Coherence Returns.

The added sequence (64–96 seconds) starts a new, ideal five-qubit experiment. A system initially in |+⟩ interacts with four environment qubits initially in |0⟩. The conditional interaction on fragment j is U_j = |0⟩⟨0| ⊗ I + |1⟩⟨1| ⊗ R_y(2α_j). The colored arrows inside each environment glyph are its two conditional states, not two actual particles.

For equal pointer priors, the optimum probability of distinguishing a fragment's conditional states is (1 + sin α_j) / 2 over the angle range shown. This is an ideal model probability, not experimental data. The global pure-state purity stays one. Its individual subsystems can be mixed because they are entangled.

Reversing E4 alone leaves three records; the remaining product of overlaps is still small. Reversing every controlled gate in reverse order restores the original product state. This demonstrates reversibility in a fully controlled, isolated finite model. It does not imply that arbitrary laboratory decoherence is easy to reverse, that records can be deleted for free, or that time itself runs backward.

The experiment contains no fitted universal threshold, speculative classicalization ratio, fabricated data, or claim of new physics. Camera paths and connections are display choices, not the spatial motion of a qubit.

Background: IBM Quantum: channels, joint systems and partial trace; W. H. Zurek: Quantum Darwinism. The explicit five-qubit example above is the model used in this film; it is not a laboratory demonstration.

The sphere represents the three Bloch coordinates of an ideal two-level quantum state. It is not a particle orbit, a photographed wave function, or a literal laboratory apparatus. The lower energy basis state |0⟩ is at the north pole by the usual Bloch convention; sphere height is not energy.

The prepared state is Rᵧ(π/2)|0⟩. A controlled relative phase ends at φ = 7π/3, equivalent to π/3. The analysis is Rᵧ(−π/2), so P(0) = (1 + cos φ)/2 = 3/4. Rotations and probability displays are computed from this state at every frame. The single readout is an illustrative ideal projective update, not a proposed mechanism of collapse. The 100-shot raster uses seeded classical samples of P(0) = 0.75 from independently re-prepared copies.

The coherent-versus-mixture comparison is a NEW preparation: |+⟩ against I/2. Both initially give equal Z-basis probabilities. The same analysis pulse maps |+⟩ to |0⟩ but leaves I/2 unchanged. The comparison shows exact model probabilities, not finite-shot frequencies.

The final sequence starts with a NEW |+⟩ preparation under a pure-dephasing channel and phase evolution. With τ = t/T₂, c = exp(−τ) and φ = (4π/3)τ, r = (c cos φ, c sin φ, 0). Purity = (1 + |r|²)/2. The state-history helix projects (r_x,r_y,r_z=0,τ) into three display coordinates. Its vertical axis is elapsed time. There is no universal classicalization threshold, speculative prediction, or experimental discovery in this film.

Pulse carrier motion, glow, camera motion, timing and the decorative stage are illustration choices. All measurement samples are simulated. The MP4 soundtrack is synthesized sound design, not the sound of a quantum system.

Sources: IBM Quantum: superposition and interference; IBM Quantum: RY gate; IBM Quantum: quantum channels and dephasing.