Why Remote Entanglement?
The scaling wall is real — and photons are the only way through it.
The single-array bottleneck
The Rydberg blockade — the mechanism behind every neutral-atom two-qubit gate — operates over a range of 5–15 μm. This means that only atoms within a few micrometres of each other can be entangled locally. A single optical tweezer array can hold ~1,000–3,000 atoms today, and even optimistic projections push this to ~10,000 in a single vacuum chamber with current technology. Fault-tolerant quantum computing for practical applications needs millions.
The solution adopted by every serious neutral-atom quantum computing company is modular architecture: build many smaller, higher-quality arrays and connect them with quantum links. The link between modules is necessarily a photonic interconnect — photons are the only quantum system that can travel macroscopic distances without losing coherence in flight.
What remote entanglement enables
- Modular scaling: Chain N tweezer arrays, each with M atoms. The effective system size grows as N × M, but each array can be individually optimised.
- Teleportation-based gates: A shared Bell pair between modules enables a deterministic quantum gate between one qubit in each module via gate teleportation — even without direct interaction.
- Quantum networks: The same photonic link that connects two modules in a lab can connect two labs across a city — enabling distributed quantum computing and quantum key distribution.
- Quantum repeaters: Long-distance quantum communication requires intermediate "repeater" nodes with quantum memories. Neutral-atom tweezers are ideal quantum memories.
- Heterogeneous architectures: Photonic links can connect different qubit platforms (atoms + superconductors) allowing each to do what it does best.
The Protocol — Step by Step
How two atoms that have never interacted become maximally entangled — one photon at a time.
Barrett-Kok / Two-Photon Heralded Entanglement Protocol
Success probability
Each attempt succeeds with probability:
$$p_{\rm success} = \frac{\eta_A \cdot \eta_B}{2}$$
where $\eta_A, \eta_B$ are the total photon detection efficiencies (collection × fiber × detector). With $\eta \approx 5\%$, $p \approx 0.125\%$ — but at a repetition rate of 1 MHz, this yields ~1,250 Bell pairs per second. With cavities ($\eta \approx 50\%$), this reaches ~125,000 pairs/second.
Why the protocol is heralded (not post-selected)
A key feature: you know whether the protocol succeeded without measuring the atoms. The coincident photon detection at D₁ and D₂ is the herald signal — it tells you the atoms are entangled without revealing which state they're in. This is fundamentally different from post-selection, which would collapse the atomic state. Failed attempts (no coincidence) simply mean you try again, with the atoms reset.
The atoms are left in one of the Bell states $|\Phi^\pm\rangle$ depending on the detector pattern. The phase $\theta$ can be tracked and corrected with single-qubit rotations. The state of the photons is irrelevant after detection — entanglement is established in the atoms.
Atom–Photon Entanglement
The fundamental resource: a single atom emitting a photon whose polarization is entangled with the atom's spin.
How it works: selection rules as the quantum interface
When an atom in state $|\uparrow\rangle$ (e.g., $m_F = +1$) decays to the ground state, it must emit a $\sigma^+$ photon to conserve angular momentum. When in $|\downarrow\rangle$ ($m_F = -1$), it emits $\sigma^-$. Prepare the atom in a superposition, and the decay process creates an entangled atom-photon state:
$$|\Psi\rangle = \frac{1}{\sqrt{2}}\!\left(|\uparrow\rangle\otimes|{\sigma^+}\rangle + |\downarrow\rangle\otimes|{\sigma^-}\rangle\right)$$
This is a maximally entangled state between the atom spin and the photon polarization. The photon then travels to the BSM station — carrying quantum information about the atom's state without ever "measuring" it.
Time-bin encoding (more practical)
Polarization encoding over optical fibers degrades due to birefringence (polarization rotation by stress, temperature). A more robust scheme uses time-bin encoding:
- $|\uparrow\rangle \rightarrow$ photon in early time window (E)
- $|\downarrow\rangle \rightarrow$ photon in late time window (L)
- Entangled state: $(1/\sqrt{2})(|\uparrow, E\rangle + |\downarrow, L\rangle)$
- Early/late defined by the timing of the excitation pulse on each atom
Time-bin qubits are insensitive to fiber birefringence and are the standard encoding for long-distance quantum links.
Photon indistinguishability — the hidden requirement
For the Bell state measurement to work, the photon from atom A and the photon from atom B must be perfectly indistinguishable — identical in frequency, bandwidth, temporal profile, and polarization. If they differ even slightly, the HOM interference is imperfect and entanglement fidelity drops.
