Quantum Decoherence

What decoherence actually is

Start with a single quantum system in a clean superposition of two distinguishable states — say a particle that is genuinely in both slit positions, written |ψ⟩ = α|L⟩ + β|R⟩. As long as nothing else interacts with it, the two branches keep a definite relative phase, and that phase produces interference: bright and dark fringes on a screen, the hallmark of quantumness.

Decoherence is what happens the instant the system starts talking to anything else — stray photons, air molecules, lattice vibrations, the readout apparatus, the rest of the universe. Each of those degrees of freedom is sensitive to which branch the system is in, so it gets dragged into the superposition too. The particle becomes entangled with its environment:

(α|L⟩ + β|R⟩) ⊗ |E₀⟩   →   α|L⟩|E_L⟩ + β|R⟩|E_R⟩

The environment states |E_L⟩ and |E_R⟩ record which path was taken. As soon as they become (nearly) orthogonal — ⟨E_L|E_R⟩ → 0 — the interference cross-terms between |L⟩ and |R⟩ can no longer be seen in any measurement on the particle alone. The which-path information has leaked out, and by the complementarity principle, leaked-out which-path information means lost interference. That loss is decoherence.

The density matrix: where the interference lives

The cleanest way to see decoherence is the reduced density matrix, ρ, of the system after you trace out (ignore) the environment. For our two-branch system before any interaction:

ρ = |α|²|L⟩⟨L| + |β|²|R⟩⟨R| + αβ*|L⟩⟨R| + α*β|R⟩⟨L|
                                  └──────── coherences (off-diagonal) ────────┘

The diagonal terms |α|² and |β|² are the ordinary probabilities of finding L or R. The off-diagonal "coherences" are what encode interference — they are the part that says the system is in a genuine superposition, not just a coin-flip mixture. Trace out an environment with ⟨E_L|E_R⟩ = γ, and those coherences get multiplied by γ:

ρ_reduced = |α|²|L⟩⟨L| + |β|²|R⟩⟨R| + γ·αβ*|L⟩⟨R| + γ*·α*β|R⟩⟨L|

As the environment branches become orthogonal, γ → 0 and the off-diagonals vanish. The reduced state collapses (in appearance) to a purely diagonal matrix — exactly the form of a classical probability distribution over definite outcomes. Crucially, nothing non-unitary happened: the global system+environment state is still pure. Decoherence is the delocalization of phase information into inaccessible correlations, not its destruction.

	Pure superposition	After full decoherence
Reduced ρ	Has large off-diagonals	Diagonal (classical mixture)
Interference	Visible fringes	Washed out
Purity Tr(ρ²)	= 1 (pure)	< 1 (mixed)
Global state	Pure	Still pure (just entangled)
Information	Local to system	Spread into environment
Reversible?	—	In principle yes, in practice no

The governing equation

For a system weakly coupled to a large, fast, memoryless environment, the reduced density matrix obeys a Lindblad (master) equation. For a free particle monitored in position by scattering, it takes the famous form:

dρ/dt = −(i/ℏ)[H, ρ]  −  Λ (x − x')² ⟨x|ρ|x'⟩

The first term is ordinary unitary Schrödinger dynamics. The second is the decoherence term: the coherence between two positions x and x' decays at a rate proportional to the square of their separation, with a localization rate Λ (the scattering constant). Integrating gives the spatial coherences dying off as

⟨x|ρ|x'⟩(t) = ⟨x|ρ|x'⟩(0) · exp[ −Λ (x − x')² t ]

The key lesson is the (x − x')² scaling. Coherence over a large superposition (big x − x') dies catastrophically faster than over a tiny one. Double the separation, and the decoherence rate quadruples. This single fact explains the entire quantum-to-classical crossover: microscopic separations survive, macroscopic ones do not.

Real numbers: why big things are classical

Joos and Zeh (1985) computed localization rates Λ for a small object scattering various environments. The decoherence time for a coherence spread of Δx is roughly τ_d ≈ 1/(Λ Δx²). For a dust grain (radius ~10 µm) delocalized over Δx = 1 µm:

Environment scattering the dust grain	Decoherence time τ_d
Air at standard pressure	~10-31 s
Laboratory vacuum (10-6 torr)	~10-19 s
Sunlight	~10-21 s
Thermal (300 K) radiation	~10-18 s
Cosmic microwave background (2.7 K)	~1 s

Even the relic glow of the Big Bang would decohere a dust grain in about a second. There is no way to hide a macroscopic object from the environment. Compare that with carefully engineered quantum systems, where the entire game is fighting decoherence:

System	Typical coherence time	Dominant decoherence channel
Macroscopic object (cat, chair)	10-20–10-30 s	Air, photons, phonons
C70 fullerene in interferometer	~10-3 s	Thermal photon emission, gas collisions
Superconducting transmon qubit	~10-4 s	Two-level defects, flux/charge noise
Electron-spin qubit (NV center)	~10-3 s	Nuclear-spin bath, phonons
Trapped-ion qubit	1–50 s	Magnetic-field noise, heating
Nuclear spin in solid	seconds–hours	Dipolar coupling, spin diffusion

Pointer states and einselection

Why does the environment single out position as the classical variable, and not momentum or some weird superposition? Because the system–environment interaction Hamiltonian is itself a function of position — photons scatter off where the object is. The states that commute with (or are least disturbed by) that interaction survive monitoring with the least entanglement. Wojciech Zurek called these the pointer states, and the dynamical process that selects them einselection (environment-induced superselection).

The predictability sieve makes this precise: of all possible initial states, the pointer states are the ones whose entropy production under environmental monitoring is minimal — they remain (approximately) pure and predictable. For everyday objects coupled through position-dependent forces, the pointer states are localized, minimum-uncertainty wave packets that follow Newtonian trajectories. That is the deep reason the classical limit looks like little balls moving on definite paths: those are the states the universe lets persist.

