Entanglement in Foundations

Entanglement sits at the center of many foundational debates because it separates the state of a composite system from any simple list of independent local properties. The same formal facts behind partial traces, Bell states, conditional states, and EPR-like correlations become the input to arguments about locality, completeness, hidden variables, steering, no-signaling, and measurement.

The state formalism distinguishes three questions: whether a joint state is separable, whether observed correlations admit a local hidden-variable explanation, and how a measurement outcome should be interpreted. They require different assumptions. No-cloning and no-signaling develops the communication consequences.

Helpful background. Bell states supplies the examples; local measurement statistics and conditional states distinguish averaging from selecting an outcome.

EPR Pairs as State Examples

In discrete variables, a standard EPR-pair example is the Bell state

$$\lvert\Phi^+\rangle = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert11\rangle \right).$$

Its one-qubit reduced states are

$$\rho_A = \rho_B = \frac12 I.$$

Locally, a computational-basis measurement has two equally likely outcomes. Jointly, the state has perfect same-outcome correlations in the computational basis and in the $x$ basis. The coherence between $\lvert00\rangle$ and $\lvert11\rangle$ is what distinguishes the Bell state from the classically correlated mixture

$$\rho_{\mathrm{cc}} = \frac12 \lvert00\rangle\langle00\rvert + \frac12 \lvert11\rangle\langle11\rvert.$$

In continuous variables, the ideal EPR pattern uses collective observables such as

$$X_1-X_2, \qquad P_1+P_2.$$

These two commute, so an idealized generalized state can have sharp values of both. The foundational tension comes from combining the corresponding remote predictability with the noncommutation of local observables such as $X_1$ and $P_1$. The normalizability and finite-energy cautions belong to the EPR State Preview.

The EPR Argument Preview

The historical EPR argument is not merely the statement “there is an entangled state.” It uses additional premises. In compressed form, it involves:

• perfect correlations in the historical idealized state;
• a locality assumption about what operations here can affect there;
• a criterion connecting certain prediction with an element of reality;
• a claim that the quantum state description is incomplete if it lacks those elements.

Bohr’s reply and later foundations work scrutinize those premises and the meaning of experimental context. Finite experimental correlations require quantitative tests with explicit noise and inference assumptions; merely replacing ‘perfect’ by ‘very strong’ is not the historical argument.

The safe habit is to use “EPR state” for a state or correlation pattern, and “EPR argument” for the locality-and-completeness reasoning built on it.

Bell Theorem Preview

Bell’s theorem is sharper than the EPR argument because it produces experimentally testable inequalities. A standard CHSH expression is built from correlations

$$E(a,b) = \langle A_a B_b\rangle,$$

where $A_a$ and $B_b$ are outcomes with values $\pm1$ for two possible measurement settings on each side. The local model has outcome probabilities

$$p(a,b|x,y)=\int d\lambda\,\mu(\lambda)p(a|x,\lambda)p(b|y,\lambda),$$

where the distribution $\mu$ is independent of the choices $x,y$. This combines locality of the response probabilities with measurement-setting independence. Under these assumptions, the CHSH correlations obey

$$\left| E(a,b) + E(a,b') + E(a',b) - E(a',b') \right| \le 2.$$

Quantum theory can violate this bound. For a singlet state, the spin correlation has the form

$$E(\mathbf a,\mathbf b) = -\mathbf a\cdot\mathbf b,$$

and suitable choices of directions reach

$$2\sqrt2.$$

For example, choose $\mathbf a=\mathbf z$, $\mathbf a'=\mathbf x$, $\mathbf b=(\mathbf z+\mathbf x)/\sqrt2$, and $\mathbf b'=(\mathbf z-\mathbf x)/\sqrt2$. The first three correlations are $-1/\sqrt2$, while the last is $+1/\sqrt2$, giving absolute CHSH value $2\sqrt2$. Detection, sampling, and setting choices determine whether an experiment tests the stated model without additional loopholes. Bell’s theorem is not the definition of entanglement.

Entanglement, Steering, and Bell Nonlocality

Several nonclassical notions form a hierarchy in many common settings:

$$\text{Bell nonlocality} \quad\Longrightarrow\quad \text{steering} \quad\Longrightarrow\quad \text{entanglement}.$$

The converses do not hold in full generality. Some entangled mixed states do not violate a chosen Bell inequality. Some entangled states are not steerable for a specified measurement class. The measurement class and trusted parties must be specified. For a concrete gap, the two-qubit Werner state $\rho_v=v\lvert\Psi^-\rangle\langle\Psi^-\rvert+(1-v)I_4/4$ is entangled for $v>1/3$. It admits a local hidden-variable model for all local projective measurements when $v\le1/2$ (Werner, 1989). At $v=1/2$, therefore, entanglement is present without Bell nonlocality in that measurement class. The familiar singlet CHSH directions give $2\sqrt2v$, below two in this example.

