Chapter 11: Tensor Products and Composite Systems — How Quantum Systems Combine

"I would not call [entanglement] one but rather the characteristic trait of quantum mechanics, the one that enforces its entire departure from classical lines of thought." — Erwin Schrödinger, 1935

Until now, every quantum system we have studied has lived alone. A single spin-1/2 particle. A single harmonic oscillator. A single particle in a box. We learned Dirac notation (Chapter 8), spectral decomposition (Chapter 9), and density matrices (Chapter 10) — all powerful tools, but all applied to individual systems. Nature, however, rarely hands us isolated quantum objects. Atoms contain multiple electrons. Photon pairs stream from nonlinear crystals. Qubits in a quantum computer must interact to compute anything useful.

This chapter answers a deceptively simple question: when two quantum systems come together, what is the mathematical structure of the whole? The answer — the tensor product — is straightforward in definition but staggering in consequence. It is the gateway to entanglement, a phenomenon with no classical counterpart, one that troubled Einstein, delighted Schrödinger, and now powers the most extraordinary technologies of the quantum information age.

11.1 When Two Quantum Systems Meet: The Tensor Product

11.1.1 The Classical Warm-Up

Before diving into quantum mechanics, let us recall how classical physics handles composite systems. If system $A$ is described by a state in some configuration space of dimension $d_A$, and system $B$ by a state in a space of dimension $d_B$, the composite system $AB$ lives in a space of dimension $d_A + d_B$. A pair of classical coins (each heads or tails) has $2 + 2 = 4$ coordinates... wait, that does not sound right. Actually, a pair of classical coins has $2 \times 2 = 4$ possible states (HH, HT, TH, TT), but the state of the pair is always fully determined by stating the state of each coin separately. There is no "composite coin state" that cannot be decomposed into "state of coin 1" and "state of coin 2."

Quantum mechanics breaks this rule spectacularly.

11.1.2 The Tensor Product: Definition

Definition 11.1 (Tensor Product of Hilbert Spaces). Let $\mathcal{H}_A$ be a Hilbert space of dimension $d_A$ and $\mathcal{H}_B$ a Hilbert space of dimension $d_B$. The tensor product $\mathcal{H}_A \otimes \mathcal{H}_B$ is a Hilbert space of dimension $d_A \cdot d_B$ equipped with a bilinear map

$$\otimes : \mathcal{H}_A \times \mathcal{H}_B \to \mathcal{H}_A \otimes \mathcal{H}_B$$

satisfying the following axioms:

1. Linearity in each factor:

$$(\alpha |\psi_1\rangle + \beta |\psi_2\rangle) \otimes |\phi\rangle = \alpha (|\psi_1\rangle \otimes |\phi\rangle) + \beta (|\psi_2\rangle \otimes |\phi\rangle)$$

$$|\psi\rangle \otimes (\alpha |\phi_1\rangle + \beta |\phi_2\rangle) = \alpha (|\psi\rangle \otimes |\phi_1\rangle) + \beta (|\psi\rangle \otimes |\phi_2\rangle)$$

2. Inner product:

$$(\langle \psi_1 | \otimes \langle \phi_1 |)(|\psi_2\rangle \otimes |\phi_2\rangle) = \langle \psi_1 | \psi_2 \rangle \cdot \langle \phi_1 | \phi_2 \rangle$$

The inner product on the tensor product space is determined by the inner products on the factor spaces.

3. Spanning: The set of all product vectors $|\psi\rangle \otimes |\phi\rangle$ spans $\mathcal{H}_A \otimes \mathcal{H}_B$.

Notation. We write $|\psi\rangle \otimes |\phi\rangle$, or more compactly $|\psi\rangle|\phi\rangle$, or $|\psi, \phi\rangle$, or even $|\psi \phi\rangle$ when the meaning is clear. All four notations mean the same thing.

💡 Key Insight: The dimension of the composite space is $d_A \cdot d_B$, not $d_A + d_B$. This multiplicative growth is the source of both quantum computing's power and entanglement's mystery.

11.1.3 Concrete Example: Two Qubits

Let us make this concrete with the system that will anchor the entire chapter: two spin-1/2 particles.

System $A$ (the first qubit) has Hilbert space $\mathcal{H}_A = \text{span}\{|0\rangle_A, |1\rangle_A\}$, dimension 2.

System $B$ (the second qubit) has Hilbert space $\mathcal{H}_B = \text{span}\{|0\rangle_B, |1\rangle_B\}$, dimension 2.

The composite space is $\mathcal{H}_{AB} = \mathcal{H}_A \otimes \mathcal{H}_B$, dimension $2 \times 2 = 4$.

A general state in this space is:

$$|\Psi\rangle_{AB} = \alpha_{00}|00\rangle + \alpha_{01}|01\rangle + \alpha_{10}|10\rangle + \alpha_{11}|11\rangle$$

where $|\alpha_{00}|^2 + |\alpha_{01}|^2 + |\alpha_{10}|^2 + |\alpha_{11}|^2 = 1$.

Here $|00\rangle \equiv |0\rangle_A \otimes |0\rangle_B$, and similarly for the other basis states.

🔄 Retrieval Practice (from Chapter 8): Write out the bra corresponding to $|\Psi\rangle_{AB}$ using Dirac notation. Verify that $\langle \Psi | \Psi \rangle = 1$ using the inner product rule from Definition 11.1.

