Quantum Experiments

Quantum Experiments · V · Records and observers · Chapter 21

Wigner’s Friend

Can an observer and an outside experimenter assign different states consistently?

In this chapter
  1. An isolated laboratory inside a superobserver’s apparatus
  2. Uncomputing a CNOT premeasurement and joint interference
  3. Environmental record leakage and observer assumptions
  4. Interactive experiment
  5. Problems and solutions

21.1 An isolated laboratory inside a superobserver’s apparatus

21.1.1 Internal record formation versus external unitary evolution

Eugene Wigner sharpened Schrödinger’s cat scenario in 1961 by replacing the inanimate pointer with an observer—Wigner’s friend \(F\)—inside a completely isolated laboratory \(L = S \otimes F\). The friend measures a spin/qubit \(S\) prepared in \(\alpha|0\rangle_S + \beta|1\rangle_S\) in the \(\{|0\rangle_S, |1\rangle_S\}\) basis and records a definite outcome in memory register \(F\). From the friend’s internal perspective upon seeing an outcome, the state appears updated to \(|0\rangle_S|0\rangle_F\) (with probability \(|\alpha|^2\)) or \(|1\rangle_S|1\rangle_F\) (with probability \(|\beta|^2\)). From the external perspective of Wigner (a “superobserver” whose laboratory is completely isolated from \(L\)), the interaction inside \(L\) is a unitary premeasurement \(U_{\mathrm{meas}} = \mathrm{CNOT}_{S\to F}\) that prepares the entangled state

\[|\Psi_{SF}\rangle = \alpha|0\rangle_S|0\rangle_F + \beta|1\rangle_S|1\rangle_F.\]

21.1.2 Reading the friend’s record versus performing a joint interference test

So long as Wigner simply opens the door and asks the friend which outcome occurred—measuring in the record basis \(\{|00\rangle_{SF}, |11\rangle_{SF}\}\)—both descriptions predict the exact same probabilities \(P(00) = |\alpha|^2\) and \(P(11) = |\beta|^2\). A physical discrepancy between the pure entangled state \(|\Psi_{SF}\rangle\) and the incoherent mixture \(\rho_{\mathrm{mix}} = |\alpha|^2|00\rangle\langle 00| + |\beta|^2|11\rangle\langle 11|\) arises only if Wigner instead performs a coherent joint measurement across the entire laboratory \(S \otimes F\) in a complementary superobserver basis.

21.1.3 Modeling the friend as a reversible quantum memory register

In circuit language, treating the friend’s apparatus and memory register as a qubit \(F\) initialized to \(|0\rangle_F\) models the internal measurement as \(\mathrm{CNOT}_{S\to F}((\alpha|0\rangle_S + \beta|1\rangle_S)|0\rangle_F) = \alpha|00\rangle_{SF} + \beta|11\rangle_{SF}\). Testing whether coherence between \(|00\rangle_{SF}\) and \(|11\rangle_{SF}\) survives requires either measuring the two-qubit parity operator \(X_S \otimes X_F\) or applying the inverse unitary \(\mathrm{CNOT}_{S\to F}^\dagger = \mathrm{CNOT}_{S\to F}\) to uncompute the friend’s memory register.

Visual animation · Section 21.1

An isolated laboratory inside a superobserver’s apparatus

An isolated laboratory inside a superobserver’s apparatusSchematic apparatus; geometry and moving indicators are explanatory, not measured trajectories.
Prepare \(|+\rangle\)
trigger qubit
Isolated laboratory (unitary coupling)
\(|0, R_0\rangle\)
branch weight \(1/2\)
\(|1, R_1\rangle\)
branch weight \(1/2\)
coherence \(\gamma\)
Superobserverrecord / interfereFriend memory register probed in record or interference basis.Section 21.1 Live State:
\(|\Psi_{SF}\rangle = \mathrm{CNOT}_{S\to F}(|+\rangle_S|0\rangle_F) = \frac{|00\rangle_{SF} + |11\rangle_{SF}}{\sqrt{2}}\)
Stage 1 (21.1.1): Internal record formation versus external unitary evolutioReady

Interactive visual model for 21.1 An isolated laboratory inside a superobserver’s apparatus (Internal record formation versus external unitary evolution · Reading the friend’s record versus performing a joint interference test · Modeling the friend as a reversible quantum memory register). Use Animate or Step to follow the physical state evolution and quantitative predictions.

