Linear Algebra

1. D’able ⇔ AM=GM for all eigenvalues

Vectorspaces
1. Non-empty set 𝑉over field 𝐹 is vectorspace if:
○ Abelian group under +: closure, associativity, identity, inverse, commutativity
○ Scalar multiplication distributivity:
i. 𝑎(α + β) = 𝑎α + 𝑎β ∀𝑎 ∈ 𝐹, α, β ∈ 𝑉
ii. (𝑎 + 𝑏)(α) = 𝑎α + 𝑏α ∀𝑎, 𝑏 ∈ 𝐹, α ∈ 𝑉
iii. 𝑎𝑏(α) = 𝑎(𝑏α) ∀𝑎, 𝑏 ∈ 𝐹, α ∈ 𝑉
iv. 1. α = α ∀α ∈ 𝑉, 1 𝑖𝑠 𝑢𝑛𝑖𝑡𝑦 𝑜𝑓 𝐹(important!)
○ 𝑉is not a vectorspace if 𝐹 is not a field; F needs to be subfield of V for V to be vectorspace
2. Subspace test: 𝑊 ⊆ 𝑉(𝐹) 𝑖𝑠 𝑠𝑢𝑏𝑠𝑝𝑎𝑐𝑒 𝑖𝑓𝑓 𝑊≠ φ, 𝑎α − 𝑏β ∈ 𝑊 ∀α, β ∈ 𝑊, 𝑎 ∈ 𝐹
3. Theorems:
○ Intersection of arbitrary number of subspaces is a subspace
○ Union is subset iff one is contained in the other
4. dim(𝑉 + 𝑊) = dim(𝑉) + dim(𝑊) − dim(𝑉∩𝑊)
5. Vector homomorphism/LT: φ: 𝑉 → 𝑊 𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡 φ(𝑎α + 𝑏β) = 𝑎φ(α) + 𝑏φ(β)
𝑊
○ First principle: φ: 𝑉 → 𝑊 ⇒ 𝑉 ≈ 𝑘𝑒𝑟 φ
○ Second principle
○ Third principle
○ Sylvester’s theorem: 𝑟𝑎𝑛𝑘(𝑇) + 𝑛𝑢𝑙𝑙(𝑇) = dim 𝑉 (proof by extending basis of 𝑛𝑢𝑙𝑙(𝑇))
○ Linear transform on a vectorspace: non-singularity ⇔ invertibility
6. If 𝐿(𝑉, 𝑊) 𝑖𝑠 𝑣𝑒𝑐𝑡𝑜𝑟𝑠𝑝𝑎𝑐𝑒 𝑜𝑓 𝑎𝑙𝑙 𝐿𝑇𝑠 𝑓𝑟𝑜𝑚 𝑉 𝑡𝑜 𝑊, dim 𝐿(𝑉, 𝑊) = dim 𝑉. dim 𝑊
○ Proof: 𝑏𝑎𝑠𝑖𝑠 = {𝑇𝑖𝑗| 𝑇𝑖𝑗(𝑣𝑖) = 𝑤𝑗, 0 𝑒𝑙𝑠𝑒𝑤ℎ𝑒𝑟𝑒}
7. Infinite set 𝑆 is linearly independent iff every finite subset is linearly independent

Matrices

Definitions
2 2 𝑛
1. Involutory: 𝐴 = 𝐼𝑛; idempotent matrix: 𝐴 = 𝐴; nilpotent matrix: 𝐴 = 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑛 ∈ ℕ
𝑇 𝑇
2. Symmetric: 𝐴 = 𝐴; skew-symmetric/anti-symmetric: 𝐴 = −𝐴
θ θ
3. Hermitian matrix: 𝐴 = 𝐴; skew-Hermitian: 𝐴 = −𝐴
𝑇 θ
4. Orthogonal matrix: 𝐴 𝐴 = 𝐼; unitary matrix: 𝐴 𝐴 = 𝐼
−1
5. Similar matrices: for square matrices 𝐴, 𝐵: 𝐴 ∼ 𝐵 ⇔ 𝐵 = 𝐶 𝐴𝐶 for non-singular 𝐶
○ Same characteristic equation and hence eigenvalues

Utkarsh Kumar
 A= ○ , B= are non similar matrices with same eigenvalue, det, trace, characteristic
polynomial; but B is non-diagonalisable
6. Rank of matrix = largest 𝑟: 𝑟 × 𝑟 non-singular minor exists in matrix
○ Rank = dim columnspace = dim rowspace
○ Proof: reduce A to normal form; show that row operations don’t change column rank (
𝐶𝑖 → 𝐸𝐶𝑖 ⇒ (𝐸𝐶𝑗 𝐿𝐼 ⇒ 𝐶𝑗 𝐿𝐼)) and column operations don’t change row rank 𝑅𝑖 → 𝑅𝑖𝐸'
○ Not affected by elementary row and column operations, pre/post multiplication with non-singular
matrix, 𝑃 (∵ 𝑃 = Π𝑅𝑖 Π𝐶𝑖)
7. Equivalent matrices (same rank): matrices of same order that can be reduced to same normal form
○ Can be obtained from each other by series of elementary row/column operations
○ 𝐵 = 𝑃𝐴𝑄, 𝑃 𝑎𝑛𝑑 𝑄 being non-singular matrices
8. Row reduced echelon form: pivots unity, all other elements in pivot column vanish, pivots are arranged

