Module 1: Linear Algebra

1.1 Vectors

Theory & Core Concepts

A vector is a mathematical object possessing both magnitude and direction, foundational to represent states in quantum
mechanics, features in machine learning, and transformations in computer graphics.

 Definition & Types: A matrix with a single column (n × 1) or single row (1 × n). In abstract linear algebra, a column
vector v ∈ ℝⁿ is written as a vertical array of components v = [v₁, v₂, ..., v] ᵀ.
 Zero & Unit Vectors: Standard zero vector 0 = [0, 0, ..., 0]ᵀ with zero magnitude. Unit vector u is any vector with
length ‖u‖ = 1. Normalized vector from v is u = v / ‖v‖.
 Basic Operations: Vector addition v + w = [v₁ + w₁, ..., vₙ + wₙ]ᵀ (head-to-tail rule). Scalar multiplication αv = [αv₁, ...,
αvₙ]ᵀ scales magnitude by |α| and reverses direction if α < 0.
 Dot Product & Geometric Interpretation: For u, v ∈ ℝⁿ, the dot product is defined as u · v = ∑ᵢ uᵢ vᵢ = uᵀ v = ‖u‖ ‖v‖
cos(θ), where θ is the angle between vectors. Geometrically, it measures directional alignment.
 Cross Product (3D): Defined specifically in 3D (u, v ∈ ℝ³), producing a vector orthogonal to both: u × v = [(u₂v₃ - u₃v₂),
(u₃v₁ - u₁v₃), (u₁v₂ - u₂v₁)]ᵀ. Magnitude ‖u × v‖ = ‖u‖ ‖v‖ sin(θ) equals the area of the parallelogram spanned by u and
v.
 Vector Norms: Norms measure vector length/magnitude:
• L₁ Norm (Manhattan): ‖v‖₁ = ∑ |vᵢ|
• L₂ Norm (Euclidean): ‖v‖₂ = √(∑ vᵢ²)
• L∞ Norm (Maximum): ‖v‖∞ = max |vᵢ|
 Orthogonality & Projections: Vectors u, v are orthogonal if u · v = 0 (θ = 90°). Vector projection of v onto u is projᵤ(v)
= ((v · u) / ‖u‖²) u.
 Linear Independence & Span: A set of vectors {v₁, ..., vₖ} is linearly independent if ∑ cᵢ vᵢ = 0 implies all scalar
coefficients cᵢ = 0. Span({v₁,...,vₖ}) is the set of all possible linear combinations.
 Basis & Dimension: A basis for a vector space V is a set of linearly independent vectors that spans V. Dimension
dim(V) is the number of vectors in any basis for V.

Graded Worked Examples (Easy to Hard)

Example 1 (Easy): Basic Operations and Vector Norms
Problem Statement:
Given vectors u = [3, -4, 0]ᵀ and v = [1, 2, 2]ᵀ in ℝ³:
a) Compute 2u - 3v.
b) Calculate the L₁, L₂, and L∞ norms of u.
c) Find the unit vector in the direction of u.
Solution / Step-by-Step Proof:
a) Scalar multiplication and subtraction:
2u = [6, -8, 0]ᵀ, 3v = [3, 6, 6]ᵀ
2u - 3v = [6 - 3, -8 - 6, 0 - 6]ᵀ = [3, -14, -6]ᵀ

b) Calculating Norms for u = [3, -4, 0]ᵀ:

 • L₁ Norm: ‖u‖₁ = |3| + |-4| + |0| = 3 + 4 + 0 = 7
• L₂ Norm: ‖u‖₂ = √(3² + (-4)² + 0²) = √(9 + 16 + 0) = √25 = 5
• L∞ Norm: ‖u‖∞ = max(|3|, |-4|, |0|) = 4

c) Unit Vector in direction of u:

û = u / ‖u‖₂ = (1/5)[3, -4, 0]ᵀ = [0.6, -0.8, 0]ᵀ.