21.2 Uncomputing a CNOT premeasurement and joint interference

21.2.1 Superobserver basis states and the coherent visibility witness

Suppose the system starts in \(|+\rangle_S = (|0\rangle_S + |1\rangle_S)/\sqrt{2}\) (\(\alpha = \beta = 1/\sqrt{2}\)) so that unitary premeasurement yields \(|\Phi^+\rangle_{SF} = (|00\rangle_{SF} + |11\rangle_{SF})/\sqrt{2}\). Define Wigner’s joint interference basis

\[|\Phi_{\pm}(\phi)\rangle_{SF} = \frac{|00\rangle_{SF} \pm e^{i\phi}|11\rangle_{SF}}{\sqrt{2}}.\]

If the laboratory \(S\otimes F\) retains coherence fraction \(W \in [0, 1]\) between the two record branches,

\[\rho_{SF}(W) = \frac{1}{2}\Big(|00\rangle\langle 00| + |11\rangle\langle 11| + W|00\rangle\langle 11| + W|11\rangle\langle 00|\Big),\]

then Wigner’s probability of obtaining outcome \(|\Phi_+(\phi)\rangle_{SF}\) is

\[P_+(\phi) = \langle\Phi_+(\phi)|\rho_{SF}(W)|\Phi_+(\phi)\rangle = \frac{1}{2}\big(1 + W\cos\phi\big).\]

An irreversible collapse inside \(L\) (\(W = 0\)) yields a flat line \(P_+(\phi) = 1/2\) for all \(\phi\), whereas reversible unitary evolution (\(W = 1\)) yields full-contrast fringes \(P_+(\phi) = \cos^2(\phi/2)\), reaching \(P_+(0) = 1\).

21.2.2 Reversing the system–memory CNOT to test interference

David Deutsch noted in 1985 that Wigner can implement the \(\phi = 0\) interference test by applying a second \(\mathrm{CNOT}_{S\to F}\) to uncompute the friend’s record:

\[\mathrm{CNOT}_{S\to F}\left(\frac{|00\rangle_{SF} + |11\rangle_{SF}}{\sqrt{2}}\right) = \left(\frac{|0\rangle_S + |1\rangle_S}{\sqrt{2}}\right)\otimes|0\rangle_F = |+\rangle_S\otimes|0\rangle_F.\]

Measuring \(S\) in the \(\{|+\rangle_S, |-\rangle_S\}\) basis then yields \(|+\rangle_S\) with probability \(P(+_S) = (1 + W)/2 = 1\) when \(W = 1\), versus \(P(+_S) = 1/2\) if the friend’s measurement collapsed the state to \(\rho_{\mathrm{mix}}\). Crucially, however, uncomputing \(F\) back to \(|0\rangle_F\) erases the friend’s outcome record: Wigner can never simultaneously retain the friend’s which-outcome record and observe \(P(+_S) = 1\).

21.2.3 Conditioned internal states versus unconditioned global states

The mathematical difference between the friend’s state assignment (\(|0\rangle_S\) or \(|1\rangle_S\)) and Wigner’s state assignment (\(|\Phi^+\rangle_{SF}\)) is the difference between a state conditioned on an internal branch record and an unconditioned joint state of system plus memory. Both assignments agree on every measurement that preserves and commutes with the friend’s memory register \(Z_F\); they differ only on superobserver operations that do not commute with \(Z_F\) and therefore overwrite the friend’s memory.

Visual animation · Section 21.2

Uncomputing a CNOT premeasurement and joint interference

21.2 Uncomputing a CNOT premeasurement and joint interferenceInteractive geometric state/phasor dial and quantitative prediction curve for Uncomputing a CNOT premeasurement and joint interference.1. State / Amplitude GeometryAmplitude / Bloch projection
Value \(= \) 0.500
2. Quantitative Law (21.2)
\(|\Psi_{SF}\rangle = \mathrm{CNOT}_{S\to F}(|+\rangle_S|0\rangle_F) = \frac{|00\rangle_{SF} + |11\rangle_{SF}}{\sqrt{2}}\)

Interactive visual model for 21.2 Uncomputing a CNOT premeasurement and joint interference (Superobserver basis states and the coherent visibility witness · Reversing the system–memory CNOT to test interference · Conditioned internal states versus unconditioned global states). Use Animate or Step to follow the physical state evolution and quantitative predictions.