Theorems
1. Every matrix can be written as sum of symmetric (Hermitian) and skew-symmetric matrices
𝑇 𝑇 𝑇 𝑇
2. 𝐴 𝐵 + 𝐵 𝐴 is symmetric, 𝐴 𝐵 − 𝐵 𝐴 is skew-symmetric
𝑛 𝑛
3. ∑ 𝑎𝑖𝑘𝐶𝑗𝑘 = ∑ 𝑎𝑘𝑖𝐶𝑘𝑗 = det(𝐴) 𝑖𝑓 𝑖 = 𝑗, 0 otherwise
𝑘 𝑘
4. Row operations on 𝐴𝐵 can be affected to pre-factor, column operations to post-factor
5. Matrix 𝐴 = (∏𝑃)𝑋𝑟(∏𝑄), where P (Q) represents elementary row (column) operations, 𝑋𝑟 is of
𝐼𝑟 𝐼 |0
normal/first canonical form ([𝐼 |0], [
𝑟 0
], [ 0𝑟 | 0 ])
6. 𝑟𝑎𝑛𝑘(𝐴𝐵) ≤ 𝑟𝑎𝑛𝑘(𝐴), 𝑟𝑎𝑛𝑘(𝐵)
a. Proof: rank doesn’t change on multiplication with non-singular matrix
b. Proof 2: 𝑅𝑖(𝐴𝐵) = Σ𝐴𝑖,𝑗𝑅 (𝐵); 𝐶𝑖(𝐴𝐵) = Σ𝐶𝑗(𝐴)𝐵𝑗,𝑖
𝑗
c. det 𝐴𝐵 = det 𝐴. det 𝐵 hint: reduce to upper triangular matrices
7. System of linear equations:
a. 𝑟𝑎𝑛𝑘(𝐴|𝐵) > 𝑟𝑎𝑛𝑘(𝐴) ⇒ 𝑛𝑜 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠
b. 𝑟𝑎𝑛𝑘(𝐴|𝐵) = 𝑟𝑎𝑛𝑘(𝐴) = #𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 ⇒ 𝑢𝑛𝑖𝑞𝑢𝑒 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛
c. 𝑟𝑎𝑛𝑘(𝐴|𝐵) < #𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 ⇒ 𝑖𝑛𝑓𝑖𝑛𝑖𝑡𝑒 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠
∆𝑖
d. Cramer’s rule: 𝑥 = ;∆ 𝑖 obtained by replacing 𝑖th column with 𝑏
𝑖 ∆
𝑛 𝑛 𝑛 2 𝑛 𝑛−1 1
8. 𝑂(𝐺𝐿(𝑛, 𝑍𝑝) = (𝑝 − 1)(𝑝 − 𝑝)(𝑝 − 𝑝 )... (𝑝 − 𝑝 ); 𝑂(𝑆𝐿(𝑛, 𝑍𝑝)) = 𝑝−1
𝑂(𝐺𝐿(𝑛, 𝑍𝑝))
9. 𝐴𝐵, 𝐵𝐴 (det 𝐴, det 𝐵 ≠ 0)have same eigenvalues; proof: (𝐴𝐵)𝑋 = λ𝑋 ⇔ 𝐵𝐴(𝐵𝑋) = λ(𝐵𝑋)

Eigenvalues and eigenvectors

1. Eigenvalues satisfy characteristic equation of the matrix

Utkarsh Kumar
 ○ Geometric multiplicity ≤ algebraic multiplicity for all eigenvalues
θ
2. Hermitian matrices have real eigenvalues (proof by pre-multiplying with 𝑋 )
○ Skew-Hermitian matrices have purely imaginary or 0 eigenvalues
○ Unitary matrix have eigenvalues with unit modulus
○ Real symmetric matrix ⇒ orthogonal eigenvectors
3. Eigenvectors corresponding to distinct eigenvalues are linearly independent
4. Cayley-Hamilton theorem: Every matrix satisfies its characteristic equation, det(𝑡 𝐼𝑛 − 𝐴) = 0
𝑖
○ Proof: 𝐵 = 𝑎𝑑𝑗(𝑡𝐼𝑛 − 𝐴) = ∑𝑡 𝐵𝑖;
𝑛−1
𝑛−1 𝑖 𝑖
○ (𝑡𝐼𝑛 − 𝐴). 𝐵 = 𝑡 𝐵𝑛−1 + ∑ 𝑡 (𝐵𝑖−1 − 𝐴𝐵𝑖 ) − 𝐴𝐵0 = 𝑝(𝑡)𝐼𝑛 = (∑𝑐𝑖𝑡 )𝐼𝑛
𝑖=1
𝑖
○ Comparing, 𝑐𝑛𝐼 = 𝐵𝑛−1, 𝑐𝑖𝐼 = 𝐵𝑖−1 − 𝐴𝐵𝑖, 𝑐0𝐼 = − 𝐴𝐵0; multiply with 𝐴 and add
5. Diagonalizability: 𝑛 linearly independent eigenvectors ⇔ diagonalizable (proof)
○ 𝑛 distinct eigenvalues ⇒ diagonalizable
○ Two matrices with same set of distinct eigenvalues are diagonalizable
○ Geometric multiplicity = algebraic multiplicity for all eigenvalues ⇒ diagonalizable
○ Real symmetric matrix ⇒ orthogonal eigenvectors ⇒ orthogonally diagonalizable

Inner product space

1. Inner product properties:
θ 𝑇
a. 𝑣. 𝑤 = 𝑤. 𝑣 = 𝑤 𝑣 = 𝑤 𝑣(over ℝ)
b. 𝑣. 𝑣 > 0, 𝑣. 𝑣 = 0 ⇔ 𝑣 = 0
c. (𝑎α + 𝑏β). 𝑣 = 𝑎(α. 𝑣) + 𝑏(β. 𝑣)
2. 𝑉(ℝ) is called Euclidean space; 𝑉(ℂ) is called unitary space
θ
3. 𝑛𝑜𝑟𝑚(𝑣) = ||𝑣|| = 𝑣. 𝑣 = 𝑣𝑣
−1 𝑇
4. Orthogonally similar: 𝐴 ∼𝑜𝑟𝑡ℎ𝑜 𝐵 ⇔ 𝐵 = 𝑃 𝐴𝑃 = 𝑃 𝐴𝑃, where 𝑃 is orthogonal

Matrix examples:

1. Non-similar matrices with same characteristic equation: ,

2. Non-diagonalisable square matrix: : eigenvalue 1, but single LI eigenvector

Utkarsh Kumar
 Calculus

Continuity of functions
1. Continuous function on closed interval ⇒ bounded, attains its bounds (proof: consider
< 𝑥𝑛 >: 𝑓(𝑥𝑛) > 𝑛; Bolzano Weierstrauss says there is a convergent subsequence)
2. 𝑓(𝑥 + 𝑦) = 𝑓(𝑥) + 𝑓(𝑦), 𝑓 𝑖𝑠 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝑎𝑡 𝑠𝑜𝑚𝑒 𝑝𝑜𝑖𝑛𝑡 ⇒ 𝑓(𝑥) = 𝑎𝑥
3. Uniform continuity: ε doesn’t depend on point in consideration (stronger than continuity)
○ ∀ϵ > 0, ∃δ > 0: ∀𝑥1, 𝑥2 |𝑥1 − 𝑥2| < δ ⇒ |𝑓(𝑥1) − 𝑓(𝑥2)| < ϵ
2
○ Continuity ⇏ uniform continuity (like 𝑓(𝑥) = 𝑥 𝑜𝑛 (0, ∞))
○ Continuity on compact set ⇒ uniform continuity (proof: for ϵ > 0, consider 𝐵(δ𝑥/2, 𝑥), that
forms open cover for compact set)
4. Intermediate Value Theorem: proof by order-completeness of ℝ (
𝑐 = sup{𝑥 ∈ [𝑎, 𝑏] | 𝑓(𝑥) < 𝑘} ⇒ 𝑓(𝑐) = 𝑘)

Differentiation

Differentiability of functions
2 1
1. Differentiable at all points ⇏ continuity of derivative (𝑦 = 𝑥 sin 𝑥
)
2. Rolle’s theorem:
𝑓 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝑜𝑛 [𝑎, 𝑏], 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑏𝑙𝑒 𝑜𝑛 (𝑎, 𝑏), 𝑓(𝑎) = 𝑓(𝑏) ⇒ 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑐 ∈ (𝑎, 𝑏), 𝑓'(𝑐) = 0
○ Proof: continuous on compact set, bounded, attains bounds, stationary point concept
𝑓(𝑏)−𝑓(𝑎)
3. LMVT: 𝑓 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝑜𝑛 [𝑎, 𝑏], 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑏𝑙𝑒 𝑜𝑛 (𝑎, 𝑏) ⇒ 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑐 ∈ (𝑎, 𝑏), 𝑓'(𝑐) = 𝑏−𝑎
○ Proof using Rolle’s theorem
4. Cauchy’s MVT:
𝑓'(𝑐) 𝑓(𝑏)−𝑓(𝑎)
𝑓, 𝑔 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝑜𝑛 [𝑎, 𝑏], 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑏𝑙𝑒 𝑜𝑛 (𝑎, 𝑏) ⇒ 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑐 ∈ (𝑎, 𝑏), 𝑔'(𝑐)
= 𝑔(𝑏)−𝑔(𝑎)
○ Proof using Rolle’s theorem
2
1 (𝑏−𝑎) 2
5. Taylor’s series: 𝑓(𝑏) = 𝑓(𝑎) + (𝑏 − 𝑎) 𝑓 (𝑎) + 2!
𝑓 (𝑎) +... + 𝑟𝑒𝑚𝑎𝑖𝑛𝑑𝑒𝑟 𝑡𝑒𝑟𝑚
𝑛+1 𝑛+1
○ Lagrange remainder = (𝑏 − 𝑎) 𝑓 (ξ)/(𝑛 + 1)!
𝑛 𝑛+1
○ Cauchy’s reminder = (𝑏 − 𝑎)(𝑏 − ξ) 𝑓 (ξ)/𝑛!
𝑏−𝑥 𝑝 2
○ Proof: 𝐺(𝑥) = 𝐹(𝑥) − ( 𝑏−𝑎
) 𝐹(𝑎); 𝐹(𝑥) = 𝑓(𝑏) − 𝑓(𝑥) − (𝑏 − 𝑥)𝑓'(𝑥) − (𝑏 − 𝑥) 𝑓''(𝑥)/2! −...
(n terms)
i. Rolle’s on 𝐺(𝑥) ⇒ 𝐺'(𝑐) = 0; 𝑝 = 1 ⇒ Cauchy’s remainder; 𝑝 = 𝑛 + 1 ⇒ Lagrange’s
remainder
○ Maclaurin expansion: take 𝑎 = 0 in Taylor series

Utkarsh Kumar
Functions of several variables
𝑥
1. Individual limits exist ⇏ simultaneous limit exists: 𝑓(𝑥, 𝑦) = 𝑥+𝑦
(individual limits not even equal)
𝑥𝑦
2. Partial derivatives exist and are equal ⇏ 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑖𝑡𝑦: 𝑓(𝑥, 𝑦) = 2 2
𝑥 +𝑦
2 2
3. Differentiability: 𝑓(𝑥 + ℎ, 𝑦 + 𝑘) − 𝑓(𝑥, 𝑦) = 𝐴𝑥 + 𝑏𝑦 + ℎ + 𝑘 ɸ(ℎ, 𝑘) ⇒ lim ɸ(ℎ, 𝑘) → 0
(ℎ,𝑘) → (0,0)

○ 𝐴 = 𝑓𝑥(𝑥, 𝑦), 𝐵 = 𝑓𝑦(𝑥, 𝑦); and hence independent of ℎ, 𝑘

○ Necessary for differentiability: 𝑓𝑥 exists and is bounded in nbd; 𝑓𝑦 exists at (𝑎, 𝑏)
○ Necessary for differentiability: 𝑓𝑥 is continuous, 𝑓𝑦 exists at (𝑎, 𝑏)
2 2
𝑥 −𝑦
4. 𝑓𝑥𝑦(𝑥, 𝑦) 𝑎𝑛𝑑 𝑓𝑦𝑥(𝑥, 𝑦) need not always be equal: 𝑓(𝑥, 𝑦) = 𝑥𝑦 2 2
𝑥 +𝑦
○ Schwarz theorem: 𝑓𝑥𝑦 continuous in nbd, 𝑓𝑥 exists ⇒ 𝑓𝑥𝑦 = 𝑓𝑦𝑥 (not necessary)
○ Young’s theorem: 𝑓𝑥, 𝑓𝑦 differentiable at (𝑎, 𝑏) ⇒ 𝑓𝑥𝑦 = 𝑓𝑦𝑥
5. Taylor’s series: 𝑓(𝑥, 𝑦) = 𝑓(𝑥0, 𝑦0) + [(𝑥 − 𝑥0) 𝑓𝑥(𝑥0, 𝑦0) + (𝑦 − 𝑦0)𝑓𝑦(𝑥0, 𝑦0)] +...