11.2 Building the Composite Hilbert Space: $|\psi\rangle \otimes |\phi\rangle$

11.2.1 Product States

The simplest states in a composite system are product states (also called separable pure states):

$$|\Psi\rangle_{AB} = |\psi\rangle_A \otimes |\phi\rangle_B$$

Such a state says: "system $A$ is in state $|\psi\rangle$ and system $B$ is in state $|\phi\rangle$, independently." There is nothing quantum-mechanically exotic about product states — they behave like classical composite states in the sense that the whole is fully determined by its parts.

Example 11.1. Suppose qubit $A$ is in the state $|+\rangle_A = \frac{1}{\sqrt{2}}(|0\rangle + |1\rangle)_A$ and qubit $B$ is in $|0\rangle_B$. The composite state is:

$$|+\rangle_A \otimes |0\rangle_B = \frac{1}{\sqrt{2}}(|0\rangle_A + |1\rangle_A) \otimes |0\rangle_B = \frac{1}{\sqrt{2}}(|00\rangle + |10\rangle)$$

This is a product state. It can be "factored" as a tensor product of a state of $A$ alone and a state of $B$ alone.

11.2.2 Matrix Representation: The Kronecker Product

When we represent states as column vectors and operators as matrices, the tensor product becomes the Kronecker product.

Definition 11.2 (Kronecker Product). If $A$ is an $m \times n$ matrix and $B$ is a $p \times q$ matrix, their Kronecker product $A \otimes B$ is the $mp \times nq$ block matrix:

$$A \otimes B = \begin{pmatrix} a_{11}B & a_{12}B & \cdots & a_{1n}B \\ a_{21}B & a_{22}B & \cdots & a_{2n}B \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1}B & a_{m2}B & \cdots & a_{mn}B \end{pmatrix}$$

Example 11.2. The state $|+\rangle \otimes |0\rangle$ in matrix form:

$$|+\rangle \otimes |0\rangle = \frac{1}{\sqrt{2}}\begin{pmatrix}1\\1\end{pmatrix} \otimes \begin{pmatrix}1\\0\end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix}1 \cdot \begin{pmatrix}1\\0\end{pmatrix} \\ 1 \cdot \begin{pmatrix}1\\0\end{pmatrix}\end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix}1\\0\\1\\0\end{pmatrix}$$

11.2.3 Operators on Composite Systems

Operators on the composite space take two forms:

Product operators: If $\hat{O}_A$ acts on $\mathcal{H}_A$ and $\hat{O}_B$ acts on $\mathcal{H}_B$, then $\hat{O}_A \otimes \hat{O}_B$ acts on $\mathcal{H}_A \otimes \mathcal{H}_B$:

$$(\hat{O}_A \otimes \hat{O}_B)(|\psi\rangle \otimes |\phi\rangle) = (\hat{O}_A|\psi\rangle) \otimes (\hat{O}_B|\phi\rangle)$$

When an operator acts only on one subsystem, we tensor with the identity on the other:

$$\hat{O}_A \text{ acting on AB} \equiv \hat{O}_A \otimes \hat{I}_B$$

Example 11.3. The operator $\hat{\sigma}_z \otimes \hat{I}$ measures the $z$-component of spin for particle $A$ while leaving particle $B$ alone:

$$(\hat{\sigma}_z \otimes \hat{I})|01\rangle = (\hat{\sigma}_z|0\rangle) \otimes (\hat{I}|1\rangle) = (+1)|0\rangle \otimes |1\rangle = |01\rangle$$

$$(\hat{\sigma}_z \otimes \hat{I})|11\rangle = (\hat{\sigma}_z|1\rangle) \otimes (\hat{I}|1\rangle) = (-1)|1\rangle \otimes |1\rangle = -|11\rangle$$

General operators on $\mathcal{H}_A \otimes \mathcal{H}_B$ need not be product operators. Any linear operator on the composite space can be written as a sum of product operators:

$$\hat{O} = \sum_{ij} c_{ij} \hat{A}_i \otimes \hat{B}_j$$

This fact is crucial for describing interactions between subsystems.

🔄 Spaced Review (from Chapter 9): Recall spectral decomposition: $\hat{O} = \sum_n \lambda_n |n\rangle\langle n|$. How does this generalize for product operators $\hat{O}_A \otimes \hat{O}_B$? Write the spectral decomposition of $\hat{\sigma}_z \otimes \hat{\sigma}_z$ in the computational basis.

11.3 Bases for Composite Systems

11.3.1 The Computational (Product) Basis

If $\{|i\rangle_A\}_{i=1}^{d_A}$ is an orthonormal basis for $\mathcal{H}_A$ and $\{|j\rangle_B\}_{j=1}^{d_B}$ is an orthonormal basis for $\mathcal{H}_B$, then

$$\{|i\rangle_A \otimes |j\rangle_B\}_{i=1,\ldots,d_A;\; j=1,\ldots,d_B}$$

is an orthonormal basis for $\mathcal{H}_A \otimes \mathcal{H}_B$. This is called the computational basis or product basis.

Proof of orthonormality:

$$(\langle i' |_A \otimes \langle j'|_B)(|i\rangle_A \otimes |j\rangle_B) = \langle i'|i\rangle_A \cdot \langle j'|j\rangle_B = \delta_{i'i}\delta_{j'j}$$

Proof of completeness: These $d_A \cdot d_B$ orthonormal vectors span a space of dimension $d_A \cdot d_B$, which is the dimension of $\mathcal{H}_A \otimes \mathcal{H}_B$ by definition.

