Atomic, Molecular & Optical Physics
Electromagnetically Induced Transparency: Opening a Window in an Opaque Atom
Shine a strong "control" laser on a gas of atoms and a probe beam that a moment earlier was completely absorbed suddenly passes through untouched — and crawls forward at 17 metres per second, roughly 20 million times slower than light in vacuum. That is electromagnetically induced transparency (EIT): a quantum-interference effect in which one laser field renders an otherwise opaque atomic medium transparent to a second, resonant field over a narrow spectral window.
The transparency is not saturation or bleaching. It is destructive interference between two excitation pathways that drives the atoms into a "dark state" — a coherent superposition of long-lived ground levels that simply does not couple to light. In the same window the medium acquires an enormously steep, dispersion, slashing the group velocity of the probe and, in the limit, freezing and storing the pulse as an atomic spin wave.
- RegimeCoherent, resonant AMO physics: 3-level Λ atom + 2 laser fields, two-photon Raman resonance
- Key relationTransparency window width ΔωEIT ≈ Ωc²/(γ√(Nσℓ)); group velocity vg ≈ ℏc ε₀ Ωc² / (2 |℘|² N)
- DiscoveredTheory: Harris, Field & Imamoğlu 1990; first observed in Sr vapour by Boller, Imamoğlu & Harris 1991
- Characteristic scalevg ≈ 17 m/s (~c/2×10⁷); windows kHz–MHz wide; storage times up to ~1 s
- Realized inUltracold Na BEC (Hau 1999), warm Rb/Cs vapour cells, cold-atom clouds, solid-state rare-earth crystals, superconducting circuits
- Matters forQuantum memories, slow/stopped light, nonlinear optics at single photons, precision magnetometry & atomic clocks
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What EIT is and why it matters
EIT is the phenomenon in which an atomic (or atom-like) medium that strongly absorbs light at a given transition becomes almost perfectly transparent to a weak probe beam when a second, stronger control beam is applied on a coupled transition. First proposed by Stephen Harris, Jonathan Field and Atac Imamoğlu in 1990 and observed a year later in strontium vapour, EIT is a purely coherent effect: it survives even for a vanishingly weak probe and requires no population change in the medium.
Its importance is twofold. First, it defeats the usual trade-off in optics between transparency and dispersion. Ordinarily a medium is dispersive only where it absorbs (Kramers–Kronig). EIT places a narrow window of near-zero absorption inside a region of extraordinarily steep normal dispersion. Second, that steep dispersion is dynamically controllable through the control-field intensity. This makes EIT the workhorse for slow light, stopped light, single-photon nonlinear optics, and quantum memories — regimes where you want strong light–matter interaction without dissipative loss.
The mechanism, step by step: interference into a dark state
Take the canonical Λ (lambda) system: two long-lived ground levels |g⟩ and |s⟩ and one excited level |e⟩ that decays fast at rate γ. The probe couples |g⟩→|e⟩ (Rabi frequency Ωp); the control couples |s⟩→|e⟩ (Rabi frequency Ωc). Normally the probe is absorbed because |g⟩→|e⟩ is resonant.
Now diagonalise the coupled system. There is a special ground-state superposition, the dark state |D⟩ ∝ Ωc|g⟩ − Ωp|s⟩, which has zero dipole coupling to |e⟩: the two amplitudes for reaching |e⟩ interfere destructively (a Fano-type interference). Optical pumping drives the atoms into |D⟩, and once there they cannot absorb — the medium is transparent. The complementary "bright" state does absorb, but it is depopulated. Crucially the effect is sharpest at two-photon resonance, δ = ωp − ωs − (ωge − ωse) = 0, where the ground-state coherence ρgs is maximised. Any decoherence of that ground-state coherence (rate γgs) fills in the window, so long-lived |g⟩–|s⟩ coherence is the essential resource.