1 𝑖 𝑗 𝑛 𝑖 𝑗
○ =∑ ∑ 𝑖! 𝑗!
(𝑥 − 𝑥0) (𝑦 − 𝑦0) [∂ 𝑓/∂ 𝑥∂ 𝑦](𝑥,𝑦)
𝑛 𝑖+𝑗=𝑛

1 𝑖 𝑗 𝑛 𝑖 𝑗
○ Remainder, 𝑅𝑛 = ∑ 𝑖! 𝑗!
(α − 𝑥0) (β − 𝑦0) [∂ 𝑓/∂ 𝑥∂ 𝑦](α,β); (α, β) lies on line segment joining
𝑖+𝑗=𝑛
(𝑥0, 𝑦0) and (𝑥, 𝑦)

Extremas
1. bivariate calculus: consider stationary points and non-differentiable points
2
○ 𝑓𝑥𝑦 − 𝑓𝑥𝑥𝑓𝑦𝑦 < 0 ⇒ 𝑚𝑎𝑥𝑖𝑚𝑎 (𝑓𝑥𝑥 < 0) 𝑜𝑟 𝑚𝑖𝑛𝑖𝑚𝑎 (𝑓𝑥𝑥 > 0)
2
○ 𝑓𝑥𝑦 − 𝑓𝑥𝑥𝑓𝑦𝑦 > 0 ⇒ 𝑠𝑎𝑑𝑑𝑙𝑒 𝑝𝑜𝑖𝑛𝑡
2
○ 𝑓𝑥𝑦 − 𝑓𝑥𝑥𝑓𝑦𝑦 = 0 ⇒ 𝑖𝑛𝑐𝑜𝑛𝑐𝑙𝑢𝑠𝑖𝑣𝑒
2
2. >3 variables: consider stationary points where 𝑑 𝑓(𝑥1, 𝑥2,...) has same sign for arbitrary 𝑑𝑥1, 𝑑𝑥2,...
(curvature up/down in all directions)
2 2
○ Maxima if 𝑑 𝑓 < 0 (concave down); Minima if 𝑑 𝑓 > 0 (concave up)
3. Lagrange’s method of multipliers: optimise ɸ = 𝑓(𝑥, 𝑦) + ∑ λ𝑖𝑐𝑖; 𝑐𝑖 being constraints
2
○ Verify maxima/minima by checking for same sign of 𝑑 ɸ

Jacobian
∂(𝑢,𝑣)
1. Jacobian = ∂(𝑥,𝑦)
= |{𝑢'} {𝑣'}| = 0 ⇔ 𝑢, 𝑣 are functionally related
∂(𝑢,𝑣) ∂(𝑥,𝑦)
2. ∂(𝑥,𝑦)
× ∂(𝑢,𝑣)
= 1

Utkarsh Kumar
 3. Jacobians are multipliable (chain rule)
∂(𝑓1,𝑓2,...) ∂(𝑦1,𝑦2,...) 𝑛 ∂(𝑓1,𝑓2,...)
4. In case of 𝑛 implicit equations 𝑓 , 𝑓 ,..., ∂(𝑦 ,𝑦 ,...)
. ∂(𝑥1,𝑥2,...)
= (− 1) ∂(𝑥1,𝑥2,...)
1 2 1 2

Uniform convergence
1. Definition: < 𝑓𝑛(𝑥) > ⇉ 𝑓(𝑥) ⇔ ∀ϵ > 0, ∃𝑛0 > 0: |𝑓(𝑥) − 𝑓𝑛(𝑥)| ≤ ϵ ∀𝑛 > 𝑛0, 𝑥 ∈ 𝐷
2. Continuity
○ 𝑓𝑛(𝑥) 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠, {𝑓𝑛(𝑥)} ⇉ 𝑓(𝑥) ⇒ 𝑓 𝑖𝑠 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠
○ 𝑢𝑛(𝑥) 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠, ∑𝑢𝑛(𝑥) ⇉ 𝑢(𝑥) ⇒ 𝑢 𝑖𝑠 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠
3. Integrability
𝑏 𝑏
○ 𝑓𝑛(𝑥) integrable, {𝑓𝑛(𝑥)} ⇉ 𝑓(𝑥) ⇒ 𝑓 𝑖𝑠 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑏𝑙𝑒, lim ∫ 𝑓𝑛(𝑥)𝑑𝑥 = ∫ 𝑓(𝑥)𝑑𝑥
𝑛→∞ 𝑎 𝑎
𝑏 𝑏
○ 𝑢𝑛(𝑥) integrable, ∑𝑢𝑛(𝑥) ⇉ 𝑢(𝑥) ⇒ 𝑢 𝑖𝑠 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑏𝑙𝑒, ∑ ∫ 𝑢𝑛(𝑥)𝑑𝑥 = ∫ 𝑢(𝑥)𝑑𝑥
𝑎 𝑎
4. Differentiability
○ 𝑓𝑛(𝑥) is differentiable, 𝑓'𝑛(𝑥) is continuous, 𝑓𝑛(𝑥) → 𝑓(𝑥), 𝑓𝑛'(𝑥) ⇉ 𝑔(𝑥) ⇒ 𝑓'(𝑥) = 𝑔(𝑥)
○ 𝑓𝑛(𝑥)differentiable, 𝑓'𝑛(𝑥) continuous, ∑𝑓𝑛(𝑥) = 𝑓(𝑥), ∑𝑓𝑛'(𝑥) ⇉ 𝑔(𝑥) ⇒ 𝑓'(𝑥) = 𝑔(𝑥)

Uniform convergence for sequences

1. ∀ϵ > 0 ∃𝑚0 ∈ ℕ: ∀𝑚 ≥ 𝑚0, | 𝑓𝑚(𝑥) − 𝑓(𝑥)| < ϵ ∀𝑥 ∈ ℝ (𝑚0 independent of 𝑥)
2. Mn test: 𝑓(𝑥) = lim 𝑓𝑛(𝑥); 𝑀𝑛 = sup𝑥∈ℝ | 𝑓(𝑥) − 𝑓𝑛(𝑥)|; lim 𝑀𝑛 = 0 ⇔ {𝑓𝑛(𝑥)} ⇉ 𝑓(𝑥)
𝑛→∞ 𝑛→∞