11.3.2 Two Qubits: Explicit Basis

For two qubits, the computational basis is:

$$|00\rangle = \begin{pmatrix}1\\0\\0\\0\end{pmatrix}, \quad |01\rangle = \begin{pmatrix}0\\1\\0\\0\end{pmatrix}, \quad |10\rangle = \begin{pmatrix}0\\0\\1\\0\end{pmatrix}, \quad |11\rangle = \begin{pmatrix}0\\0\\0\\1\end{pmatrix}$$

Any two-qubit state can be expanded:

$$|\Psi\rangle = \sum_{i,j \in \{0,1\}} \alpha_{ij}|ij\rangle$$

where the $\alpha_{ij}$ are complex amplitudes satisfying normalization.

11.3.3 Higher-Dimensional Examples

For a qubit ($d_A = 2$) tensored with a qutrit ($d_B = 3$), the composite space has dimension 6, with basis:

$$\{|0,0\rangle, |0,1\rangle, |0,2\rangle, |1,0\rangle, |1,1\rangle, |1,2\rangle\}$$

For $n$ qubits, the composite space $(\mathbb{C}^2)^{\otimes n}$ has dimension $2^n$. This exponential growth is why simulating quantum systems on classical computers is hard — and why quantum computers are potentially powerful.

⚠️ Important: The product basis is just one choice of basis for the composite space. Other bases — notably the Bell basis we will meet in Section 11.8 — consist entirely of entangled states. The choice of basis does not change the physics, but it can dramatically simplify calculations.

11.3.4 The Coefficient Matrix

A useful way to organize the amplitudes of a bipartite state is the coefficient matrix. For $|\Psi\rangle = \sum_{ij}\alpha_{ij}|i\rangle_A|j\rangle_B$, define the $d_A \times d_B$ matrix $C$ with entries $C_{ij} = \alpha_{ij}$.

For two qubits:

$$C = \begin{pmatrix}\alpha_{00} & \alpha_{01} \\ \alpha_{10} & \alpha_{11}\end{pmatrix}$$

This matrix is the key to the Schmidt decomposition (Section 11.6) and to distinguishing entangled from separable states (Section 11.5).

Theorem 11.1. A bipartite pure state $|\Psi\rangle$ is a product state if and only if its coefficient matrix $C$ has rank 1.

Proof. If $|\Psi\rangle = |\psi\rangle_A \otimes |\phi\rangle_B$ with $|\psi\rangle = \sum_i a_i |i\rangle$ and $|\phi\rangle = \sum_j b_j |j\rangle$, then $C_{ij} = a_i b_j$, so $C = \mathbf{a}\mathbf{b}^T$, which has rank 1. Conversely, if $\text{rank}(C) = 1$, then $C = \mathbf{a}\mathbf{b}^T$ for some vectors $\mathbf{a}$, $\mathbf{b}$, which means $|\Psi\rangle = |\psi\rangle_A \otimes |\phi\rangle_B$. $\square$

11.4 Entanglement: The Genuinely New Feature

11.4.1 A State That Cannot Be Factored

Consider the following two-qubit state:

$$|\Phi^+\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle)$$

Can we write this as a product $|\psi\rangle_A \otimes |\phi\rangle_B$? Let us try:

$$|\psi\rangle_A \otimes |\phi\rangle_B = (a|0\rangle + b|1\rangle) \otimes (c|0\rangle + d|1\rangle) = ac|00\rangle + ad|01\rangle + bc|10\rangle + bd|11\rangle$$

Matching coefficients with $|\Phi^+\rangle$:

$$ac = \frac{1}{\sqrt{2}}, \quad ad = 0, \quad bc = 0, \quad bd = \frac{1}{\sqrt{2}}$$

From $ad = 0$: either $a = 0$ or $d = 0$. But if $a = 0$, then $ac = 0 \neq \frac{1}{\sqrt{2}}$. If $d = 0$, then $bd = 0 \neq \frac{1}{\sqrt{2}}$. Contradiction. The state $|\Phi^+\rangle$ cannot be written as a product state.

Definition 11.3 (Entangled State). A state $|\Psi\rangle \in \mathcal{H}_A \otimes \mathcal{H}_B$ is entangled if it cannot be written as a product state $|\psi\rangle_A \otimes |\phi\rangle_B$. Otherwise, it is separable (or a product state).

11.4.2 What Entanglement Means Physically

When two particles are in the entangled state $|\Phi^+\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle)$:

• Neither particle has a definite state of its own. The state of the pair is perfectly well-defined (it is a pure state of the composite system), but the individual particles do not have pure states. Each particle, considered alone, is in a mixed state (as we will make precise via the partial trace in Section 11.7).

• Measurement outcomes are perfectly correlated. If you measure particle $A$ in the $\{|0\rangle, |1\rangle\}$ basis and get outcome 0, particle $B$ instantaneously "collapses" to $|0\rangle$ — regardless of the spatial separation between the particles. If you get outcome 1, particle $B$ collapses to $|1\rangle$.

• This is not communication. The outcomes are random (50/50 for each), so no information is transmitted faster than light. The correlations are only revealed when the two experimenters compare their results.

📊 By the Numbers: A pair of qubits lives in a 4-dimensional Hilbert space. Of those states, the separable ones form a set of measure zero — "almost all" two-qubit states are entangled. Entanglement is the rule, not the exception.

11.4.3 Entanglement Has No Classical Analogue

A common but misleading analogy: "It is like putting one red sock and one blue sock in separate boxes — when you open one and see red, you know the other is blue." This analogy captures correlation but misses the essence of entanglement. In the sock scenario, each sock had a definite color all along — you just did not know which box held which. In quantum mechanics, the particles genuinely do not have definite individual states before measurement. This distinction (hidden variables vs. genuine indeterminacy) was the crux of the EPR debate, which we will preview in Section 11.9.