The key relations, window width and slow-light scaling
The probe susceptibility near two-photon resonance takes the form χ(δ) ∝ δ / (δ(iγ/2 + δ) − Ωc²/4). Its imaginary part (absorption) vanishes at δ = 0, giving the transparency; its real part passes through zero with a large positive slope dn/dω. The window's spectral width is set by the control field:
ΔωEIT ≈ Ωc² / (γ √(αℓ)) for an optically thick sample of resonant opacity αℓ = Nσℓ (with N the density, σ the cross-section). The steep dispersion yields a group velocity
vg ≈ ℏ c ε₀ Ωc² / (2 |℘|² N) ≪ c,
so vg falls as Ωc² and rises with lower density. In Lene Hau's 1999 sodium BEC, Ωc, density and temperature conspired to give vg ≈ 17 m/s — about c/2×10⁷ — with corresponding pulse delays of milliseconds over a sub-millimetre cloud. Typical warm-vapour windows are kHz–MHz wide; the delay–bandwidth product (roughly the storable pulse count) is set by the optical depth.
How it is realized and measured
The standard signature is a probe-absorption spectrum: sweep the probe detuning with the control on, and a sharp transmission peak (the EIT window) appears inside the broad absorption line — a Lorentzian dip narrowed far below the natural linewidth γ. In steady state one measures the peak transmission and window width; in the time domain one sends a probe pulse and measures its group delay.
Platforms span the field. The cleanest are alkali systems: warm rubidium or caesium vapour cells (often with a buffer gas or paraffin coating to protect ground-state coherence), cold-atom clouds in magneto-optical traps, and the ultracold sodium and rubidium condensates used for record slow light. Hyperfine or Zeeman sublevels supply the two ground states. Beyond gases, EIT-like windows appear in rare-earth-doped crystals (Pr:YSO), in nitrogen-vacancy centres, in optomechanical resonators ("optomechanically induced transparency", Nature 2011), in metamaterials, and in superconducting artificial atoms — any three-level system with a coherent path interference reproduces the physics.
Where it operates, and how it differs from look-alikes
EIT operates wherever you have (i) a Λ-type level structure, (ii) a strong control field, and (iii) ground-state coherence that lives long compared with 1/ΔωEIT. It is often confused with two neighbours. Autler–Townes splitting also produces a dip, but there the strong field simply dresses |e⟩ into two shifted peaks; the gap is a spacing, not an interference minimum, and it requires Ωc ≫ γ. Genuine EIT persists in the weak-control regime and is destroyed by dephasing — Anisimov, Clark and others devised criteria to distinguish the two from lineshapes.
Coherent population trapping is essentially the same dark-state physics in the symmetric two-strong-field limit, pioneered by Alzetta and Arimondo in the mid-1970s; EIT is the asymmetric, weak-probe descendant. The sign-reversed cousin, electromagnetically induced absorption, arises when coherence transfer interferes constructively instead. Ordinary saturation bleaching, by contrast, is incoherent, intensity-hungry and broad.
Applications, open questions and significance
EIT's headline application is the quantum memory. Fleischhauer and Lukin (2000) showed the slow-light pulse is a dark-state polariton — a coherent mixture of photon and atomic spin wave whose mixing angle is set by Ωc. Ramp Ωc adiabatically to zero and the polariton becomes purely atomic: the light is stopped and stored as a spin coherence, then retrieved on demand by turning Ωc back up. This was demonstrated in 2001 (Phillips/Walsworth/Lukin in warm Rb; Liu/Hau in cold Na), with storage times since pushed to ~1 second and efficiencies above 90% in optimized systems — the basis for photonic quantum memories and repeaters.