Uniform convergence for series

1. Series uniformly converges iff SOPS uniformly converges
2. Weierstrass’ M test: 𝑀 ≥ | 𝑢 (𝑥)| ∀𝑥 ∈ ℝ; Σ𝑀 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑠 ⇒ Σ 𝑢 (𝑥) ⇉ 𝑢(𝑥)
𝑛 𝑛 𝑛 𝑛

Utkarsh Kumar
 Analytic Geometry

Sphere
2 2 2 2 2 2
1. Need to choose 2 spheres: 𝑥 + 𝑦 + 𝑧 + 2𝑢1𝑥 + 𝑑 = 0, 𝑥 + 𝑦 + 𝑧 + 2𝑢2𝑥 + 𝑑 = 0
2 2 2
2. Orthogonality: 𝑟1 + 𝑟2 = 𝑑

Cone
1. Condition for cone: ∆ = det(𝑎, 𝑏, 𝑐, 𝑓, 𝑔, ℎ, 𝑢, 𝑣, 𝑤, 𝑑) = 0; vertex by 𝑓𝑥 = 𝑓𝑦 = 𝑓𝑧 = 0
2. Cone with origin as vertex ⇔ homogeneous 2 degree equation
3. Cone with 3 ⊥ generators ⇔ 𝑎 + 𝑏 + 𝑐 = 0
4. Cone through coordinate axes: ℎ𝑥𝑦 + 𝑓𝑦𝑧 + 𝑔𝑧𝑥 = 0
5. Reciprocal cone: generated by lines through vertex ⊥ to tangent planes to original plane
○ Obtained by replacing 𝑎, 𝑏, 𝑐, 𝑓, 𝑔, ℎ with their cofactors in ∆' = det(𝑎, 𝑏, 𝑐, 𝑓, 𝑔, ℎ)
6. Enveloping curve: a curve that touches each curve of a given family
2
○ Enveloping cone of sphere: 𝑆𝑆1 = 𝑇
2 2 2
2 2 2 𝑙 𝑚 𝑛
7. Plane 𝑙𝑥 + 𝑚𝑦 + 𝑛𝑧 = 0 touches cone 𝑎𝑥 + 𝑏𝑦 + 𝑐𝑧 = 0: 𝑎
+ 𝑏
+ 𝑐
=0
8. Normal plane through generator 𝐺 will be perpendicular to tangent plane at G and G itself

Cylinder

General conic
1. Condition of tangency: Consider point (α, β, γ) on conicoid; then a line through it has zero roots
2 2 2
2 2 2 𝑙 𝑚 𝑛
2. Central conic: 𝑎𝑥 + 𝑏𝑦 + 𝑐𝑧 = 1 has tangent planes 𝑃: 𝑙𝑥 + 𝑚𝑦 + 𝑛𝑧 =± 𝑎
+ 𝑏
+ 𝑐
2 2 2 −1 −1 −1
a. Director circle: 𝑥 + 𝑦 + 𝑧 = 𝑎 +𝑏 +𝑐
2 2
3. Sections of 𝑎𝑥 + 𝑏𝑦 + 2ℎ𝑥𝑦 + 2𝑢𝑥 + 2𝑣𝑦 + 𝑑 = 0 by 𝑧 = 0 will be
a. Parabola if second degree terms form perfect square
b. Rectangular hyperbola if 𝑎 + 𝑏 = 0
4. Section of a conic with centre (α, β, γ) is obtained by plane 𝑇 = 𝑆1
5. Plane of contact: 𝑇 = 0; polar plane: locus of harmonic conjugate of P wrt R, S: 𝑇 = 0
6. Pole of given plane, 𝑃: 𝑙𝑥 + 𝑚𝑦 + 𝑛𝑧 = 𝑝 is point X whose polar plane is P, ie 𝑇(𝑋) ≡ 𝑃
7. Conjugate points 𝑃, 𝑄: polar plane of 𝑃 passes through 𝑄 (and vice versa)
8. Conjugate planes 𝑃1, 𝑃2: pole of 𝑃1 lies on 𝑃2 (and vice-versa)
9. Conjugate lines 𝐿1, 𝐿2: conjugate plane of each point in 𝐿1 passes through 𝐿2 (and vice-versa)

Utkarsh Kumar
 10. Conjugate semi-diameters 𝑂𝑃, 𝑂𝑄, 𝑂𝑅: conjugate plane of one passes through other two
2 2 2 2 2 2
a. Condition:Σ𝑥𝑖 = 𝑎 , Σ𝑦𝑖 = 𝑏 , Σ𝑧𝑖 = 𝑐 , Σ𝑥𝑖𝑦𝑖 = 0 = Σ𝑥𝑖𝑥𝑗 (think
𝑀𝑎𝑡(𝑥1, 𝑥2, 𝑥3, 𝑦1, 𝑦2, 𝑦3, 𝑧1, 𝑧2, 𝑧3))
11. Conjugate plane 𝑃 to (𝑙, 𝑚, 𝑛): locus of midpoints of system of chord parallel to (𝑙, 𝑚, 𝑛)
a. Conjugate planes 𝑃1, 𝑃2, 𝑃3: 𝑃1 is conjugate to 𝑃2 ∩ 𝑃3 and so on
12. Eccentricity
2 2 2
a. Ellipse, 𝑏 = 𝑎 (1 − 𝑒 ); 𝑒 < 1
2 2 2
b. Hyperbola, 𝑏 = 𝑎 (𝑒 − 1); 𝑒 > 1
c. 𝑒 = 0 for circle; 𝑒 = 1 for parabola
𝑥−𝑎cosθ 𝑦−𝑏sinθ 𝑧
13. Hyperbolic paraboloid generators through principal ellipse:
𝑎sinθ
= −𝑏cosθ
= ±𝑐