11.5 Separable vs. Entangled States: How to Tell

11.5.1 The Factoring Test

For pure bipartite states, the most direct test is to attempt factoring. Write $|\Psi\rangle = \sum_{ij}\alpha_{ij}|ij\rangle$ and form the coefficient matrix $C_{ij} = \alpha_{ij}$. By Theorem 11.1:

• $\text{rank}(C) = 1$ $\Rightarrow$ separable
• $\text{rank}(C) > 1$ $\Rightarrow$ entangled

Example 11.4. Test the state $|\Psi\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle)$.

$$C = \frac{1}{\sqrt{2}}\begin{pmatrix}1 & 0 \\ 0 & 1\end{pmatrix}$$

$\det(C) = \frac{1}{2} \neq 0$, so $\text{rank}(C) = 2$. The state is entangled. $\checkmark$

Example 11.5. Test $|\Psi\rangle = \frac{1}{2}(|00\rangle + |01\rangle + |10\rangle + |11\rangle)$.

$$C = \frac{1}{2}\begin{pmatrix}1 & 1 \\ 1 & 1\end{pmatrix}$$

$\det(C) = \frac{1}{4}(1-1) = 0$, so $\text{rank}(C) = 1$. The state is separable. Indeed, $|\Psi\rangle = |+\rangle|+\rangle$. $\checkmark$

11.5.2 The Entropy Test (Preview)

A more quantitative measure of entanglement comes from the von Neumann entropy of the reduced density matrix (Section 11.7). For a pure bipartite state $|\Psi\rangle_{AB}$:

$$S(\hat{\rho}_A) = -\text{Tr}(\hat{\rho}_A \log_2 \hat{\rho}_A)$$

where $\hat{\rho}_A = \text{Tr}_B(|\Psi\rangle\langle\Psi|)$ is the reduced density matrix of subsystem $A$.

• $S(\hat{\rho}_A) = 0$ $\Leftrightarrow$ separable
• $S(\hat{\rho}_A) > 0$ $\Leftrightarrow$ entangled
• $S(\hat{\rho}_A) = \log_2 d$ (where $d = \min(d_A, d_B)$) $\Leftrightarrow$ maximally entangled

11.5.3 Mixed State Separability (Brief Mention)

For mixed states, the separability question is far more subtle. A mixed state $\hat{\rho}_{AB}$ is separable if it can be written as a convex combination of product states:

$$\hat{\rho}_{AB} = \sum_k p_k \hat{\rho}_A^{(k)} \otimes \hat{\rho}_B^{(k)}$$

Determining whether a given mixed state is separable is, in general, NP-hard. For two qubits, the Peres-Horodecki criterion (positive partial transpose) provides a necessary and sufficient condition — a topic we revisit in a later chapter.

💡 Conceptual Checkpoint: Why can we not simply check whether $\hat{\rho}_{AB}$ has rank 1 to determine entanglement of mixed states? Because a mixed separable state can have rank > 1 (it is a mixture of product states, each of rank 1 individually). The rank test works only for pure states.

11.6 The Schmidt Decomposition

11.6.1 Statement

The Schmidt decomposition is one of the most important tools in quantum information theory. It provides a canonical form for any bipartite pure state.

Theorem 11.2 (Schmidt Decomposition). For any pure state $|\Psi\rangle \in \mathcal{H}_A \otimes \mathcal{H}_B$, there exist orthonormal bases $\{|a_k\rangle\}$ for $\mathcal{H}_A$ and $\{|b_k\rangle\}$ for $\mathcal{H}_B$ such that:

$$|\Psi\rangle = \sum_{k=1}^{r} \lambda_k |a_k\rangle |b_k\rangle$$

where: - $r \leq \min(d_A, d_B)$ is the Schmidt rank - $\lambda_k > 0$ are the Schmidt coefficients satisfying $\sum_k \lambda_k^2 = 1$ - The bases $\{|a_k\rangle\}$ and $\{|b_k\rangle\}$ are called the Schmidt bases

Key properties: - The decomposition is a single sum, not a double sum — this is what makes it powerful - The Schmidt coefficients $\lambda_k$ are unique (up to ordering), though the bases are not unique when coefficients are degenerate - $r = 1$ $\Leftrightarrow$ separable; $r > 1$ $\Leftrightarrow$ entangled

11.6.2 Proof via Singular Value Decomposition

The proof connects beautifully to linear algebra. Form the coefficient matrix $C_{ij} = \alpha_{ij}$. By the singular value decomposition (SVD):

$$C = U \Sigma V^\dagger$$

where $U$ is $d_A \times d_A$ unitary, $V$ is $d_B \times d_B$ unitary, and $\Sigma$ is $d_A \times d_B$ with non-negative real entries $\sigma_k$ on the diagonal and zeros elsewhere.

Define new bases: $$|a_k\rangle = \sum_i U_{ik} |i\rangle_A, \quad |b_k\rangle = \sum_j V_{jk}^* |j\rangle_B$$

Then: $$|\Psi\rangle = \sum_{ij} C_{ij} |i\rangle|j\rangle = \sum_{ij}\sum_k U_{ik}\sigma_k V_{jk}^*|i\rangle|j\rangle = \sum_k \sigma_k \left(\sum_i U_{ik}|i\rangle\right)\left(\sum_j V_{jk}^*|j\rangle\right) = \sum_k \sigma_k |a_k\rangle|b_k\rangle$$

Setting $\lambda_k = \sigma_k$ (the singular values of $C$), we obtain the Schmidt decomposition. The orthonormality of $\{|a_k\rangle\}$ and $\{|b_k\rangle\}$ follows from the unitarity of $U$ and $V$. $\square$

11.6.3 Worked Example: Schmidt Decomposition of a Two-Qubit State

Problem: Find the Schmidt decomposition of $|\Psi\rangle = \frac{1}{\sqrt{3}}|00\rangle + \frac{1}{\sqrt{6}}|01\rangle + \frac{1}{\sqrt{6}}|10\rangle + \frac{1}{\sqrt{3}}|11\rangle$.