Beyond memory, EIT enables giant resonant Kerr nonlinearities for single-photon gates and Rydberg-EIT photon blockade, ultra-narrow-line magnetometry and CPT atomic clocks, and precision spectroscopy. Open frontiers include maximizing the delay–bandwidth (optical-depth) product, suppressing residual four-wave mixing and ground-state dephasing, and engineering strong photon–photon interactions for deterministic quantum logic and many-body photonic states.
| Effect | Physical origin | Fields required | Signature / hallmark |
|---|---|---|---|
| Electromagnetically induced transparency (EIT) | Fano-type destructive interference of excitation pathways → dark state | Weak probe + strong coherent control (coupling) field | Sharp transparency dip in absorption with steep normal dispersion; slow light |
| Coherent population trapping (CPT) | Same dark-state physics, symmetric two-field limit | Two comparable coherent fields on a Λ system | Non-absorbing coherent superposition; used in CPT clocks/magnetometers |
| Autler–Townes splitting (AC Stark) | Level dressing by strong field, two shifted absorption peaks | Strong control field (Ωc ≫ γ, no interference needed) | Two resolved peaks; transparency is a gap, not interference |
| Saturation / hole burning | Population depletion of the ground state | Single intense field | Bleaching that scales with intensity; broad, incoherent |
| Electromagnetically induced absorption (EIA) | Constructive interference / transfer of coherence | Two fields on suitable degenerate levels | Enhanced absorption peak (opposite sign of EIT) |
Frequently asked questions
How is EIT different from simply saturating or bleaching the transition?
Saturation is an incoherent depletion of the ground-state population that requires substantial intensity in the beam being transmitted and grows with that intensity. EIT is coherent: it works for an arbitrarily weak probe and is driven by a separate control field. The transparency comes from destructive quantum interference into a dark state, not from emptying the absorbing level, and it produces a much narrower, dispersion-rich window than bleaching ever could.
Why does EIT make light travel so slowly?
The transparency window sits in a region of very steep normal dispersion, dn/dω ≫ 0. Group velocity is vg = c/(n + ω dn/dω), so a large positive slope collapses vg. Because the slope scales as 1/Ωc², reducing the control-field intensity steepens the dispersion and slows the light further, down to Hau's 17 m/s in a sodium condensate and, in the limit Ωc→0, to a full stop.
What exactly is the 'dark state'?
In a Λ system it is the ground-state superposition |D⟩ ∝ Ωc|g⟩ − Ωp|s⟩ whose two amplitudes for exciting the common upper level |e⟩ cancel by destructive interference. An atom in |D⟩ therefore cannot absorb either field — it is decoupled from the light. Optical pumping funnels the population into |D⟩, and the medium becomes transparent. It is an eigenstate of the interaction Hamiltonian with zero light-coupling eigenvalue.
How do you tell EIT apart from Autler–Townes splitting?
Both give a dip in the absorption line, but the mechanisms differ. Autler–Townes is AC-Stark dressing: a strong control field (Ωc ≫ γ) splits the upper level into two peaks, and the dip is the gap between them. True EIT is a Fano interference that persists even for a weak control field and is destroyed by ground-state dephasing. Lineshape-decomposition tests (Anisimov–Clark type) quantitatively distinguish the interference contribution from the splitting.
What limits the EIT transparency and the storage time?
The critical limit is decoherence of the two-photon (ground-state) coherence at rate γgs — from atomic motion, collisions, magnetic-field inhomogeneity, or finite laser linewidth. It sets a minimum window width and fills in the transparency. In vapour cells, buffer gases and anti-relaxation wall coatings extend the coherence; in cold atoms, low temperature does. Storage times are ultimately capped by γgs, reaching ~1 second in the best-protected room-temperature and cold systems.
Can EIT store a single photon, and is that useful?
Yes. Since the slow-light pulse is a dark-state polariton — part photon, part collective spin wave — adiabatically switching off the control field converts it entirely into an atomic spin excitation, storing the quantum state, then retrieving it on demand. This gives a bandwidth-matched, on-demand photonic quantum memory, a key building block for quantum repeaters, single-photon nonlinear gates (especially with Rydberg EIT), and synchronization of photonic quantum networks.