Steering language is especially close to the conditional-state formalism. If a measurement on $B$ produces outcome $b$, the state assigned to $A$ can change to

$$\rho_{A\vert b} = \frac{\sigma_{A\vert b}}{p(b)}.$$

Steering asks whether the subnormalized conditional states for all measurement choices $y$ admit a single local-hidden-state model,

$$\sigma_{A|b,y}=\sum_\lambda p(\lambda)p(b|y,\lambda)\rho_A^\lambda,$$

with states $\rho_A^\lambda$ independent of $y$. Ruling out every such model demonstrates steering from $B$ to $A$ under the chosen trusted-measurement assumptions. Conditional states changing with an outcome is not enough: the separable $00/11$ mixture above also prepares $\lvert0\rangle$ or $\lvert1\rangle$ conditionally. Until the outcome is communicated, the average is still $\rho_A$.

No-Signaling from Reduced States

Entanglement does not allow controllable faster-than-light signaling. The algebraic reason is that local measurement statistics for $A$ are computed from the reduced state

$$\rho_A = \operatorname{Tr}_B\rho_{AB}.$$

For a local measurement on $A$ with effect $E_a$, the probability is

$$p(a) = \operatorname{Tr}_A(\rho_A E_a).$$

If a remote party measures $B$ but the outcome is not known to $A$, the local statistics on $A$ are still those of $\rho_A$. In instrument language, averaging over all remote outcomes gives the same unconditioned reduced state for $A$ under an ordinary trace-preserving local operation on $B$:

$$\rho_A' = \operatorname{Tr}_B \left[ (\mathrm{id}_A\otimes\mathcal E_B)(\rho_{AB}) \right] = \rho_A.$$

Here $\mathrm{id}_A$ is the identity map on operators of $A$. One way to check the equality is to use Kraus operators $K_b$ for the remote channel. Cyclicity of the trace over $B$ gives a sum containing $\sum_bK_b^\dagger K_b=I_B$. If only a selected outcome is retained, that sum is not the identity; the resulting conditional subensemble is available only when the selection is identified.

This does not make the remote measurement irrelevant. After the outcome is communicated, the data can be sorted into conditional subensembles with different states. The distinction between unconditioned local marginals and conditioned subensembles is the core of no-signaling in the elementary formalism.

Measurement Entanglement

Measurement models naturally create entanglement between a system and an apparatus. If a two-state system begins in

$$\alpha\lvert0\rangle + \beta\lvert1\rangle$$

and an ideal premeasurement correlates system states with pointer states, the joint state becomes

$$\alpha\lvert0\rangle\lvert R_0\rangle + \beta\lvert1\rangle\lvert R_1\rangle.$$

If the pointer states are orthogonal, tracing out the apparatus gives

$$\rho_S = \lvert\alpha\rvert^2 \lvert0\rangle\langle0\rvert + \lvert\beta\rvert^2 \lvert1\rangle\langle1\rvert.$$

For nonorthogonal pointer states, put $\gamma=\langle R_1|R_0\rangle$. The reduced off-diagonal term is $\alpha\beta^*\gamma\lvert0\rangle\langle1\rvert$ plus its adjoint. Thus distinguishable records suppress local coherence continuously: $|\gamma|=1$ leaves it intact up to a phase, whereas $\gamma=0$ removes it. An environment can make the overlaps extremely small and interference hard to recover, but the joint unitary state still contains the correlations.

This is the beginning of the decoherence story, not the end of the measurement problem. Decoherence explains why interference between pointer alternatives becomes inaccessible in local macroscopic observables. It does not, by itself, say which interpretation of individual outcomes is correct.

What Entanglement Does Not Settle

Entanglement is not an interpretation of quantum mechanics. It is a mathematically precise feature of composite states. It supplies unavoidable input to foundational debates, but it does not by itself decide:

• whether the quantum state is complete;
• whether hidden variables are possible under relaxed assumptions;
• what counts as a physical “element of reality”;
• how to understand a single measurement outcome;
• whether collapse is physical, epistemic, effective, or absent;
• which interpretation of quantum mechanics is preferred.