Example 2 (Easy-Medium): Dot Product, Angle, and Projections

Problem Statement:
Let a = [2, 1, -2]ᵀ and b = [1, 1, 0]ᵀ.
a) Find the dot product a · b.
b) Find the angle θ between a and b.
c) Calculate the orthogonal projection of a onto b (proj_b(a)).
Solution / Step-by-Step Proof:
a) Dot Product:
a · b = (2)(1) + (1)(1) + (-2)(0) = 2 + 1 + 0 = 3

b) Angle θ:
‖a‖₂ = √(2² + 1² + (-2)²) = √(4 + 1 + 4) = 3
‖b‖₂ = √(1² + 1² + 0²) = √2
cos(θ) = (a · b) / (‖a‖₂ ‖b‖₂) = 3 / (3 √2) = 1 / √2
θ = arccos(1 / √2) = 45° or π/4 radians.

c) Vector Projection proj_b(a):

proj_b(a) = [(a · b) / ‖b‖₂²] b = [3 / (√2)²] [1, 1, 0]ᵀ = (3/2) [1, 1, 0]ᵀ = [1.5, 1.5, 0]ᵀ.

Example 3 (Medium): 3D Cross Product and Geometric Area

Problem Statement:
Find a vector orthogonal to both p = [1, 2, 3]ᵀ and q = [4, 5, 6]ᵀ using the cross product, and calculate the area of the
parallelogram formed by p and q.
Solution / Step-by-Step Proof:
a) Cross Product p × q:
p×q=|i j k|
|1 2 3|
|4 5 6|
= i(2·6 - 3·5) - j(1·6 - 3·4) + k(1·5 - 2·4)
= i(12 - 15) - j(6 - 12) + k(5 - 8)
= -3i + 6j - 3k = [-3, 6, -3]ᵀ.

b) Verification of Orthogonality:
(p × q) · p = (-3)(1) + (6)(2) + (-3)(3) = -3 + 12 - 9 = 0
(p × q) · q = (-3)(4) + (6)(5) + (-3)(6) = -12 + 30 - 18 = 0

c) Area of Parallelogram:
Area = ‖p × q‖₂ = √((-3)² + 6² + (-3)²) = √(9 + 36 + 9) = √54 = 3√6 ≈ 7.348 square units.
Example 4 (Medium-Hard): Testing Linear Independence and Span
Problem Statement:
Determine whether the set of vectors V = {v₁ = [1, 2, 3]ᵀ, v₂ = [0, 1, 2]ᵀ, v₃ = [2, 1, 0]ᵀ} is linearly independent in ℝ³, and
determine if V spans ℝ³.
Solution / Step-by-Step Proof:
Set up the equation c₁ v₁ + c₂ v₂ + c₃ v₃ = 0:
c₁[1, 2, 3]ᵀ + c₂[0, 1, 2]ᵀ + c₃[2, 1, 0]ᵀ = [0, 0, 0]ᵀ
Form the coefficient matrix A with vectors as columns:
A=[1 0 2]
[2 1 1]
[3 2 0]

Compute the determinant det(A):

det(A) = 1·(1·0 - 1·2) - 0 + 2·(2·2 - 1·3)
det(A) = 1(-2) + 2(4 - 3) = -2 + 2 = 0.

Conclusion:
Since det(A) = 0, the matrix columns are linearly dependent. Thus, {v₁, v₂, v₃} is NOT linearly independent.
Because 3 vectors in ℝ³ are linearly dependent, they cannot span ℝ³ (their span forms a 2D plane in ℝ³).

Example 5 (Hard): Gram-Schmidt Orthogonalization (Quantum / ML Basis construction)

Problem Statement:
Using the Gram-Schmidt process, convert the linearly independent set {u₁ = [1, 1, 0]ᵀ, u₂ = [1, 0, 1]ᵀ} into an orthonormal
basis {q₁, q₂}.
Solution / Step-by-Step Proof:
Step 1: Construct first orthogonal vector v₁ and unit vector q₁:
v₁ = u₁ = [1, 1, 0]ᵀ
‖v₁‖ = √(1² + 1² + 0) = √2
q₁ = v₁ / ‖v₁‖ = [1/√2, 1/√2, 0]ᵀ

Step 2: Construct second orthogonal vector v₂ by subtracting projection onto q₁:

proj_q₁(u₂) = (u₂ · q₁) q₁
u₂ · q₁ = (1)(1/√2) + (0)(1/√2) + (1)(0) = 1/√2
proj_q₁(u₂) = (1/√2) [1/√2, 1/√2, 0]ᵀ = [1/2, 1/2, 0]ᵀ
v₂ = u₂ - proj_q₁(u₂) = [1, 0, 1]ᵀ - [1/2, 1/2, 0]ᵀ = [1/2, -1/2, 1]ᵀ

Step 3: Normalize v₂ to get q₂:

‖v₂‖ = √((1/2)² + (-1/2)² + 1²) = √(1/4 + 1/4 + 1) = √(3/2)
q₂ = v₂ / ‖v₂‖ = √(2/3) [1/2, -1/2, 1]ᵀ = [1/√6, -1/√6, √(2/3)]ᵀ

Verification: q₁ · q₂ = (1/√2)(1/√6) + (1/√2)(-1/√6) + 0 = 1/√12 - 1/√12 = 0. Orthonormal basis constructed successfully.

