W State Preparation Protocols

• W state preparation is a set of protocols that generate multipartite single-excitation entangled states crucial for resilient quantum information processing.
• Techniques such as gate synthesis, measurement-based schemes, and dissipative pumping are employed to balance fidelity, depth, and resource tradeoffs.
• Experimental implementations across platforms like trapped ions, circuit-QED, and Rydberg rings have achieved high fidelities, scalability, and deterministic state expansion.

W state preparation comprises the family of protocols that generate multipartite single-excitation entangled states of the form

$$|W_n\rangle=\frac{1}{\sqrt{n}\left(|100\cdots 0\rangle+|010\cdots 0\rangle+\cdots+|000\cdots 1\rangle\right)},$$

together with closely related $W$-class targets that differ by amplitudes, phases, chirality, or many-body encoding. Across the literature, these protocols appear in gate-based synthesis, optimal control, ancilla-mediated measurement schemes, remote preparation, dissipative pumping, and analog many-body dynamics. The class is technically important because $W$ states combine multipartite and bipartite entanglement and retain nontrivial entanglement after losing or tracing out one qubit, while exact $W$ states are not available as unique ground states of local gapped Hamiltonians, which motivates dynamical, dissipative, and frustration-based preparation routes rather than simple ground-state cooling (Catalano et al., 20 Oct 2025).

1. Target states and state-class variants

The standard target is the symmetric one-excitation Dicke state $|W_n\rangle$. Several papers, however, work with nonstandard but closely related targets. In deterministic entanglement swapping, the relevant three-qubit state is not the symmetric $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$, but the asymmetric $W$-class state

$$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$

chosen because its algebra closes exactly under a specific joint-measurement decomposition (Harraz et al., 2024). In dissipative Lindblad constructions, the target can be the generalized phased state

$|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$

which is related to the standard $W$ state by a local gauge transformation (Wampler et al., 2024). In three-ion trapped-ion protocols, the relevant basis contains six entangled $W$0 states $W$1, distinguished by excitation number and chirality (Cole et al., 2021). In frustrated Rydberg rings, the target is the equal superposition of single-kink antiferromagnetic configurations,

$W$2

which is explicitly not exactly the standard $W$3 state at finite $W$4, although the two converge in the thermodynamic limit and can be mapped by Clifford circuits or an ancilla construction (Catalano et al., 20 Oct 2025).

Target family	Representative definition	Representative use
Standard $W$5	$W$6	gate synthesis, remote preparation, dissipative pumping
Asymmetric three-qubit $W$7-class	$W$8	deterministic entanglement swapping
Generalized phased $W$9	$W$0 with site phases $W$1	gauge-twisted dissipative preparation
Chiral three-ion $W$2 states	$W$3 and achiral $W$4	trapped-ion sympathetic-cooling protocol
Frustration-induced kink state	$W$5	analog Rydberg-ring preparation

This variation is not a terminological accident. It reflects the fact that different preparation mechanisms privilege different conserved sectors, symmetry classes, or correction structures. A recurrent technical point is therefore that “W state preparation” may mean exact preparation of the standard symmetric state, preparation of a locally equivalent $W$6-basis element, or preparation of a finite-size many-body analogue whose entanglement resource is the equal coherent delocalization over a symmetry-related manifold (Moreno et al., 2015).

2. Circuit, gate, and optimal-control constructions

Coherent circuit-based preparation ranges from analytically structured expansions to optimal-control protocols. In tunable three-qubit gmon hardware, the preparation problem reduces to the one-excitation basis $W$7. A simple adiabatic interpolation fails completely because the initial Hamiltonian has an exact degeneracy; the final error remains $W$8 independent of protocol duration. The optimal solution is instead bang-bang: $W$9 and $W$0 throughout, while $W$1 follow an on-off-on pattern with $W$2, $W$3, $W$4 for $W$5, and critical preparation time $W$6. The protocol remains robust for timing inaccuracies of order $W$7 (Jones et al., 2019).