21.3 Environmental record leakage and observer assumptions

21.3.1 How interpretations resolve the friend–superobserver assignment gap

Different interpretations of quantum mechanics locate the resolution of Wigner’s friend in different places:

21.3.2 Scope of few-qubit photonic and superconducting friend demonstrations

Recent experiments on entangled photons and superconducting circuits implement the Wigner’s friend circuit by using a single ancillary qubit or photon path as the “friend” register \(F\). These experiments confirm \(W \approx 1\) and verify the two-qubit circuit identities cleanly, but because a one-qubit register is not a macroscopic observer with thermodynamically irreversible records, they test two-qubit unitary mechanics rather than macroscopic observer collapse.

21.3.3 Why unreversed environmental copies destroy superobserver interference

If even a single environmental mode \(E\) outside Wigner’s control correlates with the friend’s record (\(|00\rangle|E_0\rangle\) vs.\ \(|11\rangle|E_1\rangle\) with \(\langle E_0|E_1\rangle = 0\)), applying \(\mathrm{CNOT}_{S\to F}\) to \(S\) and \(F\) alone produces \((|0\rangle_S|E_0\rangle + |1\rangle_S|E_1\rangle)|0\rangle_F/\sqrt{2}\), whose reduced state on \(S\) is \(\frac{1}{2}I_S\) with \(P(+_S) = 1/2\). Reversing a measurement requires reversing every physical copy of the record.

Visual animation · Section 21.3

Environmental record leakage and observer assumptions

21.3 Environmental record leakage and observer assumptionsCompares ideal prediction against finite efficiency/coherence and audits the experimental assumptions for Environmental record leakage and observer assumptions.1. Ideal Model vs. Imperfect Apparatus
\(P_+(\phi) = \frac{1}{2}(1 + \gamma\cos\phi), \quad [Z_F,\, X_S\otimes X_F] = 2i(X_S\otimes Y_F) \neq 0\)
2. Audit Ledger (21.3)
Fidelity / efficiency \(\eta = \) 0.85
21.3.1 CheckHow interpretations resolve the friend21.3.2 CheckScope of few-qubit photonic and superc21.3.3 ScopeWhy unreversed environmental copies de
Ideal \(= \) 1.00 \(\;\mid\;\) Obs. \(= \) 0.85
Audit status: calibrated

Interactive visual model for 21.3 Environmental record leakage and observer assumptions (How interpretations resolve the friend–superobserver assignment gap · Scope of few-qubit photonic and superconducting friend demonstrations · Why unreversed environmental copies destroy superobserver interference). Use Animate or Step to follow the physical state evolution and quantitative predictions.

Explore the Apparatus

Prepare the system state \(\alpha|0\rangle_S + \beta|1\rangle_S\), run the internal \(\mathrm{CNOT}_{S\to F}\) premeasurement, and adjust the environmental record leakage \(1 - W\) outside the isolated laboratory. Compare Wigner’s outcomes when reading the friend’s record (\(Z_F\)) versus uncomputing the record and measuring interference in the superobserver basis \(|\Phi_{\pm}(\phi)\rangle_{SF}\).

Interactive model · Idealized predictions

Inside the laboratory; outside the laboratory

Inside the laboratory; outside the laboratorySchematic apparatus; geometry and moving indicators are explanatory, not measured trajectories.
Prepare \(|+\rangle\)
trigger qubit
Isolated laboratory (unitary coupling)
\(|0, R_0\rangle\)
branch weight \(1/2\)
\(|1, R_1\rangle\)
branch weight \(1/2\)
coherence \(\gamma\)
Superobserverrecord / interfereFriend memory register probed in record or interference basis.

On a small screen, swipe the diagram to inspect its components. Moving marks illustrate the model; they are not measured trajectories.

Can an observer and an outside experimenter assign different states consistently? The text and worked examples remain available without JavaScript. Animation starts only when requested; Step gives a motion-free way to follow the model.

Worked Check

Problem: System \(S\) is initialized in \(|+\rangle_S = (|0\rangle_S + |1\rangle_S)/\sqrt{2}\) and the friend’s memory register \(F\) starts in \(|0\rangle_F\). After the internal measurement \(\mathrm{CNOT}_{S\to F}\), residual coupling to an unreversed environment mode leaves coherence \(W = |\langle E_0|E_1\rangle| = 0.76\). Wigner then applies \(\mathrm{CNOT}_{S\to F}\) to uncompute the friend’s register and measures \(S\) in the \(\{|+\rangle_S, |-\rangle_S\}\) basis. Find the probabilities \(P(+_S)\) and \(P(-_S)\), and compare them with the case of complete collapse (\(W = 0\)).