1.2 Matrices

Theory & Core Concepts

Matrices represent linear transformations between vector spaces and serve as the backbone for solving system
equations, graph embeddings, dimensionality reduction (PCA), and quantum operator representations.
 Matrix Types: Square (m=n), Diagonal (a_ij=0 for i≠j), Identity (I), Symmetric (A = Aᵀ), Sparse (majority elements
zero).
 Matrix Operations: Addition A + B (element-wise), Multiplication C = AB where C_ij = ∑_k A_ik B_kj (requires inner
dimensions to match), Transpose Aᵀ where (Aᵀ)_ij = A_ji.
 Inversion & Determinant: Inversion: A⁻¹ exists iff det(A) ≠ 0. Inverse satisfies A A⁻¹ = A⁻¹ A = I. Determinant det(A)
measures volume scaling factor under matrix transformation.
 Rank & Row Echelon Form: Rank r(A) is the maximum number of linearly independent columns (or rows). Nullity = n
- rank(A) (Rank-Nullity Theorem).
 Systems of Equations (Ax = b): Gaussian Elimination reduces [A|b] to Row Echelon Form (REF) using row operations.
Gauss-Jordan Elimination further reduces to Reduced Row Echelon Form (RREF) to solve Ax = b directly.
 Matrix Factorizations: Core Matrix Factorizations:
• LU Decomposition: A = LU (Lower and Upper triangular) for efficient linear system solving.
• QR Decomposition: A = QR (Orthogonal Q, Upper triangular R) for least-squares fitting.
• Singular Value Decomposition (SVD): A = U Σ Vᵀ (U, V orthogonal; Σ singular values) for low-rank approximations
and data compression.
 Eigenvalues & Eigenvectors: Satisfies A v = λ v where λ is eigenvalue and v ≠ 0 eigenvector. Characteristic equation:
det(A - λI) = 0.
 Diagonalization & Spectral Theorem: A = P D P⁻¹ where D is diagonal containing eigenvalues, P contains
eigenvectors. Spectral Theorem: Any real symmetric matrix is orthogonally diagonalizable (A = Q D Qᵀ).

Graded Worked Examples (Easy to Hard)

Example 1 (Easy): Matrix Multiplication and Transpose Properties
Problem Statement:
Given matrices A = [[1, 2], [3, 4]] and B = [[2, 0], [1, 5]], calculate:
a) AB
b) BA
c) Verify that (AB)ᵀ = Bᵀ Aᵀ.
Solution / Step-by-Step Proof:
a) AB = [[1·2 + 2·1, 1·0 + 2·5], [3·2 + 4·1, 3·0 + 4·5]]
= [[2 + 2, 0 + 10], [6 + 4, 0 + 20]] = [[4, 10], [10, 20]]

b) BA = [[2·1 + 0·3, 2·2 + 0·4], [1·1 + 5·3, 1·2 + 5·4]]

= [[2, 4], [16, 22]] (Note: Matrix multiplication is non-commutative, AB ≠ BA).

c) Transpose Verification:
(AB)ᵀ = [[4, 10], [10, 20]]ᵀ = [[4, 10], [10, 20]]
Aᵀ = [[1, 3], [2, 4]], Bᵀ = [[2, 1], [0, 5]]
Bᵀ Aᵀ = [[2·1 + 1·2, 2·3 + 1·4], [0·1 + 5·2, 0·3 + 5·4]] = [[4, 10], [10, 20]].
Thus, (AB)ᵀ = Bᵀ Aᵀ holds true.
Example 2 (Easy-Medium): Matrix Inversion and Systems Ax = b
Problem Statement:
Solve the system of linear equations using matrix inversion A⁻¹:
2x + y = 5
4x + 3y = 11
Solution / Step-by-Step Proof:
Step 1: Write in matrix form Ax = b:
A = [[2, 1], [4, 3]], x = [x, y]ᵀ, b = [5, 11]ᵀ

Step 2: Find determinant det(A):

det(A) = (2)(3) - (1)(4) = 6 - 4 = 2 ≠ 0 (Inverse exists).