Decoherence is not collapse

It is tempting to declare the measurement problem solved: the environment "measures" the system, the off-diagonals vanish, done. Not quite. Decoherence explains two things superbly:

• The preferred-basis problem. Why we see definite positions/pointer readings rather than superpositions of them — einselection answers this.
• The appearance of classicality. Why interference between macroscopically distinct alternatives is unobservable — the (x−x')² scaling answers this.

But it does not explain why we observe one particular outcome rather than the others. The diagonal density matrix is a mixture as far as local observations go, but the global state is still a grand superposition of all branches. Whether that residual problem is solved by many-worlds (all branches are real, you ride one), by an objective collapse mechanism (GRW, Penrose), or by a Bayesian reading of the state, is interpretation-dependent and still debated. Decoherence narrows the problem; it does not eliminate it.

	Decoherence	Wavefunction collapse
Dynamics	Unitary, continuous	Non-unitary, discontinuous
Selects one outcome?	No	Yes (by postulate)
Reversible in principle?	Yes	No
Derivable from Schrödinger eq.?	Yes	No (extra postulate)
What it explains	Loss of interference, classical basis	Definite single result

Simulating two-level decoherence

A qubit dephasing in the energy basis is the simplest model — only the off-diagonal coherences decay, the populations stay put. Here is the exponential decay of the off-diagonal element and the resulting loss of interference fringe contrast:

// Pure-dephasing of a single qubit: rho is a 2x2 density matrix.
// Diagonal = populations (preserved). Off-diagonal = coherence (decays).

function dephase(rho, t, T2) {
  const decay = Math.exp(-t / T2);          // coherence factor gamma(t)
  return [
    [rho[0][0],            rho[0][1] * decay],
    [rho[1][0] * decay,    rho[1][1]]
  ];
}

// Start in an equal superposition (|0> + |1>)/sqrt(2): full coherence = 0.5
let rho = [[0.5, 0.5],
           [0.5, 0.5]];

const T2 = 100e-6;  // 100 microsecond coherence time (e.g. a transmon)
for (const t of [0, 25e-6, 50e-6, 100e-6, 300e-6]) {
  const r = dephase(rho, t, T2);
  // Fringe contrast (visibility) = 2 * |off-diagonal|
  const visibility = 2 * Math.abs(r[0][1]);
  console.log(`t=${(t*1e6).toFixed(0)} us  coherence=${r[0][1].toFixed(3)}  visibility=${visibility.toFixed(3)}`);
}
// t=0 us    coherence=0.500  visibility=1.000   (full interference)
// t=25 us   coherence=0.389  visibility=0.779
// t=50 us   coherence=0.303  visibility=0.607
// t=100 us  coherence=0.184  visibility=0.368   (1/e point)
// t=300 us  coherence=0.025  visibility=0.050   (essentially classical)

// Spatial decoherence: coherence dies as exp(-Lambda * dx^2 * t).
// Note the SQUARE of the separation -- big superpositions die far faster.
function spatialCoherence(dx, t, Lambda) {
  return Math.exp(-Lambda * dx * dx * t);
}

const Lambda = 1e19;  // localization rate (1/m^2/s), thermal-photon scale
const t = 1e-6;
for (const dx of [1e-9, 1e-7, 1e-6]) {  // nm, 100 nm, 1 um
  const g = spatialCoherence(dx, t, Lambda);
  console.log(`dx=${(dx*1e9).toFixed(0)} nm  coherence=${g.toExponential(2)}`);
}
// dx=1 nm     coherence=1.00e+0    (survives -- microscopic)
// dx=100 nm   coherence=9.05e-1
// dx=1000 nm  coherence=4.54e-5    (gone -- macroscopic separation)

Where decoherence matters

• Quantum computing. Decoherence is the central enemy. Coherence times (T1 relaxation, T2 dephasing) set how many gates you can run; quantum error correction and dynamical decoupling exist to fight it.
• The measurement problem. Decoherence theory (Zeh 1970, Zurek 1981, Joos–Zeh 1985) reframed the quantum-to-classical transition as a dynamical, calculable process rather than a metaphysical postulate.
• Matter-wave interferometry. Experiments with large molecules (C60, C70, and molecules of thousands of atomic mass units) deliberately reduce gas and thermal-photon scattering to push the coherence boundary upward.
• Superconductivity and SQUIDs. Superpositions of macroscopic current states are observable only because the circuit is cooled and shielded to suppress decoherence channels.
• Quantum biology. Whether coherence survives long enough in warm, wet systems (photosynthetic complexes, avian magnetoreception) to play a functional role is a decoherence-rate question.
• Foundations and gravity. Penrose and Diósi propose that gravity itself induces an objective decoherence/collapse, giving a testable timescale tied to mass and separation.

Common misconceptions

• "Decoherence collapses the wavefunction." It does not. The global state stays pure and unitary; only the local reduced state looks mixed. No single outcome is selected.
• "Decoherence is just the wavefunction spreading out." No — that is free dispersion and is reversible. Decoherence is entanglement with external degrees of freedom that you cannot track.
• "Observers cause decoherence." Any environment does — air, photons, phonons. No conscious observer is needed; a single scattered photon is enough.
• "Decoherence destroys information." It delocalizes information into system–environment correlations. The information is still there in the total state, just inaccessible to local measurement.
• "Small and large systems decohere at similar rates." The rate scales as the square of the separation and with the size of the object's coupling to the environment, so macroscopic superpositions decohere astronomically faster than microscopic ones.
• "Coherence and entanglement are the same thing." A system can be coherent (have superposition) without being entangled. Decoherence is precisely the conversion of internal coherence into external entanglement.