The Spin Echo: Rewinding Quantum Dephasing

Larmor precession and the birth of dephasing

Place a spin-½ nucleus — a proton, say — in a static field B₀ and its magnetic moment precesses about that field at the Larmor frequency:

• ω₀ = γB₀, where γ is the gyromagnetic ratio. For ¹H, γ/2π = 42.58 MHz/T, so a 1.5 T clinical MRI runs at ν₀ ≈ 63.9 MHz; a 3 T scanner at ≈ 127.7 MHz.
• The equilibrium magnetization M₀ points along B₀ (call it ẑ). A resonant 90° (π/2) radio pulse tips the whole bulk magnetization into the transverse x–y plane, where it precesses and induces a voltage in the pickup coil — the free induction decay (FID).

Here is the problem. No real magnet is perfectly uniform. Across a sample of even a few centimetres, B₀ varies by parts per million, and each nucleus also feels a slightly different local field from its chemical environment. So spin i precesses at ω₀ + δωᵢ. Within milliseconds the individual spin vectors — all launched together along, say, +y — fan out around the circle like the second hands of a thousand slightly mis-set watches. Their vector sum, the measured signal, collapses to zero. This is dephasing, and its time constant is T₂*, typically ≤10 ms in tissue even though nothing irreversible has yet happened to any single spin.

Hahn's insight: the 180° pulse that reverses time

The genius of the echo is that the fanning-out is deterministic. A spin that is 5 Hz fast has accumulated a definite lead; one that is 5 Hz slow has a definite lag. If you could flip the whole fan over a mirror line, the leaders would become laggers and vice versa — and since each keeps its own precession rate, they would all converge back to a point.

That mirror flip is exactly a 180° (π) pulse. The canonical Hahn echo sequence is:

• Apply a 90°ₓ pulse at t = 0 → magnetization into the transverse plane; spins begin to fan.
• Wait a time τ. Spin i has now advanced a phase φᵢ = δωᵢ·τ relative to the mean.
• Apply a 180°ᵧ pulse. This rotates each spin about the y-axis, sending phase φᵢ → −φᵢ. The fast spins are now behind, the slow ones ahead.
• Wait another τ. Each spin advances by the same δωᵢ·τ again — but now this precisely cancels its flipped phase: −φᵢ + δωᵢ·τ = 0.

At t = 2τ ≡ TE (the echo time) every spin is back in phase, the magnetization re-coheres, and the coil sees a revived signal — the echo — that rises, peaks, and decays symmetrically. Crucially, the sequence never needed to know the individual δωᵢ. It refocuses any static frequency spread automatically, which is why it is so powerful.

The runners-on-a-track intuition

The cleanest mental picture, and the one Hahn himself used, is a race. Fire a starting gun (the 90° pulse) and a field of runners sets off around a track at slightly different speeds. After time τ they are strung out — fast ones ahead, slow ones behind — and the pack has spread so thin that, viewed as a crowd, there is no coherent bunch left. That is the dead FID.

• Now blow a second whistle at t = τ (the 180° pulse) that means "turn around and run back at the same speed."
• The runner who got farthest ahead now has the farthest to return; the slowpoke who lagged is closest to the finish. Running for another equal interval τ, every runner arrives at the starting line at the same instant, regardless of their speed.

The echo is that moment of simultaneous arrival. Notice what the picture also teaches: if a runner randomly stumbles and changes pace after the whistle — an unpredictable event — the reversal no longer works for them. Those random pace changes are the irreversible part. The whistle only rewinds the part of the spreading that came from each runner having a fixed, unchanging speed.

T₂* versus T₂: what the echo can and cannot save

The echo cleanly separates two kinds of signal loss, and confusing them is the single most common misconception.

• T₂* — reversible dephasing. Caused by static, spin-specific frequency offsets: magnet inhomogeneity ΔB₀, susceptibility gradients, chemical-shift spread. Because each offset is constant in time, its accumulated phase is exactly undone by the 180° flip. The FID decays with 1/T₂* ≈ 1/T₂ + γΔB₀/2, and T₂* can be far shorter than T₂.
• T₂ — irreversible decoherence ("true" transverse relaxation). Caused by fluctuating local fields: neighbouring spins flipping, molecular tumbling modulating dipolar couplings, diffusion through field gradients. Between the two τ intervals a spin's frequency changes unpredictably, so the phase it picks up before the flip is not the phase it un-picks after. This part is genuinely lost.

So the peak amplitude of successive echoes does not recover to the original — it follows the true envelope M(2τ) = M₀ · exp(−2τ/T₂). By stepping τ and plotting echo height, you measure the real T₂ while cleanly stripping away the magnet's imperfections. In water at room temperature T₂ can reach several seconds; in brain grey matter T₂ ≈ 80–100 ms at 1.5 T while T₂* might be only ~50 ms; near iron-rich deoxyhaemoglobin or an implant, T₂* crashes to milliseconds while T₂ stays long — the basis of the BOLD contrast in functional MRI.