Approximate synthesis on NISQ devices emphasizes a different tradeoff. GASP uses the gate set $W$8, a genetic search over circuit structure, and SLSQP optimization of rotation angles. For $W$9 states from 2 to 10 qubits, it targets approximate preparation at specified fidelity rather than exact synthesis. The clearest reported comparison is the 6-qubit case: a GASP circuit of depth 22 and 59 total gates versus Qiskit initialize at depth 120 and 125 total gates, with the reported experiments using 99% target fidelity, the ibmq_guadalupe noise model, and 16384 shots. The paper further states that GASP consistently reduces both total gate count and CNOT count relative to Qiskit, including by more than two orders of magnitude in higher-qubit tests (Creevey et al., 2023).

A more structural construction is deterministic local expansion. The basic two-qubit gadget consists of a controlled-Hadamard followed by a CNOT. Applied to one accessed qubit of a $|W_n\rangle$0 state with one $|W_n\rangle$1 ancilla, it yields probabilistic growth $|W_n\rangle$2 with success probability

$|W_n\rangle$3

Applied in parallel to all $|W_n\rangle$4 qubits, each with its own ancilla, it deterministically doubles the state,

$|W_n\rangle$5

The paper also gives two routes to odd sizes: a final probabilistic one-qubit addition, or projection from the next larger even-size state (Yesilyurt et al., 2016).

Ancilla-mediated spin-photon constructions implement the same expansion logic without direct logical-qubit interaction. The fundamental three-qubit primitive $|W_n\rangle$6 uses 4 controlled-$|W_n\rangle$7 gates, 6 Hadamard gates, and 2 $|W_n\rangle$8 gates. It creates a $|W_n\rangle$9-type EPR pair from two separable logical qubits and one ancilla, expands $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$0 by one qubit, and supports deterministic size doubling $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$1. The ancilla is restored at the end of the operation, and neither post-measurement nor post-processing is required (Ozaydin et al., 2021).

3. Measurement-mediated and remote protocols

Measurement-based W-state preparation appears in both remote-state and swapping settings. A fully remote protocol uses three bipartite pairs, regrouped as $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$2 and $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$3, and an orthonormal three-qubit $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$4-basis $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$5. In the ideal case,

$$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$6

so measuring subsystem $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$7 in the $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$8-basis deterministically prepares the corresponding $$\frac{1}{\sqrt 3}(|100\rangle+|010\rangle+|001\rangle)$$9 on subsystem $W$0, after which local corrections map every outcome to the standard target $W$1. For the imperfect source $W$2, outcomes $W$3 and $W$4 still yield exact $W$5 states after correction with total postselected probability

$W$6

If no postselection is used, the output is mixed but its fidelity

$W$7

satisfies $W$8, where $W$9 is the imperfect-source fidelity relative to the ideal triple-EPR resource (Moreno et al., 2015).

Deterministic entanglement swapping uses a different measurement structure. Two copies of

$$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$0

are prepared at distant nodes, and Charlie performs a projective measurement on three transmitted qubits in the orthogonal basis

$$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$1

Each outcome leaves the remote subsystem in the target $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$2 state up to a known Pauli correction $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$3, so all four outcomes are accepted and the swapping is deterministic. Under amplitude damping, the reported fidelity is

$$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$4

and a weak-measurement filter raises this to

$$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$5

but only for the $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$6 branches; the purification-enhanced version therefore becomes probabilistic with maximal success probability $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$7. The paper is explicit that the elementary gate decomposition of the initial W-state preparation block is not given in the manuscript text (Harraz et al., 2024).

The measurement-based literature also includes broader state-preparation frameworks with partial relevance to $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$8 states. The abstract of a low-depth amplitude-encoding paper states that it gives a deterministic measurement-based method with circuit depth $$|W\rangle=\frac{1}{2}\left(|100\rangle+|010\rangle+\sqrt{2}|001\rangle\right),$$9, $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$0 ancilla qubits, and worked examples of both a dense state and a W-state, but the supplied extract contains no body-text details for the W-state construction (Nmaju et al., 9 Oct 2025). Measurement-induced steering on digital hardware similarly demonstrates ancilla-measurement-reset state preparation for arbitrary qubits and a qutrit, yet does not provide an explicit $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$1 protocol; its significance for W-state preparation is therefore conceptual rather than constructive (Volya et al., 2023).