Reduction of 2nd degree equations to canonical forms

1. Compute λ𝑖’s as eigenvalues of ∆ = 𝑀𝑎𝑡(𝑎, 𝑏, 𝑐, 𝑓, 𝑔, ℎ)
a. λ𝑖 ≠ 0 ∀𝑖
i. 𝑑' = 𝑑 + 𝑢α + 𝑣β + 𝑤γ
2 2 2
ii. λ1𝑥 + λ2𝑦 + λ3𝑧 + 𝑑' = 0
iii. 𝑓𝑥 = 𝑓𝑦 = 𝑓𝑧 = 0 ⇒ (α, β, γ) is centre
b. Else, (𝑙, 𝑚, 𝑛) is eigenvector corresponding to 0, normalised
i. 𝑘 = 𝑙𝑢 + 𝑚𝑣 + 𝑛𝑤 ≠ 0
● Centre: 𝑓𝑥/𝑙 = 𝑓𝑦/𝑚 = 𝑓𝑧/𝑛 = 2𝑘; 𝑘Σ𝑙𝑥 + Σ𝑢𝑥 + 𝑑 = 0
● 𝑑' = 0
ii. 𝑘 = 𝑙𝑢 + 𝑚𝑣 + 𝑛𝑤 = 0
● Line of centres: 𝑓𝑥 = 𝑓𝑦 = 𝑓𝑧 = 0
● 𝑑' = 𝑑 + 𝑢α + 𝑣β + 𝑤γ
2 2
iii. λ1𝑥 + λ2𝑦 + 2𝑘𝑧 + 𝑑' = 0
2. If 2nd degree terms form perfect square,
a. Linear terms line up with square term ⇒ pair of planes
2 2
b. Else, (𝑝𝑥 + 𝑞𝑦 + 𝑟𝑧 + λ) = − 2𝑢𝑥 − 2𝑣𝑦 − 2𝑤𝑧 − 𝑑 + λ + 2λ(𝑝𝑥 + 𝑞𝑦 + 𝑟𝑧)
i. Normal vectors of planes on LHS, RHS are ⊥⇒ λchosen
3. Direction ratios of normal axes/planes are eigenvectors for ∆
4. Type of conicoid:
2 2 2
𝑥 𝑦 𝑧
a. Ellipsoid: 2 + 2 + 2 =1
𝑎 𝑏 𝑐
2 2 2
𝑥 𝑦 𝑧
b. Hyperboloid of one sheet: 2 + 2 − 2 =1
𝑎 𝑏 𝑐

Utkarsh Kumar
 2 2 2
𝑥 𝑦 𝑧
c. Hyperboloid of two sheets: 2 − 2 − 2 =1
𝑎 𝑏 𝑐
2 2
𝑥 𝑦 2𝑧
d. Elliptic paraboloid: 2 + 2 = 𝑐
𝑎 𝑏
2 2
𝑥 𝑦 2𝑧
e. Hyperbolic paraboloid: 2 − 2 = 𝑐
𝑎 𝑏
2
f. Parabolic cylinder: 𝑥 = 4𝑎𝑦
2 2
𝑥 𝑦
g. Elliptic cylinder: 2 + 2 =1
𝑎 𝑏
h. Conicoid of revolution: λ1 = λ2
5. Generators for hyperboloids: two sets(λ, µ)

Miscellaneous
1. Check degree of parameterised curve, 𝑥 = 𝑋(λ), 𝑦 = 𝑌(λ), 𝑧 = 𝑍(λ): intersection with arbitrary plane
𝑃: 𝑙𝑥 + 𝑚𝑦 + 𝑛𝑧 = 𝑝, count solutions for λ
2. Plane of contact: locus of points on conicoid, tangent planes at which pass through given point
3. Polar points and planes
4. Enveloping cone

Utkarsh Kumar
 Ordinary Differential Equations
1. Order: highest derivative term involved; Degree: power of highest degree term when made free of
negative and fraction powers
2. Exact form ⇔ ∂𝑀/∂𝑦 = ∂𝑁/∂𝑥
−1
a. Homogeneous eqn ⇒ 𝐼𝐹 = (𝑀𝑥 + 𝑁𝑦)
1 𝑑𝑥 𝑑𝑦 1 𝑑𝑥 𝑑𝑦
■ proof:𝑀𝑑𝑥 + 𝑁𝑑𝑦 = 2
(𝑀𝑥 + 𝑁𝑦)( 𝑥
+ 𝑦
)+ 2
(𝑀𝑥 − 𝑁𝑦)( 𝑥
− 𝑦
)
−1
b. 𝑀 = 𝑦𝑓(𝑥𝑦), 𝑁 = 𝑥𝑔(𝑥𝑦) ⇒ 𝐼𝐹 = (𝑀𝑥 − 𝑁𝑦) ; proof like above
∫ 𝑓(𝑥) 𝑑𝑥
c. (𝑀𝑦 − 𝑁𝑥)/𝑁 = 𝑓(𝑥) ⇒ 𝐼𝐹 = 𝑒
∫ 𝑔(𝑦) 𝑑𝑦
d. (𝑁𝑥 − 𝑀𝑦)/𝑀 = 𝑔(𝑦) ⇒ 𝐼𝐹 = 𝑒
3. Don’t forget to check for linear in 𝑥 or 𝑦
𝑖 𝑖 𝑧
4. Euler-Cauchy form: Σ 𝑥 𝐷 𝑦 = 𝑓(𝑥) ⇒ 𝑥 = 𝑒
α −1−α
a. 1/(𝐷 − α) 𝑋(𝑥) = 𝑥 ∫ 𝑥 𝑋(𝑥) 𝑑𝑥; 𝐷 ≡ 𝑑/𝑑𝑧; 𝑧 = log 𝑥
5. CF, PI:
α𝑥 −α𝑥
a. 1/(𝐷 − α) 𝑋(𝑥) = 𝑒 ∫ 𝑒 𝑋(𝑥) 𝑑𝑥; 𝐷 ≡ 𝑑/𝑑𝑥
6. Solving 𝑦" + 𝑃𝑦' + 𝑄𝑦 = 𝑅 (2 deg eqn with nonconstant coefficients)
a. When 𝑧, part of CF is known: 𝑦 = 𝑉𝑧; 𝑉" + (2𝑧'/𝑧 + 𝑃)𝑉' = 𝑅/𝑧
−∫ 𝑃/2 𝑑𝑥 2
b. Normal form: 𝑦 = 𝑉𝑧; 𝑧 = 𝑒 ; 𝑉" + (𝑄 − 𝑃 /4 − 𝑃'/2)𝑉 = 𝑅/𝑧
c. Change of variable 𝑥 → 𝑧: coeff of 𝑦 vanishes or coeff of 𝑦 becomes 1
d. Variation of parameters to find 𝑦𝑝
i. 𝑊 = det (𝑢, 𝑣, 𝑢', 𝑣'); 𝑢, 𝑣 are solutions of the homogeneous equation
ii. 𝑃𝐼 = 𝑢φ + 𝑣ψ; φ = ∫ − 𝑅𝑣/𝑊; ψ = ∫ 𝑅𝑢/𝑊
iii. Proof: find φ and ψ: φ𝑢 + ψ𝑣 is a particular solution; take 𝑢φ' + 𝑣ψ' = 0
𝑧'
iv. Can find other part of CF using 𝑣'' + (𝑃 + 2 𝑧
)𝑣' = 0