The extra assumptions, rather than the existence of entanglement alone, determine what a proposed interpretation or experiment establishes.

References

• J. S. Bell, “On the Einstein Podolsky Rosen paradox,” Physics Physique Fizika 1, 195–200 (1964), doi:10.1103/PhysicsPhysiqueFizika.1.195.
• J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, “Proposed experiment to test local hidden-variable theories,” Physical Review Letters 23, 880–884 (1969), doi:10.1103/PhysRevLett.23.880.
• A. Einstein, B. Podolsky, and N. Rosen, “Can quantum-mechanical description of physical reality be considered complete?,” Physical Review 47, 777–780 (1935), doi:10.1103/PhysRev.47.777.
• R. F. Werner, “Quantum states with Einstein–Podolsky–Rosen correlations admitting a hidden-variable model,” Physical Review A 40, 4277–4281 (1989), doi:10.1103/PhysRevA.40.4277.
• H. M. Wiseman, S. J. Jones, and A. C. Doherty, “Steering, entanglement, nonlocality, and the Einstein–Podolsky–Rosen paradox,” Physical Review Letters 98, 140402 (2007), doi:10.1103/PhysRevLett.98.140402.
• W. H. Zurek, “Decoherence, einselection, and the quantum origins of the classical,” Reviews of Modern Physics 75, 715–775 (2003), doi:10.1103/RevModPhys.75.715.

Exercises

1. Bell pair local state. Show that the reduced state of either qubit of $\lvert\Phi^+\rangle$ is $I/2$.

Solution

For

$$\lvert\Phi^+\rangle = \frac{\lvert00\rangle+\lvert11\rangle}{\sqrt2},$$

the density operator has diagonal terms $\lvert00\rangle\langle00\rvert$ and $\lvert11\rangle\langle11\rvert$ plus cross terms. Tracing over the second qubit kills the cross terms because $\langle0\vert1\rangle=0$. Therefore

$$\rho_A = \frac12 \lvert0\rangle\langle0\rvert + \frac12 \lvert1\rangle\langle1\rvert = \frac12 I.$$

2. Conditional but not signaling. If $B$ measures the second qubit of $\lvert\Phi^+\rangle$ in the computational basis, what conditional states can be assigned to $A$, and what is their average?

Solution

Outcome $0$ on $B$ occurs with probability $1/2$ and assigns $\lvert0\rangle$ to $A$. Outcome $1$ occurs with probability $1/2$ and assigns $\lvert1\rangle$ to $A$. Averaging over outcomes gives

$$\frac12 \lvert0\rangle\langle0\rvert + \frac12 \lvert1\rangle\langle1\rvert = \frac12 I.$$

Thus conditioning changes the subensemble after the outcome is known, but the unconditioned local state is unchanged.

3. Bell violation is not the definition. Why is Bell-inequality violation not the definition of entanglement?

Solution

Entanglement is defined by nonseparability of a state relative to a subsystem split. Bell-inequality violation is a stronger operational property involving measurement settings and locality assumptions. Some entangled mixed states do not violate a given Bell inequality, so Bell violation cannot be the definition of entanglement.

4. Premeasurement reduced state. Suppose

$$\lvert\Psi\rangle = \alpha\lvert0\rangle\lvert R_0\rangle + \beta\lvert1\rangle\lvert R_1\rangle,$$

with $\langle R_0\vert R_1\rangle=0$. Trace out the apparatus and find $\rho_S$.

Solution

The diagonal terms survive. The off-diagonal terms contain $\langle R_1\vert R_0\rangle$ or $\langle R_0\vert R_1\rangle$, which vanish. Therefore

$$\rho_S = \lvert\alpha\rvert^2 \lvert0\rangle\langle0\rvert + \lvert\beta\rvert^2 \lvert1\rangle\langle1\rvert.$$

5. Imperfect records. Take $\alpha=\beta=1/\sqrt2$ and real pointer overlap $\gamma=0.6$. Find the eigenvalues and purity of the reduced system state. Does the reduction select one definite outcome?

Solution

The reduced matrix is $\rho_S=\tfrac12\begin{pmatrix}1&0.6\\0.6&1\end{pmatrix}$. Its eigenvalues are $(1\pm0.6)/2=0.8,0.2$, and its purity is $0.8^2+0.2^2=0.68$. This is a prediction for local statistics after ignoring the apparatus. It does not by itself select a single outcome or specify an interpretation of outcomes.