω × 𝑑𝑟 = 0)
e. Equipotential lines: ψ = 𝑐 (perpendicular to streamlines)
→ ^ → →→ →
7. Acceleration = 𝑑𝑞/𝑑𝑡 = Σ(∂𝑢/∂𝑡 + 𝑢∂𝑢/∂𝑥 + 𝑣∂𝑢/∂𝑦 + 𝑤∂𝑢/∂𝑧) 𝑖 = ∂𝑞/∂𝑡 + (𝑞. ∇) 𝑞
8. Circulation of flow 𝑓 round curve 𝐶: ∮𝐶 𝑓 𝑑𝑠
9. Boundary surface condition (for 𝐹 = 0): ∂𝐹/∂𝑡 + Σ𝑢∂𝐹/∂𝑥 = 0
10. Lagrangian specs: observe motion of a particular particle over time 𝑥 = 𝑥(𝑥0, 𝑦0, 𝑧0, 𝑡)
→ →
11. Eulerian specs: observe particles passing through a particular point over time: 𝑞 = 𝑞(𝑥, 𝑦, 𝑧, 𝑡)

Utkarsh Kumar
 Statics
1. Common catenary
a. 𝑠 = 𝑐 𝑡𝑎𝑛 ψ
2 2 2
b. 𝑦 = 𝑠 + 𝑐 ; 𝑦 = 𝑐 cosh(𝑥/𝑐) = 𝑐 𝑠𝑒𝑐 ψ
c. 𝑇 = λ𝑦; 2𝑇 sin φ = λ𝑙
2. Virtual work principle:
2 2 2 2
a. Apollonius theorem: 𝐴𝐵 + 𝐴𝐶 = 2(𝐴𝑀 + 𝐵𝑀 ); M being midpoint of BC

Utkarsh Kumar
 Appendix
𝑥 2
2 −𝑡
1. Error function, 𝑒𝑟𝑓(𝑥) = ∫𝑒 𝑑𝑡
π
0
𝑛
2. Homogeneous function, 𝑓(𝑥, 𝑦) can be written as 𝑥 𝑔(𝑦/𝑥); 𝑑𝑒𝑔 𝑓 = 𝑛
∞
𝑥−1 −𝑡 1
3. Gamma function, Γ(𝑥) = ∫ 𝑡 𝑒 𝑑𝑡 𝑓𝑜𝑟 𝑥 > 0; Γ(𝑛) = (𝑛 − 1)! 𝑖𝑓 𝑥 ∈ ℕ; Γ( 2 ) = π
0
a. Γ(𝑥) = (𝑥 − 1)Γ(𝑥 − 1)
1 π/2
𝑚−1 𝑛−1 2𝑚−1 2𝑛−1 Γ(𝑚)Γ(𝑛)
4. Beta function, β(𝑚, 𝑛) = ∫ 𝑥 (1 − 𝑥) 𝑑𝑥 = 2 ∫ 𝑠𝑖𝑛 𝑥 𝑐𝑜𝑠 𝑥 𝑑𝑥 = Γ(𝑚+𝑛)
𝑓𝑜𝑟 𝑚, 𝑛 > 0
0 0
−𝑧 𝑛−1
a. Proof: consider Γ(𝑚): 𝑥 → 𝑎𝑥, 𝑎 → 𝑧, multiply both sides by 𝑒 𝑧 and integrate both sides
5. To find surfaces/curves (ODE/PDE solution) passing through a given curve,
a. Eliminate constants in ODE, PDE by substituting given curve
b. Parametrise with given integral, ensure equal roots in parameter (discriminant vanishes)
c. Do (b) as many times as needed to eliminate all arbitrary constants