The indistinguishability $M$ is related to the visibility of Hong-Ou-Mandel interference. The resulting Bell state fidelity is bounded by:
$$\mathcal{F} \leq \frac{1 + M}{2}$$
What degrades indistinguishability for neutral atoms:
- Motional broadening: A warm atom Doppler-shifts the emitted photon. Ground-state cooling narrows the emission spectrum dramatically — another reason cooling matters for QC.
- Inhomogeneous light shifts: Trapping laser shifts the transition frequency by an amount that varies across the array. Requires magic wavelengths or careful trap design.
- Spectral diffusion: Laser noise, magnetic field fluctuations, and phonon coupling all cause shot-to-shot frequency jitter.
Bell State Measurement & Hong-Ou-Mandel Effect
The quantum magic at the beamsplitter — and how imperfect photons degrade entanglement fidelity.
Hong-Ou-Mandel (HOM) effect
When two identical photons enter a 50-50 beamsplitter from different input ports, quantum interference forces both photons to exit the same port — they always bunch. No coincidence is detected. This is the HOM effect (1987, Nobel Prize 2022 to Aspect, Clauser, Zeilinger).
For the BSM to work, we need photons from atoms A and B to bunch when they're in the same polarization state and anti-bunch (leave separately) when in orthogonal states. The coincidence probability as a function of time delay $\tau$ between the two photons is:
$$P_{\rm coinc}(\tau) = \frac{1}{2}\!\left[1 - \mathcal{M}\cdot e^{-|\tau|/\tau_c}\right]$$
where $\mathcal{M}$ is the photon indistinguishability (0 = distinguishable, 1 = identical) and $\tau_c$ is the photon coherence time $\tau_c = 1/\Delta\nu$. The dip at $\tau = 0$ is the HOM dip — deeper dip = more identical photons = higher entanglement fidelity.
Interactive HOM Dip
Coincidence rate vs arrival-time delay τ between photons from atom A and atom B. The HOM dip depth directly measures $\mathcal{M}$ and sets the maximum Bell state fidelity.
Linear optics BSM and its 50% limitation
A linear optics Bell state measurement using only beamsplitters and phase shifters can at most distinguish two of the four Bell states $|\Phi^\pm\rangle$ and $|\Psi^\pm\rangle$ with certainty. The other two give ambiguous outcomes. This imposes a fundamental 50% efficiency ceiling on the BSM — meaning only half of all entanglement attempts that produce photons result in an unambiguous herald. The remaining 50% are discarded (not errors — just failures that are retried). Overcoming this requires either photon-number resolving detectors in specific configurations, or ancilla photons (which add complexity). In practice, the 50% BSM efficiency is accepted and compensated by increasing the repetition rate.
Challenges Specific to Neutral Atoms
Neutral atoms are excellent quantum computers — but photon emission is not their natural language.
1. Collection efficiency
Ions in Paul traps can be engineered to sit at the focus of a high-NA mirror. Neutral atoms in tweezers face a fundamental geometric challenge: the lens collecting the photon is on one side, and the trapping laser also comes from that side. Without a cavity, collection efficiency is:
$$\eta_{\rm coll} = \frac{1 - \sqrt{1-\text{NA}^2}}{2} \approx \frac{\text{NA}^2}{4}$$
For NA = 0.5 (typical tweezer): $\eta_{\rm coll} \approx 7\%$. Most photons escape sideways. Cavities can boost this to 50–80% but add considerable complexity.
2. Telecom wavelength mismatch
Alkali atoms (Rb at 780 nm, Cs at 852 nm) emit at wavelengths with 3–5 dB/km fiber loss. At 10 km, only 0.1–0.3% of photons survive. Long-distance links require quantum frequency conversion (QFC) to telecom wavelengths (1310 nm or 1550 nm, where loss is 0.3–0.35 dB/km). Alkaline-earth atoms (Sr at 461/689 nm, Yb at 399/578 nm) face even worse fiber transmission without QFC. State-of-the-art QFC achieves 70–90% internal efficiency.
3. Memory coherence vs communication time
After the atom emits a photon, it must wait for:
- Photon transit time: L/c (10 km → 50 μs)
- Classical herald to return: another L/c
- Total latency: 2L/c — the "round-trip" classical communication time
At L = 10 km, this is ~100 μs. The atom must remain coherent for this long — easy for nuclear-spin qubits ($T_2 \sim$ seconds), challenging for microwave-dressed states. The separation between communication qubit and memory qubit addresses this.
4. Spectral indistinguishability
Atoms A and B emit photons at slightly different frequencies due to different trapping potentials, residual magnetic fields, and AC Stark shifts. Two photons with spectral overlap less than 1 produce a reduced HOM visibility. Requires active stabilisation: matched trap depths, common-mode field cancellation, optical cavities to narrow linewidth, or real-time spectral demultiplexing.
