NOON States: Super-Resolution at the Heisenberg Limit

What a NOON state is and why it matters

A NOON state is the two-mode, N-photon entangled superposition

|ψ⟩ = (|N,0⟩ + |0,N⟩)/√2,

meaning all N photons occupy interferometer arm A or all N occupy arm B, with no term in which the photons split between arms. It is the maximally path-entangled state of fixed photon number — a Greenberger–Horne–Zeilinger (GHZ) state written in the photon-number (Fock) basis of two modes.

Its importance is that it saturates the ultimate quantum bound on phase estimation. In any interferometer, one arm imprints a phase φ; reading φ precisely is the core task of gravitational-wave detectors, gyroscopes, magnetometers, and lithography. With classical light of mean photon number N̄, the best achievable resolution is the standard quantum limit (SQL), Δφ ∼ 1/√N̄, set by the Poissonian shot noise of independent photons. Entangling the photons so they act as a single quantum object of N-fold coherence lets the NOON state reach the Heisenberg limit Δφ ∼ 1/N — a genuine √N advantage that is the canonical demonstration of a quantum resource beating a classical protocol.

The mechanism, step by step

Follow the phase. In arm A the photons pick up phase per photon, so the whole N-photon amplitude acquires e^{iNφ}: the branch |N,0⟩ → e^{iNφ}|N,0⟩, while |0,N⟩ is unchanged. The relative phase between the two branches is therefore Nφ, not φ — the entire super-resolution effect springs from this single fact. The N indistinguishable photons behave as one particle of N times the momentum, giving an effective de Broglie wavelength λ_dB = λ/N.

To read that N-fold phase you cannot use ordinary single-photon detection; the enhanced fringe lives only in the N-photon coincidence. One recombines the modes and measures the parity or the N-fold coincidence probability. The probability of detecting all N photons in one output port becomes

P_N(φ) = ½ [1 + cos(Nφ)],

which oscillates N times as fast as the classical (1 + cos φ)/2 fringe. Each fringe is N times narrower in φ, so a given phase shift moves the signal N times further — that is the super-resolution. Crucially, the phase information is carried collectively; no single photon knows φ, only the joint N-photon correlation does.

The key equation, characteristic numbers, and scales

The sensitivity follows from error propagation on the N-fold fringe. With signal ⟨Â⟩ oscillating as cos(Nφ), the slope is N times steeper while the shot noise per shot is unchanged, so

Δφ = ΔÂ / |∂⟨Â⟩/∂φ| = 1/N.

This is the Heisenberg limit, the √N improvement over the SQL 1/√N. Equivalently, the quantum Fisher information for a NOON state is F_Q = N², versus F_Q = N for N independent photons; since Δφ ≥ 1/√F_Q (the quantum Cramér–Rao bound), the NOON state saturates 1/N.

Characteristic numbers: the fringe period shrinks from 2π to 2π/N; in quantum lithography (Boto et al., 2000) the writable feature size shrinks from the Rayleigh λ/2 to λ/(2N). For N = 5 photons at λ = 800 nm, the effective wavelength is 160 nm and features down to ~80 nm are in principle addressable. The offsetting cost is loss: if each photon survives with probability η, the N-photon fringe visibility falls as η^N, so the whole payload must arrive intact.

How NOON states are realized and measured

Small NOON states arise naturally from Hong–Ou–Mandel interference: two indistinguishable photons meeting at a 50:50 beam splitter bunch into (|2,0⟩+|0,2⟩)/√2, the N=2 NOON state, and its cos(2φ) fringe was an early super-resolution demonstration. Pushing to higher N is hard because simply feeding more spontaneous-parametric-down-conversion (SPDC) photons into linear optics gives the wrong (non-NOON) superpositions with vanishing efficiency.

The breakthrough was Afek, Ambar, and Silberberg (Weizmann Institute, Science 2010): they mixed quantum down-converted light with a classical coherent beam at a beam splitter, tuning their relative amplitude and phase so that unwanted terms interfere away. This hybrid scheme is inherently scalable and reached N = 5 — the highest NOON state to date — with super-resolving fringes exceeding the classical visibility ceiling and a projected fidelity of ~92% for arbitrary N. Detection uses N-fold coincidence counting or photon-number-resolving detectors; the signature is a sinusoid oscillating N times per 2π of the scanned phase, sharper than any classical light of the same intensity could produce.

Where it operates, and how it differs from related states

NOON states are the extreme point of a family of quantum-metrological probes. They live wherever an interferometric phase must be read at fixed, small photon number and low loss: photonic phase estimation, optical gyroscopes, phase microscopy, and two-photon/N-photon lithography. Related to them, but distinct, are spin-squeezed and squeezed-vacuum states, which give sub-shot-noise (though generally sub-Heisenberg) sensitivity with far greater loss tolerance — the route actually used in gravitational-wave detectors like LIGO, which inject squeezed vacuum rather than NOON states.

Formally the NOON state is a two-mode GHZ/'cat' state; it is maximally entangled and therefore maximally fragile. It should not be confused with the N-photon de Broglie effect in general (which NOON states realize most sharply), nor with 'super-sensitivity' alone — a state can be super-resolving (narrow fringes) without being super-sensitive (better Δφ), and only entanglement delivers both. Optimized states such as Holland–Burnett (twin-Fock) states sacrifice a little of the ideal 1/N to gain robustness, sitting between coherent and NOON states.

Applications, open problems, and significance

NOON states are the clean proof-of-principle that entanglement beats the shot-noise limit, and they seed real applications: quantum lithography aiming below the diffraction limit, super-resolving phase and polarization microscopy that limits photon dose on delicate or biological samples, quantum-enhanced spectroscopy, and precision refractometry. They have also been generated beyond photons — in trapped ions, superconducting microwave cavities, and nitrogen-vacancy spin ensembles — as a benchmark of many-body coherence.

The open problem is loss and decoherence. Losing or measuring a single photon reveals which-arm information and collapses |ψ⟩ to a classical mixture with no phase enhancement, and visibility scales as η^N, so at realistic η the Heisenberg advantage evaporates before N grows large — the reason no NOON state above N=5 has been demonstrated. In lossy channels the true optimum is not the NOON state; the sensitivity crosses over from 1/N back to 1/√N scaling as loss increases. Active research pursues loss-tolerant metrological states, adaptive and error-corrected NOON protocols, and hybrid squeezed/entangled probes, so the state's ultimate legacy may be conceptual — defining the Heisenberg limit that every practical scheme is measured against.

Classical (coherent state) vs. NOON-state interferometry, and the scaling that separates them

Property	Coherent / classical light	N-photon NOON state
State	|α⟩, Poissonian photon number	(|N,0⟩+|0,N⟩)/√2, definite total N
Fringe period in phase φ	2π (single-photon fringe)	2π/N (N-fold super-resolution)
Effective wavelength	λ	λ/N (N-photon de Broglie)
Phase uncertainty Δφ	1/√N̄ (standard quantum / shot-noise limit)	1/N (Heisenberg limit)
Sensitivity gain	baseline	factor √N over SQL
Loss sensitivity	graceful, √η per photon	catastrophic: one lost photon collapses state; visibility ∝ η^N