Clairaut’s equation
1. 𝑦 = 𝑝𝑥 + 𝑓(𝑝) ⇒ 𝑦 = 𝑝𝑐 + 𝑓(𝑐)
2. Check for solvability for 𝑥, 𝑦, 𝑝
3. Useful substitutions:
2 2
a. 𝑥 = 𝑋, 𝑦 = 𝑌 ⇒ (𝑦𝑝/𝑥) = 𝑃
2 2
b. 1/𝑥 = 𝑋, 1/𝑦 = 𝑌 ⇒ 𝑥 𝑝/𝑦 = 𝑃
c. Identify expressions from derivative terms: 𝑥𝑝 + 𝑦 ⇒ 𝑥𝑦 = 𝑣 etc
2 2
d. Watch out for expressions: 𝑥 + 𝑦 ⇒ polar coordinates etc

Utkarsh Kumar
 1

1
Image taken from IMS notes

Utkarsh Kumar
Singular solution
1. Obtained as special case of given ODE (satisfies it); usually not part of general solution
2. ODE 𝑓 = 0 with solution φ = 0: 𝑒𝑙𝑖𝑚𝑖𝑛𝑎𝑡𝑒 𝑝 𝑓𝑟𝑜𝑚 𝑓 = 0 = ∂𝑓/∂𝑝 𝑜𝑟 𝑐 𝑓𝑟𝑜𝑚 φ = 0 = ∂φ/∂𝑐
3. Extraneous loci: singular solution but doesn’t satisfy given ODE; obtain from both
∂𝑓 ∂φ
𝑓= ∂𝑝
= 0; φ = ∂𝑐
=0
2
○ p-discriminant = 𝐸𝑇 𝐶
3 2
○ c-discriminant = 𝐸𝐶 𝑁

Family of curves
𝑑𝑟 2 𝑑θ
1. Orthogonal family: replace 𝑝 → (− 1)/𝑝; 𝑑θ
→− 𝑟 𝑑𝑟
2. Oblique family at angle α: 𝑝 → (𝑝 + tan α)/(1 − 𝑝 tan α)

Laplace transform
∞
−𝑠𝑡
1. ℒ ( 𝑓(𝑡) ) = ∫ 𝑒 𝑓(𝑡) 𝑑𝑡 = 𝐹(𝑠)
0
𝑎𝑡
2. ℒ ( 𝑓(𝑎𝑡 ) = 1/𝑎 𝐹(𝑠/𝑎); ℒ (𝑒 𝑓(𝑡 ) = 𝐹(𝑠 − 𝑎)
∞
𝑛 𝑛 𝑛
3. ℒ (𝑡 𝑓(𝑡) ) = (− 1) 𝐷 𝐹(𝑠); ℒ ( 𝑓(𝑡)/𝑡 ) = ∫ 𝐹(𝑠) 𝑑𝑠
𝑠
𝑡
1
4. ℒ ( 𝑓'(𝑡) ) = 𝑠ℒ (𝑓) − 𝑓(0); ℒ ( ∫ 𝑓(𝑡) 𝑑𝑡 ) = 𝑠
ℒ (𝑓(𝑡) )
0
−𝑎𝑠
5. ℒ ( 𝑓(𝑡 − 𝑎). 𝐻(𝑡 − 𝑎) ) = 𝑒 ℒ (𝑓)
𝑡
−1
6. Convolution: ℒ (𝐹𝐺) = 𝑓 * 𝑔 = ∫ 𝑓(𝑢). 𝑔(𝑡 − 𝑢) 𝑑𝑢(proof by changing order of integration)
0
𝑛 𝑛+1 𝑎 𝑠
7. ℒ (𝑡 ) = Γ(𝑛 + 1)/𝑠 (𝑛≠ − 1); ℒ(sin 𝑎𝑡) = 2 2 ; ℒ(cos 𝑎𝑡) = 2 2
𝑠 +𝑎 𝑠 +𝑎
−1 1 1 −1 1 −1/𝑠
8. As a last resort, consider taylor series: ℒ(sin 𝑡), ℒ ( 𝑠 cos 𝑠
), ℒ ( 𝑠 𝑒 )
9. Initial value theorem: lim 𝑓(𝑡) = lim 𝑠. 𝐹(𝑠); final value theorem: lim 𝑓(𝑡) = lim 𝑠. 𝐹(𝑠)
𝑡→0 𝑠→∞ 𝑡→∞ 𝑠→0

Miscellaneous
∂𝑓 ∂𝑓
1. Euler’s equation: 𝑓(𝑥, 𝑦) is homogeneous of degree 𝑛 ⇒ 𝑥 ∂𝑥
+𝑦 ∂𝑦
= 𝑛 𝑓(𝑥, 𝑦)
2. Wronskian, 𝑊 = 𝑑𝑒𝑡(𝑢, 𝑣,..; 𝑢', 𝑣',...; 𝑢'', 𝑣'', 𝑤'',...)
a. 𝑊 ≠ 0 ⇒ linearly independent functions
b. 𝑊 = 0 is necessary, not sufficient condition for linear dependence of functions
c. If 𝑢, 𝑣 are solutions to ODE, 𝑊 is either identically 0 or never 0; 𝑊 = 0 ⇔ linear independence