Coordinate systems
1. Cartesian: ℎ1 = ℎ2 = ℎ3 = 1; spherical: ℎ1 = 1, ℎ2 = 𝑟, ℎ3 = 𝑟 sin θ; cylindrical: ℎ1 = 1, ℎ2 = 𝑟, ℎ3 = 1
1 ∂ψ ^
2. Gradient: ∇ψ = Σ 𝑒
ℎ1 ∂𝑢1 1
→ → 1 ∂ψ1ℎ2ℎ3 ^
3. Divergence: ∇. ψ = Σ( ) 𝑒1
ℎ1ℎ2ℎ3 ∂𝑢1
→ → 1 ^ ∂
4. Curl: ∇ ×ψ= ∆(ℎ1𝑒1, , ψ1ℎ1)
ℎ1ℎ2ℎ3 ∂𝑢1
ℎ1𝑑𝑢1 ℎ2𝑑𝑢2 ℎ3𝑑𝑢3
5. Continuity equation:
𝑞𝑢
= 𝑞𝑢
= 𝑞𝑢
1 2 3

2
6. Arc length: 𝑑𝑠 = Σ(ℎ1𝑑𝑢1)
2 1 ∂ ℎ2ℎ3 ∂ψ1
7. Laplace equation: ∇ ψ= Σ ( )
ℎ1ℎ2ℎ3 ∂𝑢1 ℎ1 ∂𝑢1
2
1 ∂ ∂ 1 ∂
a. Polar: 𝑟 ∂𝑟
(𝑟 ∂𝑟
)+ 2 2
𝑟 ∂θ
2 2
1 ∂ ∂ 1 ∂ ∂
b. Cylindrical: 𝑟 ∂𝑟
(𝑟 ∂𝑟
)+ 2 2 + 2
𝑟 ∂θ ∂𝑧
2
1 ∂ 2 ∂ 1
∂ ∂ 1 ∂
c. Spherical: 2 ∂𝑟
(𝑟 ∂𝑟
)+ 2 (sin θ )+ 2 2 2
𝑟 𝑟 sinθ ∂θ ∂θ 𝑟 sin θ ∂φ

Utkarsh Kumar
 Asymptotes
1. Divide by highest power of 𝑥 (or 𝑦) and let 𝑥 → ∞
2. Put 𝑦 = 𝑚𝑥 + 𝑐 and choose 𝑚, 𝑐 such that higher degree terms vanish

Important expansions/series
𝑥 2
1. 𝑒 = 1 + 𝑥/1! + 𝑥 /2! +...
2 3
2. log(1 + 𝑥) = 𝑥 − 𝑥 /2 + 𝑥 /3 +...
3 5 2 4
3. sin(𝑥) = 𝑥 − 𝑥 /3! + 𝑥 /5! +...; cos 𝑥 = 1 − 𝑥 /2! + 𝑥 /4! +...
2 3 2
4. ∑𝑛 = 𝑛(𝑛 + 1)/2; ∑𝑛 = 𝑛(𝑛 + 1)(2𝑛 + 1)/6; ∑𝑛 = [𝑛(𝑛 + 1)/2]
2 2 𝑛 2 2
5. ∑1/𝑛 = π /6; ∑(− 1) /𝑛 = π /12

Utkarsh Kumar
Computer Programming
1. Flowcharts
2. Storage of integers, floating integers:
a. Double: 2 16-bit registers
b. Float: scientific notation (1 bit sign, 24 bit mantissa, 7 bit exponent)
c. Negative numbers: MSB sign bit, 1’s complement (flip all bits), 2’s complement (flip, add 1)

Utkarsh Kumar
Physics appendix

Common MIs:
2
1. Solid sphere: 2/5 𝑀𝑎
2
2. Hollow sphere: 2/3 𝑀𝑎
2
3. Disk: 1/2𝑀𝑎
2 2
4. Centre of rod: 1/12 𝑀𝑎 ; end of rod: 1/3 𝑀𝑎
5. Parallel axis theorem; perpendicular axis theorem

Common CoMs:
1. Hemisphere: shell: 𝑎/2; solid: 3𝑎/8
2. Semi-circle: ring: 2𝑟/π; lamina: 4𝑟/3π
3. Right circular cone: shell: ℎ/3; solid: ℎ/4

Common volumes:
1. Ellipsoid: 4/3π𝑎𝑏𝑐
2. Pyramid: 1/3 𝐴ℎ

Utkarsh Kumar