5. The duty cycle problem
Remote entanglement is generated probabilistically. While waiting for a successful BSM herald, the atoms must be held in the trap without performing any other operations. The duty cycle for entanglement generation is:
$$\text{duty cycle} = p_{\rm success} \times f_{\rm rep} \times \tau_{\rm attempt}$$
With $p \approx 0.1\%$ and $f_{\rm rep} = 1$ MHz, about 1 ms is spent per Bell pair — during which atoms in other modules sit idle, reducing overall computational throughput.
6. No native strong coupling to photons
Unlike NV centers (spin-selective emission) or quantum dots (Purcell-enhanced emission into a mode), neutral alkali atoms in free space are optically weak — the atom-photon cooperativity is $C = g^2/(\kappa\gamma) \ll 1$ without a cavity. This means each excitation has a low probability of producing a useful photon. Alkaline-earth atoms with narrow "clock" transitions have different challenges: long excited-state lifetimes mean slower photon production rates.
Approaches to Photon Collection & Emission
The field has converged on several complementary strategies, each with distinct engineering trade-offs.
The most direct approach: use the highest available NA objective (0.5–0.9) to collect emitted photons and couple them into single-mode fiber. Requires careful mode matching between the objective's focal mode and the fiber's LP01 mode. A parabolic mirror around the atom can nearly double the collection efficiency by capturing photons from both sides.
- No moving parts, no cavity alignment, integrates naturally with tweezer control optics.
- Collection efficiency limited by NA. For NA = 0.65: $\eta_{\rm coll} \approx 12\%$. Single-mode fiber coupling reduces this to ~6–8% total.
- Harvard (Covey → now UIUC): Demonstrated single-photon collection from single Cs atoms in tweezers with improved fiber coupling efficiency. Lukin group: Working on free-space Rb photon collection in tweezer arrays.
- Key challenge: With $\eta \sim 8\%$, the two-photon BSM rate is $\eta^2/2 \approx 0.3\%$ — about 3,000 Bell pairs/second at 1 MHz repetition rate. Adequate for demonstrations; marginal for FTQC requirements.
A high-finesse optical cavity surrounding the atom enhances emission into the cavity mode via the Purcell effect. The fraction of photons emitted into the cavity mode (the "beta factor") is:
β = C / (1 + C)
where $C = g^2/(\kappa\gamma)$ is the cooperativity. For $C = 10$: $\beta = 91\%$. The photon exits the cavity mirror and can be coupled directly into single-mode fiber with 60–80% efficiency, giving total $\eta \sim 55–72\%$.
- Rempe group (MPQ): Pioneered this approach with Rb in macroscopic Fabry-Pérot cavities. Demonstrated 2012: heralded entanglement between two single Rb atoms 21 m apart. 2021 (Daiss et al., Science): photonic quantum gate between two distant atoms via a shared fiber cavity — the first remote two-qubit gate.
- Thompson group (Princeton/Caltech): Cs atoms in a 2D optical lattice coupled to a Fabry-Pérot cavity; cooperativity $C > 100$.
- Fiber Fabry-Pérot cavities (FFP): Micro-machined tips on optical fibers that form a cavity around a single atom. Miniature (100s μm long), compatible with tweezer arrays, scalable. Groups: Hunger (LMU), Vollmer, Schiller.
- Key challenge: Cavity alignment stability, acoustic/thermal vibration of cavity mirrors, and integration with the tweezer control optics all require careful engineering. The cavity finesse typically degrades near atomic density gradients.
Alkali atoms emit photons at 780–852 nm where standard SMF-28 fiber has 3–5 dB/km loss. At 1550 nm telecom C-band, loss is only 0.18–0.20 dB/km. Quantum frequency conversion (QFC) maps the atom's photon to telecom wavelengths while preserving all quantum information, including polarization and time-bin encoding.
Implemented via:
- Periodically poled lithium niobate (PPLN) waveguides: Difference-frequency generation (DFG) with a strong pump laser. Rb 780 nm + pump 1450 nm → 1550 nm (telecom C-band). Demonstrated with $>$70% internal efficiency and photon indistinguishability preserved.
- Nanophotonic AlGaAs and GaP waveguides: Higher nonlinearity, smaller footprint, CMOS compatible. Emerging platform.
- Key demonstration: van Leent et al. (2022, Nature): Rb atoms in separate labs entangled over 33 km of fiber, with QFC from 780 to 1517 nm. First demonstration of neutral-atom remote entanglement over deployed fiber distance.
- Memory qubit + communication qubit architecture: In $^{171}$Yb and $^{87}$Sr, the nuclear spin $I = 1/2$ serves as the memory qubit (coherence time seconds to minutes), while the electron spin handles photon emission. QFC from Yb 399/556 nm to telecom is an active area.