Utkarsh Kumar
 Vector Analysis
1. Serret Frenet formulae:
𝑑𝑇 𝑑𝐵
○ Curvature vector, 𝑑𝑠 = 𝑘 𝑁; Torsion vector, 𝑑𝑠
= − τ𝑁
˙ →¨
→ →˙ →¨ →¨˙
|𝑟 × 𝑟| [𝑟 𝑟 𝑟]
○𝑘 = ˙2
→
, τ= ˙ →¨ 2
→
|𝑟| |𝑟 × 𝑟|
˙
→ ¨
→ →˙ →¨ →¨˙
[𝑟 𝑟 𝑟]
i. Proof using 𝑘 = |𝑟 × 𝑟|, τ = ˙ →¨ 2
→
, derivatives taken wrt 𝑠
|𝑟 × 𝑟|
∂𝑄 ∂𝑃
2. Green’s identity: ∮ 𝑃𝑑𝑥 + 𝑄𝑑𝑦 = ∬𝐴( ∂𝑥
− ∂𝑦
) 𝑑𝑥𝑑𝑦; A is area inside simple closed curve C
𝐶
○ Check that the function is defined both on C and within A
∂𝑄
○ Proof: divide curve into two parts; show ∬𝐴 ∂𝑥
𝑑𝑥𝑑𝑦 = ∮ 𝑄 𝑑𝑥
𝐶
→ → → → ^
3. Stoke’s theorem: ∮ 𝐹. 𝑑𝑟 = ∬𝑆 ∇ × 𝐹. 𝑛 𝑑𝑆; C is border of surface S
𝐶
○ check for continuity and differentiability within surface
○ Take clockwise line integral (seen from +Z) if surface is below curve
→ ^ → →
4. Gauss’ divergence theorem: ∫ 𝐹. 𝑛 𝑑𝑆 = ∭𝑉∇. 𝐹 𝑑𝑉; S is surface of volume V
𝑆
○ check for continuity and differentiability within volume
→→ → → → → → → →
5. Scalar triple product, [𝑢 𝑣 𝑤] = (𝑢 × 𝑣). 𝑤 = 𝑑𝑒𝑡 (𝑢, 𝑣, 𝑤)
→ → →
○ [𝑢 𝑣 𝑤] is volume of parallelepiped, and 6 × volume of tetrahedron

Miscellaneous terms
→ → → → → →
1. Reciprocal system of vectors for 𝑢, 𝑣, 𝑤 is 𝑢', 𝑣', 𝑤':
→ → → → → →
a. 𝑢. 𝑢' = 𝑣. 𝑣' = 𝑤. 𝑤' = 1
→ → → →
b. 𝑢. 𝑣' = 𝑣. 𝑤' =... = 0
→ →
→ 𝑣×𝑤
c. Thus, 𝑢' = →→ → ...
[𝑢 𝑣 𝑤]
→ → → → → →
2. 𝑎 is irrotational ⇒ ∇ × 𝑎 = 0; 𝑎 is solenoidal ⇒ ∇. 𝑎 = 0
∂𝐹 → ^
3. ^ = ∇𝐹. 𝑛
∂𝑛

Utkarsh Kumar
Vector integration
→ 𝑑𝑢→ 1 → →
1. ∫ 𝑢. 𝑑𝑡 𝑑𝑡 = 2
𝑢. 𝑢
→
→ 2
𝑑𝑢 → 𝑑𝑢
2. ∫ 𝑢 × 2 𝑑𝑡 = 𝑢 × 𝑑𝑡
𝑑𝑡
^ 2 2
3. ∫∫ φ∇φ. 𝑛 𝑑𝑆 = ∫∫∫ (|∇φ| + φ∇ φ )𝑑𝑉

Utkarsh Kumar
 Dynamics and Statics

Dynamics
● Lagrangian
○ 𝐿=𝐾−𝑈
○ 𝑑(∂𝐿/∂𝑥˙)/𝑑𝑡 = ∂𝐿/∂𝑥
● Hamiltonian
○ 𝑝 = ∂𝐿/∂𝑥˙
𝑥

○ 𝐻 = Σ 𝑝𝑥 𝑥˙ − 𝐿 = 𝐾 + 𝑈 (for conservative forces)

○ ∂𝐻/∂𝑝𝑥 = 𝑥˙; ∂𝐻/∂𝑥 = − 𝑝̇𝑥

● Motion under central force:
^ ^ 2 ^ ^ 2
(𝑟˙ 𝑟 + 𝑟θ̇ θ) = (𝑟¨ − θ̇ 𝑟) 𝑟 + (2𝑟˙θ̇ + 𝑟θ̈) θ ⇒ 𝑟¨ − θ̇ 𝑟 =
𝑑𝑣 𝑑 2
○ 𝑑𝑡
= 𝑑𝑡
− 𝑃, 𝑟 θ̇ = ℎ (constant)
𝑑𝑟 ^ 𝑑θ ^
○ 𝑑𝑡
= θ̇θ, 𝑑𝑡
=− θ̇𝑟
● Motion constrained on cycloid: 𝑠 = 4𝑎 sin ψ; θ = 2ψ; 𝑥 = 𝑎(θ + sin θ), 𝑦 = 𝑎(1 − cos θ)
2
𝑑𝑠
○ 𝑚 2 =− 𝑚𝑔 sin ψ ⇒ SHM with angular frequency ω = 𝑔/4𝑎
𝑑𝑡
2
○ 𝑁 − 𝑚𝑔 cos ψ = 𝑚𝑣 /ρ; ρ = 𝑑𝑠/𝑑ψ
● Kepler’s laws
2𝐹
● Principal axis: tan 2θ = 𝐵−𝐴
○ L, T are principal axes of lamina: ∫ 𝑝1𝑝2 𝑑𝑥 𝑑𝑦 = 0; 𝑝1, 𝑝2are signed distances from L, T

Utkarsh Kumar