Step 3: Compute inverse A⁻¹:

For 2×2 matrix [[a, b], [c, d]], A⁻¹ = (1/det(A)) [[d, -b], [-c, a]]
A⁻¹ = (1/2) [[3, -1], [-4, 2]] = [[1.5, -0.5], [-2, 1]]

Step 4: Solve x = A⁻¹ b:

x = [[1.5, -0.5], [-2, 1]] [5, 11]ᵀ
x = (1.5)(5) + (-0.5)(11) = 7.5 - 5.5 = 2
y = (-2)(5) + (1)(11) = -10 + 11 = 1
Solution: x = 2, y = 1.

Example 3 (Medium): Gauss-Jordan Elimination and Matrix Rank

Problem Statement:
Find the Reduced Row Echelon Form (RREF) and rank of matrix A:
A = [ 1 2 -1 3 ]
[2 4 1 0]
[3 6 0 3]
Solution / Step-by-Step Proof:
Step 1: Row operations to reach REF:
R₂ ← R₂ - 2R₁: [0, 0, 3, -6]
R₃ ← R₃ - 3R₁: [0, 0, 3, -6]

Step 2: Eliminate R₃:

R₃ ← R₃ - R₂: [0, 0, 0, 0]

Step 3: Normalize pivot row 2:

R₂ ← (1/3) R₂: [0, 0, 1, -2]

Step 4: Eliminate above pivot in R₁:

R₁ ← R₁ + R₂: [1, 2, 0, 1]

Final RREF(A):
RREF = [ 1 2 0 1 ]
[ 0 0 1 -2 ]
 [0 0 0 0]

Rank Analysis: There are 2 non-zero pivot rows. Therefore, Rank(A) = 2, Nullity(A) = 4 - 2 = 2.

Example 4 (Medium-Hard): Eigenvalues, Eigenvectors, and Diagonalization

Problem Statement:
Find the eigenvalues and corresponding eigenvectors of matrix M = [[4, 1], [2, 3]], and express M in diagonalized form M
= P D P⁻¹.
Solution / Step-by-Step Proof:
Step 1: Characteristic Equation det(M - λI) = 0:
det([[4 - λ, 1], [2, 3 - λ]]) = (4 - λ)(3 - λ) - 2 = λ² - 7λ + 12 - 2 = λ² - 7λ + 10 = 0
(λ - 5)(λ - 2) = 0 => Eigenvalues: λ₁ = 5, λ₂ = 2.

Step 2: Find Eigenvector for λ₁ = 5:

(M - 5I) v = [[-1, 1], [2, -2]] [x, y]ᵀ = 0 => -x + y = 0 => x = y.
Eigenvector v₁ = [1, 1]ᵀ.

Step 3: Find Eigenvector for λ₂ = 2:

(M - 2I) v = [[2, 1], [2, 1]] [x, y]ᵀ = 0 => 2x + y = 0 => y = -2x.
Eigenvector v₂ = [1, -2]ᵀ.

Step 4: Form P, D, P⁻¹:

P = [[1, 1], [1, -2]], D = [[5, 0], [0, 2]]
det(P) = -2 - 1 = -3
P⁻¹ = (-1/3) [[-2, -1], [-1, 1]] = [[2/3, 1/3], [1/3, -1/3]]
Verification: M = P D P⁻¹ holds true.

Example 5 (Hard): Singular Value Decomposition (SVD) and PCA Application

Problem Statement:
Compute the SVD of matrix A = [[3, 0], [0, -2]] and explain how SVD is utilized for Dimensionality Reduction / Principal
Component Analysis (PCA).
Solution / Step-by-Step Proof:
Step 1: SVD formulation A = U Σ Vᵀ:
Aᵀ A = [[9, 0], [0, 4]]
Eigenvalues of Aᵀ A: λ₁ = 9, λ₂ = 4.
Singular values: σ₁ = √9 = 3, σ₂ = √4 = 2.
Diagonal matrix Σ = [[3, 0], [0, 2]].

Step 2: Right Singular Vectors (V):

Eigenvectors of Aᵀ A: v₁ = [1, 0]ᵀ, v₂ = [0, 1]ᵀ => V = [[1, 0], [0, 1]] = I₂.