Step 1: Form the coefficient matrix.

$$C = \begin{pmatrix} 1/\sqrt{3} & 1/\sqrt{6} \\ 1/\sqrt{6} & 1/\sqrt{3}\end{pmatrix}$$

Step 2: Compute $CC^\dagger$.

Since $C$ is real and symmetric, $CC^\dagger = C^2$:

$$C^2 = \begin{pmatrix} 1/3 + 1/6 & 1/\sqrt{18} + 1/\sqrt{18} \\ 1/\sqrt{18} + 1/\sqrt{18} & 1/6 + 1/3 \end{pmatrix} = \begin{pmatrix} 1/2 & 2/\sqrt{18} \\ 2/\sqrt{18} & 1/2\end{pmatrix} = \begin{pmatrix} 1/2 & \sqrt{2}/3 \\ \sqrt{2}/3 & 1/2 \end{pmatrix}$$

Step 3: Find eigenvalues of $CC^\dagger$.

$$\det(CC^\dagger - \mu I) = (1/2 - \mu)^2 - 2/9 = 0$$

$$\mu = \frac{1}{2} \pm \frac{\sqrt{2}}{3}$$

So $\lambda_1^2 = \frac{1}{2} + \frac{\sqrt{2}}{3}$ and $\lambda_2^2 = \frac{1}{2} - \frac{\sqrt{2}}{3}$.

The Schmidt coefficients are $\lambda_1 = \sqrt{\frac{1}{2} + \frac{\sqrt{2}}{3}} \approx 0.888$ and $\lambda_2 = \sqrt{\frac{1}{2} - \frac{\sqrt{2}}{3}} \approx 0.460$.

Step 4: Find Schmidt bases from the eigenvectors of $CC^\dagger$ and $C^\dagger C$.

The eigenvectors of $CC^\dagger$ (which is symmetric with off-diagonal $\sqrt{2}/3 > 0$) are:

$$|a_1\rangle = \frac{1}{\sqrt{2}}(|0\rangle + |1\rangle), \quad |a_2\rangle = \frac{1}{\sqrt{2}}(|0\rangle - |1\rangle)$$

By symmetry ($C = C^T$ in this case), $|b_k\rangle = |a_k\rangle$, so:

$$|\Psi\rangle = \lambda_1 |+\rangle_A|+\rangle_B + \lambda_2 |-\rangle_A|-\rangle_B$$

Verification: The Schmidt rank is 2 (both coefficients are nonzero), confirming that $|\Psi\rangle$ is entangled.

🔗 Connection to Earlier Material: The SVD connection means that everything you know about singular values from linear algebra transfers directly to quantum entanglement theory. The Schmidt coefficients are the singular values of the coefficient matrix.

11.6.4 Schmidt Number as an Entanglement Measure

The Schmidt rank (or Schmidt number) $r$ provides a coarse-grained measure of entanglement:

• $r = 1$: No entanglement (product state)
• $r = \min(d_A, d_B)$ with all $\lambda_k$ equal: Maximum entanglement
• Intermediate values: Partial entanglement

A more refined measure uses the Schmidt coefficients themselves. The entanglement entropy is:

$$E(|\Psi\rangle) = -\sum_k \lambda_k^2 \log_2(\lambda_k^2)$$

This equals the von Neumann entropy of the reduced density matrix, as we show next.

11.7 Reduced Density Matrices and the Partial Trace

11.7.1 The Problem: Describing a Subsystem

When a composite system $AB$ is in a pure entangled state $|\Psi\rangle_{AB}$, how do we describe subsystem $A$ alone? We cannot assign it a pure state (that is the whole point of entanglement). The answer, foreshadowed in Chapter 10, is the density matrix.

🔄 Retrieval Practice (from Chapter 10): Before reading on, recall: what is the density matrix of a system in a pure state $|\psi\rangle$? What is it for a mixed state? What physical quantity determines whether $\hat{\rho}$ describes a pure or mixed state?

11.7.2 Definition of the Partial Trace

Definition 11.4 (Partial Trace). The partial trace over $B$ is the linear map $\text{Tr}_B : \mathcal{L}(\mathcal{H}_A \otimes \mathcal{H}_B) \to \mathcal{L}(\mathcal{H}_A)$ defined on product operators by:

$$\text{Tr}_B(|a_1\rangle\langle a_2| \otimes |b_1\rangle\langle b_2|) = |a_1\rangle\langle a_2| \cdot \langle b_2|b_1\rangle$$

and extended to all operators by linearity.

Equivalently, using an orthonormal basis $\{|j\rangle_B\}$ of $\mathcal{H}_B$:

$$\text{Tr}_B(\hat{O}_{AB}) = \sum_j \langle j|_B \hat{O}_{AB} |j\rangle_B$$

The result is an operator on $\mathcal{H}_A$ alone.

Definition 11.5 (Reduced Density Matrix). The reduced density matrix of subsystem $A$ is:

$$\hat{\rho}_A = \text{Tr}_B(\hat{\rho}_{AB}) = \text{Tr}_B(|\Psi\rangle_{AB}\langle\Psi|)$$

11.7.3 Explicit Calculation for Two Qubits

Example 11.6. Compute $\hat{\rho}_A$ for $|\Phi^+\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle)$.

