Interpretations of Quantum Mechanics · Chapter 07
What would it mean for reduction to be an actual physical process?
Starting point: Chapters 1–6; familiarity with density operators and linear algebra.
Anchor example: Poisson-distributed position localizations and a qubit dephasing toy model.
Method: Calculate a concrete experiment, identify the assumptions, then compare what different accounts explain.
Watch for: An ensemble master equation alone does not show that individual trajectories localize.
This lesson includes explanations, worked examples, and problems with solutions. Advanced results whose proofs are outside the chapter are identified explicitly.
What would it mean for reduction to be an actual physical process?
Objective-collapse models change quantum dynamics so that macroscopic alternatives do not indefinitely coexist as one uncollapsed superposition. They are therefore more than different verbal readings of an unchanged formalism. A complete model must specify when localization occurs, how the state changes, and how outcomes are distributed.
In a simple GRW-type position-jump model for distinguishable particles, each particle undergoes localization events at rate λ. A jump for particle i with center x acts through
\[L_i(x)=(\pi r_C^2)^{-3/4}\exp[-(\hat q_i-x)^2/(2r_C^2)],\qquad \int d^3x\,L_i(x)^\dagger L_i(x)=I.\]The center has probability density \(\|L_i(x)\Psi\|^2\), and the updated vector is \(L_i(x)\Psi/\|L_i(x)\Psi\|\). Between jumps the state evolves with the Hamiltonian. Here r_C fixes the Gaussian convention; other texts can parameterize the same width differently.
The jump multiplies the configuration wavefunction by a localized Gaussian. It narrows spatial alternatives probabilistically, rather than asking an observer to declare a measurement. Gaussian tails remain nonzero, however, so the interpretation of macroscopic definiteness and the choice of material ontology still need attention. A wavefunction law alone is not a complete description of flashes or mass density.
Worked check 7.1. For Nλt=3, find the probability of at least one jump.
It is 1−e⁻³≈0.9502. This is an event probability, not a universal branch-resolution probability.
If N distinguishable constituents each have independent rate λ, the probability of no event over time t is exp(−Nλt). For a rigid superposition whose corresponding particle locations are separated far beyond r_C, one localization can strongly suppress one branch relative to the other. This is the basic amplification mechanism. The formula does not claim that every event resolves every possible pair of states: separations and geometry matter.
Averaging the jump process gives an ensemble master equation of the form
\[\dot\rho=-\frac{i}{\hbar}[H,\rho]+\lambda\sum_i\left(\int d^3x\,L_i(x)\rho L_i(x)^\dagger-\rho\right).\]The jump law supplies more information than this average. To see why, consider the toy qubit equation \(\dot\rho=\kappa(Z\rho Z-\rho)\). Its off-diagonal entries decay at rate 2κ. Random unitary Z flips generate this same equation, yet an individual |+⟩ trajectory then merely switches between |+⟩ and |−⟩. It does not localize into a Z eigenstate.
Thus decay of an ensemble coherence is not enough to establish collapse of individual states. The state-vector stochastic process, its ontology, and the ensemble predictions must be specified separately. The qubit example is a warning about inference, not a replacement for the spatial collapse model.
Worked check 7.2. Solve the toy master equation for ρ₀₁.
Since (ZρZ)₀₁=−ρ₀₁, its derivative is −2κρ₀₁ and ρ₀₁(t)=e⁻²κᵗρ₀₁(0).
A modified theory can be investigated experimentally when its additional dynamics predicts observable differences. In collapse models, suppressed interference, momentum diffusion, or heating can provide signatures, depending on the model and coupling. Ordinary gas collisions, radiation, vibration, and detector noise can also suppress interference. A credible test must compare quantitative models rather than treating any missing visibility as evidence for collapse.
For an explicitly simplified visibility model, write \(V(t)=V_0\exp[-(\Gamma_{\rm env}+\Gamma_{\rm new})t]\). If environmental decay is independently known and the measured visibility has a lower bound V_min, the allowed extra rate satisfies
\[\Gamma_{\rm new}\le-\frac1t\log(V_{\min}/V_0)-\Gamma_{\rm env}.\]This bound is conditional on the exponential model, calibrated visibility, and uncertainty treatment. A negative right side signals incompatibility of those inputs with a nonnegative additional rate; it is not a measured negative collapse rate. Real constraints on λ and r_C require the model-specific relationship between localization parameters and the measured signal.
The research habit here is transferable: publish assumptions, units, nuisance parameters, and statistical confidence. Distinguish a bound on one parameterized collapse model from a verdict about all interpretations. No current experimental limit is asserted by our hypothetical examples.
Worked check 7.3. With V₀=1, V_min=0.9, t=1 second, and Γ_env=0.05 s⁻¹, bound Γ_new.
The hypothetical bound is −ln(0.9)−0.05≈0.05536 s⁻¹, assuming the stated environmental estimate and visibility model.
These exercises draw on the whole chapter. Attempt each question before opening a hint, explain your reasoning, then compare it with the worked solution.
Why is the jump-center expression a probability density rather than one probability?
The center ranges continuously over space.
The probability in a region is the integral of ||L(x)Ψ||² over that region. The completeness integral ensures total probability one.
Verify the one-dimensional Gaussian completeness normalization.
Square (πr²)⁻¹/⁴ exp[−(q−x)²/(2r²)] and integrate over x.
The squared prefactor is 1/(√π r), and the Gaussian integral is √π r, giving one for each q.
Does a Gaussian jump eliminate all distant amplitude exactly?
Examine its tails.
No. The multiplier is nonzero at finite separation. Suppression can be very strong without compact support or exact elimination.
For a Poisson process with total rate Nλ, find the mean waiting time.
Use the exponential survival distribution.
The mean is 1/(Nλ), assuming independent constant-rate constituent events.
Why might a jump fail to distinguish two branches?
Compare their position separation with r_C.
If their positions nearly coincide on the localization scale, the Gaussian factors are similar. The event gives little relative suppression.
Show trace preservation of the ensemble jump term.
Use cyclicity and the completeness integral.
The integrated gain has trace Tr[ρ∫L†L]=Trρ, which cancels the loss term for each particle.
Does observed decoherence uniquely identify objective collapse?
List an ordinary competing cause.
No. Environmental entanglement can produce the same reduced coherence loss. Discriminating tests need calibrated alternatives and model-specific signatures.
What does a negative upper bound in the visibility toy model mean?
Require Γ_new≥0.
The stated inputs cannot all fit that model with a nonnegative extra rate. Check calibration, uncertainties, and environmental assumptions.
Why can’t a bound on Γ_new be directly called a bound on λ?
Look for a map between the quantities.
The effective visibility rate depends on geometry, particle content, separation, and r_C as well as λ. That relation must be derived for the selected collapse model.
Design a minimal report for a hypothetical collapse test.
Include models and uncertainty.
Specify preparation, mass and geometry, separation and time, environmental budget, measured observable and confidence procedure, collapse equation and parameter convention, and the resulting conditional parameter region.