Carr–Purcell, Meiboom–Gill, and beating diffusion

One echo is good; a train of them is better. In 1954 Carr and Purcell applied a string of 180° pulses — 90° – τ – (180° – 2τ –)ⁿ — producing an echo after every pulse and letting you watch the entire T₂ decay in a single shot instead of one point per experiment. This is the Carr–Purcell (CP) sequence.

• The imperfect-pulse problem. Real 180° pulses are never exactly 180°; a 2° error compounds catastrophically over hundreds of echoes. In 1958 Meiboom and Gill fixed this by phase-shifting the refocusing pulses 90° relative to the excitation pulse and applying them about the y-axis. Now flip errors alternate in sign and self-correct on even echoes. The result — CPMG — is the standard robust T₂ measurement everywhere from benchtop NMR to oil-well logging tools.
• Beating diffusion. A single Hahn echo with long τ lets molecules diffuse into different field regions between refocus and echo, adding an extra, τ³-dependent decay. Packing many closely spaced 180° pulses (short τ) gives each molecule little time to wander, suppressing the diffusion loss and revealing a longer, truer T₂. Deliberately tuning that sensitivity, by contrast, is exactly how diffusion-weighted MRI and pulsed-field-gradient NMR measure molecular self-diffusion coefficients (~2.3 × 10⁻⁹ m²/s for water at 25 °C).

Echoes everywhere: from MRI voxels to living qubits

The spin echo is not a laboratory curiosity — it is embedded in technologies used millions of times a day.

• MRI. Spin-echo sequences produce the clean T₂-weighted images radiologists love because they are immune to the static B₀ inhomogeneity that plagues gradient-echo scans. TE and TR (repetition time) are the two knobs that dial contrast: long TE, long TR → bright fluid (T₂ weighting); short TE, short TR → T₁ weighting.
• NMR spectroscopy and MRI quantitation. CPMG lets chemists strip inhomogeneous broadening to resolve true linewidths, and lets physicists measure T₂ to characterize samples and molecular motion.
• Quantum computing. A qubit is just a two-level spin, and its free-induction (Ramsey) coherence time T₂* is often microseconds, throttled by slow environmental noise (nuclear spin baths, 1/f flux noise). A single Hahn echo pushes coherence out to T₂; long dynamical-decoupling trains (CPMG, XY-8, Uhrig sequences) filter out low-frequency noise and can extend usable coherence by 10–100×. NV centres in diamond, superconducting transmons, and trapped ions all rely on echo-based decoupling to protect quantum information.
• Photon and other echoes. The same refocusing idea appears optically as the photon echo (1964) in inhomogeneously broadened atomic ensembles, underpinning proposed optical quantum memories, and in plasmas as cyclotron and plasma-wave "echoes."

Subtleties, limits, and honest bookkeeping

Because the echo looks like it "reverses time," it invites overreach. A few clarifications keep the physics honest.

• It does not violate the second law. The refocused phases were never truly randomized — the information was stored, deterministically, in each spin's fixed offset. The echo simply reads it back. Genuine entropy production (T₂ processes) is not recovered, and the echo amplitude strictly decays. Hahn-style echoes are a favourite illustration of the difference between apparent and real irreversibility, closely related to Loschmidt's reversibility paradox.
• The 180° pulse costs energy and adds error. Pulse imperfections, off-resonance effects, and finite pulse duration all leak coherence; CPMG/MG phasing exists precisely to tame this.
• Stimulated echoes. Three 90° pulses (90–τ–90–T–90) store magnetization along ẑ during the middle interval T, where it decays with the usually much longer T₁ rather than T₂ — a trick for probing very slow motions and long-time correlations.
• Order of magnitude to remember. γ/2π = 42.58 MHz/T for protons means a mere 1 ppm field spread over the sample (ΔB₀ ≈ 3 μT at 3 T) already gives a ~128 Hz frequency spread — enough to dephase the FID in a few milliseconds. That fragility is why, before Hahn, the transverse signal seemed hopelessly lost, and why a single well-timed pulse to bring it back felt like sorcery.

Reversible dephasing (T₂*) versus irreversible decoherence (T₂) — the echo recovers only the first.

Property	T₂* (FID decay)	T₂ (true decoherence)
Physical origin	Static field inhomogeneity ΔB₀, chemical-shift spread	Fluctuating fields, spin–spin flips, molecular tumbling
Reversible?	Yes — echo refocuses it	No — echo cannot recover it
Typical value (¹H tissue)	~10 ms or less	~40–200 ms
Relation	1/T₂* = 1/T₂ + γΔB₀/2 (approx.)	Sets the true envelope of echo amplitude
What kills it	Magnet shimming can improve it	Fundamental — only sequence tricks slow it