Step 3: Left Singular Vectors (U):

u₁ = (1/σ₁) A v₁ = (1/3) [3, 0]ᵀ = [1, 0]ᵀ
u₂ = (1/σ₂) A v₂ = (1/2) [0, -2]ᵀ = [0, -1]ᵀ
U = [[1, 0], [0, -1]].
SVD Result: A = [[1, 0], [0, -1]] [[3, 0], [0, 2]] [[1, 0], [0, 1]]ᵀ.

PCA Connection:
In PCA, for zero-centered data matrix X (N × d), the right singular vectors V represent principal directions (eigenvectors
of covariance matrix XᵀX), and singular values σᵢ are proportional to variance explained along each principal axis.
Truncating SVD to top k components minimizes reconstruction MSE.

1.3 Tensor Products

Theory & Core Concepts

Tensors extend scalars, vectors, and matrices to arbitrary dimensions (orders/ranks), forming the mathematical
backbone of quantum entanglement state spaces and multidimensional deep learning representation tensors.

 Tensors & Order (Rank): Order/Rank 0 = Scalar (s ∈ ℝ); Order/Rank 1 = Vector (v ∈ ℝⁿ); Order/Rank 2 = Matrix (M ∈
ℝᵐⁿ); Order/Rank k = k-dimensional array (T ∈ ℝⁿ¹×ⁿ²×...×ⁿᵏ).
 Tensor Product (Kronecker Product): The Kronecker product A ⊗B for A (m × n) and B (p × q) produces a block
matrix of size (mp × nq):
A ⊗B = [[a₁₁B, a₁₂B, ...], [a₂₁B, a₂₂B, ...]]
 Mathematical Properties: Fundamental properties of Tensor Products:
• Bilinearity: A ⊗(B + C) = (A ⊗B) + (A ⊗C)
• Mixed-Product Property: (A ⊗B)(C ⊗D) = (AC) ⊗(BD)
• Transpose & Inverse: (A ⊗B)ᵀ = Aᵀ ⊗Bᵀ, (A ⊗B)⁻¹ = A⁻¹ ⊗B⁻¹
• Determinant & Trace: tr(A ⊗B) = tr(A) tr(B)
 Applications: Primary Applications across Modern Disciplines:
• Quantum Computing: Multi-qubit joint state spaces |Ψ⟩ = |q₁⟩ ⊗|q₂⟩ . Tensor product expands single qubit 2D
state space to 2ⁿ-dimensional Hilbert space.
• Deep Learning: CNN feature maps, multidimensional gradient propagation, attention mechanisms in
Transformers.
• Multilinear Algebra: Higher-order singular value decomposition (HOSVD), tensor factorization.
 Basic Operations: Tensor Operations:
• Contraction: Summing over repeated indices (generalizing matrix trace and dot product).
• Reshaping: Re-arranging tensor dimensions without altering element memory order.
• Broadcasting: Automatically expanding smaller tensor dimensions during element-wise operations.

Graded Worked Examples (Easy to Hard)

Example 1 (Easy): Vector Kronecker Product (Quantum Qubit State Construction)
Problem Statement:
In quantum computing, single qubit states |0⟩ = [1, 0]ᵀ and |1⟩ = [0, 1]ᵀ form the computational basis. Compute the joint
state vector for a 2-qubit system in state |01⟩ = |0⟩ ⊗|1⟩ .
Solution / Step-by-Step Proof:
State |0⟩ = [1, 0]ᵀ, State |1⟩ = [0, 1]ᵀ.

Compute Tensor Product |0⟩ ⊗|1⟩ :

|0⟩ ⊗|1⟩ = 1 · [0, 1]ᵀ
 0 · [0, 1]ᵀ
= [ 1·0, 1·1, 0·0, 0·1 ]ᵀ = [0, 1, 0, 0]ᵀ.

Interpretation: The resulting 4D vector represents the quantum state |01⟩ in standard basis {|00⟩ , |01⟩ , |10⟩ , |11⟩ }.

Example 2 (Easy-Medium): Matrix Kronecker Product (Quantum Gate Operations)

Problem Statement:
Given the Pauli-X gate X = [[0, 1], [1, 0]] and Identity gate I = [[1, 0], [0, 1]], compute the composite 2-qubit operator X ⊗
I.
Solution / Step-by-Step Proof:
X ⊗ I = [[ 0·I, 1·I ],
[ 1·I, 0·I ]]

Substituting matrix I = [[1, 0], [0, 1]]:

X ⊗I = [ 0 0 1 0 ]
[0 0 0 1]
[1 0 0 0]
[0 1 0 0]

Interpretation: This 4×4 operator applies the Pauli-X (NOT) gate to the first qubit while leaving the second qubit
unchanged.