First, form the full density matrix:

$$\hat{\rho}_{AB} = |\Phi^+\rangle\langle\Phi^+| = \frac{1}{2}(|00\rangle + |11\rangle)(\langle 00| + \langle 11|)$$

$$= \frac{1}{2}(|00\rangle\langle 00| + |00\rangle\langle 11| + |11\rangle\langle 00| + |11\rangle\langle 11|)$$

Now trace over $B$:

$$\hat{\rho}_A = \text{Tr}_B(\hat{\rho}_{AB}) = \sum_{j \in \{0,1\}} \langle j|_B \hat{\rho}_{AB} |j\rangle_B$$

For $j = 0$: $$\langle 0|_B \hat{\rho}_{AB} |0\rangle_B = \frac{1}{2}(\langle 0|0\rangle\langle 0|0\rangle |0\rangle_A\langle 0|_A + \langle 0|0\rangle\langle 1|0\rangle |0\rangle_A\langle 1|_A + \langle 0|1\rangle\langle 0|0\rangle |1\rangle_A\langle 0|_A + \langle 0|1\rangle\langle 1|0\rangle |1\rangle_A\langle 1|_A)$$ $$= \frac{1}{2}|0\rangle_A\langle 0|_A$$

For $j = 1$: $$\langle 1|_B \hat{\rho}_{AB} |1\rangle_B = \frac{1}{2}|1\rangle_A\langle 1|_A$$

Therefore:

$$\hat{\rho}_A = \frac{1}{2}|0\rangle\langle 0| + \frac{1}{2}|1\rangle\langle 1| = \frac{1}{2}\hat{I}$$

This is the maximally mixed state — complete ignorance about subsystem $A$, even though the composite system is in a perfectly well-defined pure state. This is the signature of maximal entanglement.

11.7.4 The Partial Trace and Schmidt Decomposition

The connection between the partial trace and the Schmidt decomposition is elegant. If $|\Psi\rangle = \sum_k \lambda_k |a_k\rangle|b_k\rangle$, then:

$$\hat{\rho}_A = \text{Tr}_B(|\Psi\rangle\langle\Psi|) = \sum_k \lambda_k^2 |a_k\rangle\langle a_k|$$

$$\hat{\rho}_B = \text{Tr}_A(|\Psi\rangle\langle\Psi|) = \sum_k \lambda_k^2 |b_k\rangle\langle b_k|$$

Proof. Starting from $|\Psi\rangle\langle\Psi| = \sum_{k,k'}\lambda_k\lambda_{k'}|a_k\rangle\langle a_{k'}| \otimes |b_k\rangle\langle b_{k'}|$, take the partial trace over $B$:

$$\hat{\rho}_A = \sum_{k,k'}\lambda_k\lambda_{k'}|a_k\rangle\langle a_{k'}|\cdot\langle b_{k'}|b_k\rangle = \sum_{k,k'}\lambda_k\lambda_{k'}|a_k\rangle\langle a_{k'}|\cdot\delta_{k'k} = \sum_k \lambda_k^2 |a_k\rangle\langle a_k|$$

using the orthonormality of the Schmidt basis $\{|b_k\rangle\}$. $\square$

Corollary 11.1. The reduced density matrices $\hat{\rho}_A$ and $\hat{\rho}_B$ have the same nonzero eigenvalues $\{\lambda_k^2\}$. Consequently: - $S(\hat{\rho}_A) = S(\hat{\rho}_B)$ — the entanglement entropy is the same whether computed from $A$ or $B$ - $\hat{\rho}_A$ is pure $\Leftrightarrow$ $\hat{\rho}_B$ is pure $\Leftrightarrow$ $|\Psi\rangle_{AB}$ is a product state

11.7.5 Physical Meaning of the Partial Trace

Why is the partial trace the correct way to describe a subsystem? Because it preserves expectation values. For any observable $\hat{O}_A$ acting only on subsystem $A$:

$$\langle \hat{O}_A \rangle = \text{Tr}_{AB}(\hat{\rho}_{AB} (\hat{O}_A \otimes \hat{I}_B)) = \text{Tr}_A(\hat{\rho}_A \hat{O}_A)$$

The partial trace is the unique operation that makes local predictions consistent with the global state. It is the quantum analogue of marginalizing a joint probability distribution.

⚠️ Common Misconception: Students sometimes think that if $\hat{\rho}_A$ is mixed, the composite system must also be in a mixed state. This is false! A pure entangled state of $AB$ produces mixed reduced states for $A$ and $B$ individually. The "mixedness" comes from entanglement, not from classical ignorance.

11.8 Bell States: The Maximally Entangled Basis

11.8.1 The Four Bell States

The Bell states (or EPR pairs) form an orthonormal basis for the two-qubit Hilbert space, and every element of this basis is maximally entangled:

$$|\Phi^+\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle)$$

$$|\Phi^-\rangle = \frac{1}{\sqrt{2}}(|00\rangle - |11\rangle)$$

$$|\Psi^+\rangle = \frac{1}{\sqrt{2}}(|01\rangle + |10\rangle)$$

$$|\Psi^-\rangle = \frac{1}{\sqrt{2}}(|01\rangle - |10\rangle)$$

11.8.2 Verification of Key Properties

Orthonormality. Let us verify $\langle\Phi^+|\Psi^+\rangle = 0$:

$$\langle\Phi^+|\Psi^+\rangle = \frac{1}{2}(\langle 00| + \langle 11|)(|01\rangle + |10\rangle) = \frac{1}{2}(\langle 00|01\rangle + \langle 00|10\rangle + \langle 11|01\rangle + \langle 11|10\rangle) = \frac{1}{2}(0 + 0 + 0 + 0) = 0$$

The remaining orthogonality relations follow similarly.