Alkaline-earth-like (AEL) atoms (¹⁷¹Yb, ⁸⁷Sr, ¹⁷³Yb) have a unique two-qubit architecture within a single atom. The nuclear spin ($I = 1/2$ for ¹⁷¹Yb) forms an ultra-long-lived memory qubit — largely decoupled from the environment since it has no magnetic moment to first order and no hyperfine structure in the $^1S_0$ ground state. The electronic state interacts with photons via the $^3P_1$ or $^3P_2$ transitions and serves as the communication qubit.
- Protocol: (1) Entangle nuclear spin and electron spin via a local quantum gate. (2) Use electron spin to emit a photon entangled with itself. (3) BSM creates remote atom-atom entanglement between electron spins. (4) Entanglement swapping transfers this to the nuclear spins. (5) Nuclear spins are used for computation while communication repeats.
- Key advantage: Memory qubits are coherent for seconds while communication is happening (100s of μs latency). No need to halt computation during entanglement generation.
- Groups: Atom Computing (¹⁷¹Yb), Ye group (JILA, ⁸⁷Sr), Kaufman group (JILA, ⁸⁷Sr and ⁸⁴Sr), Brown group (Duke).
The DLCZ protocol (Duan, Lukin, Cirac, Zoller, 2001) was the first practically feasible quantum repeater scheme and remains conceptually foundational. Instead of single atoms, it uses atomic ensembles where a write laser pulse creates a probabilistic spin wave:
|vacuum⟩ → |0⟩_ensemble|0⟩_photon + ε|1-excitation spin wave⟩|1-photon⟩ + O(ε²)
Detecting the photon heralds the spin wave. Two such nodes interfere their photons at a BSM, creating entanglement between the ensembles. A read pulse then converts the spin wave deterministically into a photon — enabling entanglement swapping across repeater nodes.
- Key advantage over single-atom schemes: Enhanced emission rate (collective enhancement by $\sqrt{N}$) — probabilistic but with much higher rate per attempt.
- Experiments: Kuzmich group (Georgia Tech, 2004) — first atom-atom entanglement via photons using ensembles. Vuletic group (MIT) — pushed rates significantly. Numerous demonstrations of quantum memory using DLCZ-type ensembles in cold atom clouds.
- Limitation for computing: Ensemble-based qubits are harder to use as computational qubits compared to single-atom registers. DLCZ is most relevant for quantum communication networks rather than modular quantum computing.
Historical Milestones
From the EPR paradox to metropolitan quantum networks — a 90-year arc toward practical quantum interconnects.
Entanglement Rate Calculator
Estimate Bell pair generation rates for different hardware configurations.
Photonic Link Parameters
Where Remote Entanglement Fits in the Computing Landscape
Not a replacement for the Rydberg gate — a complement that removes the scaling ceiling.
The two-layer architecture
The mature vision for neutral-atom quantum computing is a two-layer architecture that separates timescales:
- Layer 1 — Intra-module: Rydberg gates at 0.2–5 μs. High fidelity (99.5–99.9%). Deterministic. Operates within a single tweezer array. All the work in this site's other QC tools addresses this layer.
- Layer 2 — Inter-module: Photonic entanglement at ~10 μs–10 ms per Bell pair. Probabilistic but heralded. Operates between tweezer arrays in different vacuum chambers. Error-corrected logical operations are teleported across this layer.
The layers are complementary. Fault-tolerant quantum computing requires both: logical qubits encoded across many physical qubits (using Rydberg gates within modules) and logical operations on qubits in different modules (using entanglement-based gate teleportation).
Who's building this?
- QuEra Computing: Roadmap explicitly includes inter-module photonic links by 2027–2028. The 2023 Harvard/MIT/QuEra 48 logical qubit paper acknowledged remote entanglement as the critical missing piece for large-scale FTQC.
- Atom Computing (acquired by Google): ¹⁷¹Yb architecture's nuclear-spin memory is specifically suited for photonic links with long coherence during communication.
- Infleqtion (formerly ColdQuanta): Cs-based systems; working on photonic interface integration.
- Pasqal: Rb-based systems; photonic links are a stated roadmap item for their fault-tolerant tier.
- IonQ: Trapped ions demonstrated inter-module entanglement (2016-present); serving as benchmark for neutral-atom teams.
- Academic leaders: Lukin (Harvard), Ye (JILA), Kaufman (JILA), Covey (UIUC), Thompson (Princeton), Rempe (MPQ), Weinfurter (LMU).
07 References & Further Reading
Seminal theory papers, key experimental demonstrations, and review articles on remote entanglement and quantum networking.