Example 3 (Medium): Tensor Properties - Trace and Mixed-Product Property

Problem Statement:
Given A = [[1, 2], [3, 4]] and B = [[2, -1], [0, 3]]:
a) Compute tr(A), tr(B), and verify tr(A ⊗B) = tr(A) tr(B).
b) Verify the mixed-product property (A ⊗I)(I ⊗B) = A ⊗B.
Solution / Step-by-Step Proof:
a) Trace Verification:
tr(A) = 1 + 4 = 5
tr(B) = 2 + 3 = 5
tr(A) tr(B) = 5 × 5 = 25.

Now compute A ⊗ B:
A ⊗B = [[ 1·B, 2·B ],
[ 3·B, 4·B ]]
Diagonal elements of A ⊗B are:
(A ⊗B)₁₁ = 1·b₁₁ = 2
(A ⊗B)₂₂ = 1·b₂₂ = 3
(A ⊗B)₃₃ = 4·b₁₁ = 8
(A ⊗B)₄₄ = 4·b₂₂ = 12
tr(A ⊗B) = 2 + 3 + 8 + 12 = 25. (Verified!)

b) Mixed Product Property:

(A ⊗I)(I ⊗B) = (A·I) ⊗(I·B) = A ⊗B. (Verified!)
Example 4 (Medium-Hard): Deep Learning Tensor Reshaping & Contraction
Problem Statement:
A batch of image features in deep learning is represented as a 3rd-order tensor T ∈ ℝ²ˣ³ˣ⁴ (Batch=2, Channels=3,
Features=4).
a) Compute total elements in T.
b) Perform tensor contraction over the feature dimension (index 3) with weight vector w ∈ ℝ⁴.
c) Reshape T into a 2D matrix M for matrix multiplication in dense layers.
Solution / Step-by-Step Proof:
a) Total elements: N = 2 × 3 × 4 = 24 elements.

b) Tensor Contraction:
Contracting T_ijk with w_k along dimension k=1..4 yields output tensor Y ∈ ℝ²ˣ³:
Y_ij = ∑ₖ₌₁⁴ T_ijk · w_k
This reduces tensor order from 3 to 2 (matrix of dimensions 2 × 3).

c) Reshaping for Linear Layers:

Flatten spatial/channel dimensions: Combine (Channels=3, Features=4) into single feature vector of length 12.
New shape M ∈ ℝ²ˣ¹² where row 1 contains all 12 feature elements of batch item 1, and row 2 contains batch item 2.

Example 5 (Hard): Bell State Entanglement & Tensor Rank in Quantum Systems
Problem Statement:
Consider the Bell state qubit system |Ψ⁺⟩ = (1/√2) [ |00⟩ + |11⟩ ] = [ 1/√2, 0, 0, 1/√2 ]ᵀ.
Show that |Ψ⁺⟩ CANNOT be written as a simple tensor product |ϕa⟩ ⊗|ϕb⟩ of two single-qubit states (i.e., prove that
the Bell state is quantum entangled).
Solution / Step-by-Step Proof:
Let single-qubit states be |ϕa⟩ = [a₁, a₂]ᵀ and |ϕb⟩ = [b₁, b₂]ᵀ.
Their tensor product is:
|ϕa⟩ ⊗|ϕb⟩ = [ a₁b₁, a₁b₂, a₂b₁, a₂b₂ ]ᵀ.

Set this equal to state vector |Ψ⁺⟩ = [ 1/√2, 0, 0, 1/√2 ]ᵀ:

1) a₁b₁ = 1/√2 => a₁ ≠ 0 and b₁ ≠ 0
2) a₁b₂ = 0 => since a₁ ≠ 0, b₂ MUST be 0
3) a₂b₁ = 0 => since b₁ ≠ 0, a₂ MUST be 0
4) a₂b₂ = 1/√2 => substituting a₂=0 and b₂=0 yields (0)(0) = 0 ≠ 1/√2!

Contradiction:
No choice of single-qubit states |ϕa⟩ and |ϕb⟩ can satisfy the set of equations.
Thus, |Ψ⁺⟩ has Tensor Rank > 1 and CANNOT be decomposed into separable state tensor product. This rigorously proves
quantum entanglement!