Completeness. Any two-qubit state $|\Psi\rangle = \alpha|00\rangle + \beta|01\rangle + \gamma|10\rangle + \delta|11\rangle$ can be expanded in the Bell basis:

$$|\Psi\rangle = \frac{\alpha + \delta}{\sqrt{2}}|\Phi^+\rangle + \frac{\alpha - \delta}{\sqrt{2}}|\Phi^-\rangle + \frac{\beta + \gamma}{\sqrt{2}}|\Psi^+\rangle + \frac{\beta - \gamma}{\sqrt{2}}|\Psi^-\rangle$$

(Verify this by substituting the Bell state definitions and collecting terms.)

Maximal entanglement. For each Bell state, the coefficient matrix has rank 2, and the Schmidt coefficients are $\lambda_1 = \lambda_2 = \frac{1}{\sqrt{2}}$. The reduced density matrix is $\hat{\rho}_A = \frac{1}{2}\hat{I}$, and the entanglement entropy is $S = \log_2 2 = 1$ ebit.

11.8.3 Generating Bell States from Product States

Bell states can be created from product states using two quantum gates: the Hadamard gate $H$ and the CNOT gate:

$$H = \frac{1}{\sqrt{2}}\begin{pmatrix}1 & 1 \\ 1 & -1\end{pmatrix}, \quad \text{CNOT} = \begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&0&1\\0&0&1&0\end{pmatrix}$$

The circuit $\text{CNOT} \cdot (H \otimes I)$ maps:

$$|00\rangle \to |\Phi^+\rangle, \quad |01\rangle \to |\Psi^+\rangle, \quad |10\rangle \to |\Phi^-\rangle, \quad |11\rangle \to |\Psi^-\rangle$$

Derivation for $|00\rangle \to |\Phi^+\rangle$:

Step 1: Apply $H \otimes I$ to $|00\rangle$:

$$(H \otimes I)|00\rangle = H|0\rangle \otimes I|0\rangle = \frac{1}{\sqrt{2}}(|0\rangle + |1\rangle) \otimes |0\rangle = \frac{1}{\sqrt{2}}(|00\rangle + |10\rangle)$$

Step 2: Apply CNOT (flips the second qubit if the first is $|1\rangle$):

$$\text{CNOT}\left[\frac{1}{\sqrt{2}}(|00\rangle + |10\rangle)\right] = \frac{1}{\sqrt{2}}(|00\rangle + |11\rangle) = |\Phi^+\rangle$$

This simple two-gate circuit is one of the most important in all of quantum computing.

11.8.4 Bell States and the Pauli Matrices

An elegant relationship connects the four Bell states via local Pauli operations on one qubit:

$$|\Phi^+\rangle = (\hat{I} \otimes \hat{I})|\Phi^+\rangle$$ $$|\Phi^-\rangle = (\hat{\sigma}_z \otimes \hat{I})|\Phi^+\rangle$$ $$|\Psi^+\rangle = (\hat{\sigma}_x \otimes \hat{I})|\Phi^+\rangle$$ $$|\Psi^-\rangle = (i\hat{\sigma}_y \otimes \hat{I})|\Phi^+\rangle$$

So all four Bell states are related by local unitary operations on one qubit. This means they all have the same amount of entanglement — they differ only in the relative phase and which basis states are correlated.

11.8.5 The Singlet State and Total Spin

The state $|\Psi^-\rangle = \frac{1}{\sqrt{2}}(|01\rangle - |10\rangle)$ has special physical significance. Using the spin notation $|0\rangle = |\!\uparrow\rangle$ and $|1\rangle = |\!\downarrow\rangle$:

$$|\Psi^-\rangle = \frac{1}{\sqrt{2}}(|\!\uparrow\downarrow\rangle - |\!\downarrow\uparrow\rangle)$$

This is the singlet state with total spin $S = 0$. It is the unique state that is rotationally invariant: it looks the same regardless of which direction you choose for the spin quantization axis. The other three Bell states, together with $|\Psi^-\rangle$, span the triplet ($S = 1$) subspace... actually, no — $|\Phi^+\rangle$, $|\Phi^-\rangle$, and $|\Psi^+\rangle$ are the $m_s = +1, -1, 0$ components of the spin-1 triplet (in a different form than the standard $|S, m_s\rangle$ eigenstates, but spanning the same subspace).

The singlet-triplet decomposition is:

$$\mathcal{H}_A \otimes \mathcal{H}_B = \underbrace{\text{span}\{|\Psi^-\rangle\}}_{S=0\text{ (singlet)}} \oplus \underbrace{\text{span}\{|\Phi^+\rangle, |\Phi^-\rangle, |\Psi^+\rangle\}}_{S=1\text{ (triplet)}}$$

This decomposition, $\frac{1}{2} \otimes \frac{1}{2} = 0 \oplus 1$, is the simplest example of the addition of angular momenta, which we treat in full generality in a later chapter.