Solution:

1. Before Wigner’s reversal, the reduced state of \(S\otimes F\) is

\[\rho_{SF}(W) = \frac{1}{2}\Big(|00\rangle\langle 00| + |11\rangle\langle 11| + W|00\rangle\langle 11| + W|11\rangle\langle 00|\Big).\]

2. Applying \(\mathrm{CNOT}_{S\to F}\) maps \(|00\rangle \mapsto |00\rangle\) and \(|11\rangle \mapsto |10\rangle\), factoring out \(|0\rangle_F\langle 0|_F\) and leaving system \(S\) in

\[\rho_S^{\mathrm{rev}}(W) = \frac{1}{2}\begin{pmatrix} 1 & W \\ W & 1 \end{pmatrix} = \frac{1+W}{2}|+\rangle\langle +| + \frac{1-W}{2}|-\rangle\langle -|.\]

3. For \(W = 0.76\), Wigner observes

\[P(+_S) = \frac{1 + 0.76}{2} = 0.880, \qquad P(-_S) = \frac{1 - 0.76}{2} = 0.120,\]

whereas complete collapse (\(W = 0\)) predicts \(P(+_S) = P(-_S) = 0.500\).

Summary

Practice Problems

Problem 21.1: Expectation value of the joint parity witness \(X_S \otimes X_F\)

For the partially decohered laboratory state \(\rho_{SF}(W) = \frac{1}{2}(|00\rangle\langle 00| + |11\rangle\langle 11| + W|00\rangle\langle 11| + W|11\rangle\langle 00|)\), calculate \(\langle Z_S\otimes Z_F\rangle\) and \(\langle X_S\otimes X_F\rangle\).

Hint

Use \((X_S\otimes X_F)|00\rangle = |11\rangle\) and \((X_S\otimes X_F)|11\rangle = |00\rangle\).

Solution

Because both \(|00\rangle\) and \(|11\rangle\) have \(Z_S Z_F = +1\), \(\langle Z_S\otimes Z_F\rangle = +1\) for every \(W \in [0,1]\). By contrast, \(\mathrm{Tr}[\rho_{SF}(W)(X_S\otimes X_F)] = \frac{W}{2}\langle 11|11\rangle + \frac{W}{2}\langle 00|00\rangle = W\). Thus \(Z_S Z_F\) verifies record correlation while \(X_S X_F\) directly measures the surviving coherence \(W\).

Problem 21.2: Why retaining a copy of the friend’s record blocks interference

Suppose that before Wigner applies \(\mathrm{CNOT}_{S\to F}\) to uncompute \(F\), the friend copies the record into a notebook qubit \(N\) via \(\mathrm{CNOT}_{F\to N}\), preparing \((|000\rangle_{SFN} + |111\rangle_{SFN})/\sqrt{2}\). If Wigner reverses only \(S\) and \(F\) (leaving \(N\) untouched) and measures \(S\) in the \(\{|+\rangle_S, |-\rangle_S\}\) basis, what is \(P(+_S)\)?

Hint

Apply \(\mathrm{CNOT}_{S\to F}\) to \((|000\rangle_{SFN} + |111\rangle_{SFN})/\sqrt{2}\) and trace over \(N\).

Solution

Applying \(\mathrm{CNOT}_{S\to F}\) produces \((|000\rangle_{SFN} + |101\rangle_{SFN})/\sqrt{2} = \frac{1}{\sqrt{2}}(|0\rangle_S|0\rangle_N + |1\rangle_S|1\rangle_N)\otimes|0\rangle_F\). Because \(S\) remains maximally entangled with the unreversed notebook \(N\), \(\rho_S = \frac{1}{2}I_S\) and \(P(+_S) = 1/2\).

Problem 21.3: Noncommutativity of record readout and interference readout

Compute the commutator \([Z_F,\, X_S\otimes X_F]\) between reading the friend’s record \(Z_F = I_S\otimes Z_F\) and measuring Wigner’s joint coherence witness \(X_S\otimes X_F\).

Hint

Use \([Z, X] = 2iY\) on the second tensor factor.

Solution

\([I_S\otimes Z_F,\, X_S\otimes X_F] = X_S\otimes [Z_F, X_F] = 2i(X_S\otimes Y_F) \neq 0\). Because the two observables do not commute, no measurement can simultaneously read the friend’s definite outcome \(Z_F\) and certify nonvanishing superobserver coherence \(X_S\otimes X_F\).

Key Takeaways

Sources and Further Reading

Search Quantum Experiments

Search all 24 chapters.