11.9 Why Entanglement Confused Einstein (EPR Preview)

11.9.1 The EPR Argument (1935)

In their landmark paper, Einstein, Podolsky, and Rosen (EPR) considered a state equivalent to one of our Bell states. Their argument ran as follows:

1. Locality: What happens at one location cannot instantaneously affect a distant location.
2. Realism: If we can predict the value of a physical quantity with certainty and without disturbing the system, then that quantity has a definite value (an "element of physical reality").
3. The setup: Alice and Bob share the state $|\Psi^-\rangle = \frac{1}{\sqrt{2}}(|\!\uparrow\downarrow\rangle - |\!\downarrow\uparrow\rangle)$ and are far apart.
4. Alice measures $\hat{S}_z$: If she gets $+\hbar/2$, she can predict with certainty that Bob's spin is $-\hbar/2$ in the $z$-direction. By criterion (2), Bob's $S_z$ must be an element of reality.
5. But Alice could have measured $\hat{S}_x$ instead: Then she could predict Bob's $S_x$ with certainty. By criterion (2), Bob's $S_x$ must also be an element of reality.
6. Conclusion: Both $S_z$ and $S_x$ of Bob's particle are simultaneously real. But quantum mechanics says these are incompatible observables (they do not commute) and cannot both have definite values simultaneously.

EPR concluded that quantum mechanics must be incomplete — there must be "hidden variables" that determine the outcomes in advance.

11.9.2 Schrödinger's Response: Naming "Entanglement"

Schrödinger, responding to the EPR paper, coined the term Verschränkung (entanglement) and argued that the EPR correlations arise because:

"The best possible knowledge of a whole does not include the best possible knowledge of its parts."

This is precisely what the partial trace tells us: a pure entangled state of $AB$ yields mixed states for $A$ and $B$ individually.

11.9.3 Bell's Theorem (1964): Nature Settles the Debate

John Bell showed that any local hidden variable theory must satisfy certain inequalities (the Bell inequalities) on measurement correlations. Quantum mechanics predicts violations of these inequalities for entangled states. Experiments (Aspect et al. 1982, and many since, culminating in loophole-free tests in 2015) have decisively confirmed the quantum predictions.

The verdict: Nature is not locally realistic in the EPR sense. Entanglement is not a sign of incompleteness — it is a fundamental feature of quantum mechanics.

📊 Quantifying the violation: For the CHSH inequality (the most common form of Bell's inequality), classical correlations satisfy $|S| \leq 2$. Quantum mechanics allows $|S| \leq 2\sqrt{2} \approx 2.83$ (the Tsirelson bound). The state achieving this maximum is $|\Phi^+\rangle$ with appropriately chosen measurement bases. We will derive this in detail in the chapter on Bell's theorem.

11.10 Summary and Project Checkpoint

11.10.1 Chapter Summary

This chapter introduced the mathematical framework for describing composite quantum systems:

The tensor product $\mathcal{H}_A \otimes \mathcal{H}_B$ is the Hilbert space of a composite system, with dimension $d_A \cdot d_B$. States in this space are represented using the Kronecker product for vectors and matrices.

Product states $|\psi\rangle \otimes |\phi\rangle$ describe systems where each subsystem has a well-defined individual state. These are the analogues of classical composite states.

Entangled states cannot be written as product states. They are the generic case: almost all states in a composite Hilbert space are entangled. Entanglement means that the whole system is in a definite pure state while its parts are not.

The coefficient matrix $C_{ij} = \alpha_{ij}$ organizes the amplitudes. Its rank determines separability: rank 1 = separable, rank > 1 = entangled.

The Schmidt decomposition $|\Psi\rangle = \sum_k \lambda_k |a_k\rangle|b_k\rangle$ provides a canonical form. The Schmidt coefficients (singular values of $C$) quantify entanglement. The Schmidt rank equals the number of nonzero coefficients.

The partial trace $\hat{\rho}_A = \text{Tr}_B(|\Psi\rangle\langle\Psi|)$ gives the reduced density matrix — the correct description of a subsystem. For entangled states, $\hat{\rho}_A$ is mixed even when $|\Psi\rangle_{AB}$ is pure.

Bell states form a maximally entangled basis for two qubits:

Each Bell state has entanglement entropy $S = 1$ ebit and reduced density matrix $\hat{\rho}_A = \frac{1}{2}\hat{I}$.

11.10.2 Key Terms

11.10.3 Project Checkpoint: Toolkit v1.1 — Tensor Products and Entanglement

Extend your quantum mechanics toolkit from previous chapters with the following capabilities:

1. Tensor product construction: Given state vectors for systems $A$ and $B$, compute their tensor product.
2. Entanglement detection: Given a two-qubit state vector, determine whether it is entangled using the coefficient matrix rank test.
3. Schmidt decomposition: Compute the Schmidt decomposition of a bipartite state.
4. Partial trace: Compute reduced density matrices.
5. Bell state analysis: Express any two-qubit state in the Bell basis.
6. Entanglement entropy: Calculate the von Neumann entropy of the reduced density matrix.

See code/project-checkpoint.py for the implementation.

11.10.4 Looking Ahead

The tensor product formalism and entanglement introduced here are the foundation for everything that follows. In the next chapters, we will use these tools to:

• Add angular momenta using Clebsch-Gordan coefficients (the generalization of singlet-triplet decomposition)
• Prove Bell's theorem and derive Bell inequalities quantitatively
• Build quantum gates that operate on multi-qubit systems
• Analyze decoherence through entanglement with the environment

Entanglement is not just a mathematical curiosity — it is the resource that powers quantum teleportation, quantum cryptography, quantum error correction, and quantum computation. You have now met the mathematical tools to work with it rigorously.

🔄 Spaced Review Prompt: Before moving to Chapter 12, make sure you can: (1) write down the four Bell states from memory, (2) compute a partial trace for a two-qubit state, (3) explain why an entangled state cannot be described by pure states of its subsystems. If any of these feel shaky, revisit the relevant section before proceeding.

Glossary of Notation Introduced in This Chapter