4. Dissipative preparation and stabilization

A major line of work treats dissipation not as an error source but as the preparation mechanism itself. In cavity QED with identical $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$2 atoms, collective $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$3 symmetry permits restriction to the fully symmetric irreducible subspace, so spontaneous emission and cavity loss induce effective irreversible transport among symmetric ground states. In the single-mode scheme, the effective dynamics are one-way,

$|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$4

and the target $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$5 is the symmetric one-excitation state $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$6. In the bimodal scheme, one mode yields left-to-right transport and the other right-to-left transport, enabling a two-stage protocol: first pump an arbitrary thermal ground-manifold state to $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$7, then pump $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$8 to $|W_{\text{Gen}}\rangle=\frac{1}{\sqrt{N} \sum_{a=0}^{N-1} e^{i\phi_a} |\downarrow_0 \cdots \downarrow_{a-1}\uparrow_a\downarrow_{a+1}\cdots \downarrow_{N-1}\rangle ,$9. The paper frames this as dissipative preparation of large $W$0 states, of order ten qubits, using only global addressing (Sweke et al., 2014).

In trapped-ion systems, dissipative preparation is realized by sympathetic cooling plus chirality-selective collective couplings. The three-qubit Hilbert space is organized into states $W$1, and the protocol can generate any one of the six entangled $W$2 states $W$3. Global Mølmer-Sørensen and sideband interactions, together with tailored destructive interference, pump population into the selected target while cooling removes phonons irreversibly. The target state is trapped in the motional ground state because the relevant red-sideband loss channel is blocked by $W$4. For realistic $W$5 parameters, the reported estimate is about $W$6 fidelity in about $W$7, without individual addressing (Cole et al., 2021).

A more abstract dissipative construction uses purely Lindbladian dynamics with local two-site jumps. The preparation jumps $W$8 and $W$9 prepare an exact $W$00 state: $W$01 creates local symmetric coherence within fixed particle-number sectors, and $W$02 pumps higher-excitation sectors down to the one-particle sector. The result is that every initial state except the all-down vacuum converges to the W-state. The same formalism prepares the generalized phased state $W$03 by gauge-transforming the jump operators. The error jumps $W$04 and $W$05 preserve the $W$06 state as a dark state but can destroy its attraction basin, leading to an absorbing-state/decohering transition. The all-to-all version is gapped and fast in the ideal limit but fragile, with critical error scale $W$07; the nearest-neighbor version is slower but robust, with threshold scale $W$08, and coherence remnants persist even after the transition (Wampler et al., 2024).

5. Fast coherent implementations and experimental platforms

Shortcut-to-adiabaticity methods supply hardware-specific high-fidelity coherent routes. In a circuit-QED architecture with four rf SQUIDs coupled to a cavity, the dynamics reduce to an effective three-level system $W$09. The dressed-state method cancels nonadiabatic couplings without introducing extra physical interactions, and the corrected pulses can be fitted by linear combinations of Gaussian functions. With $W$10 and $W$11, the dressed-state protocol reaches essentially unit fidelity at $W$12, whereas STIRAP with the same amplitudes gives only $W$13; with experimentally realistic values $W$14, $W$15, $W$16, and $W$17, the predicted fidelity is $W$18 (Kang et al., 2016).

Rydberg-blockade super-atom schemes offer both heralded and coherent-superadiabatic variants. In the heralded version, strong $W$19 dressing isolates the collective transition $W$20, with

$W$21

at $W$22 and $W$23. Field ionization of the single Rydberg excitation then projects the remaining $W$24 atoms into the symmetric Dicke state $W$25. At the optimal high-fidelity operating point, the success probability is approximately $W$26, and for a Poissonian atom-number distribution with mean $W$27 the weighted average infidelity $W$28 is essentially unchanged from the $W$29 value for exactly $W$30 (Gärttner, 2015). In the cavity-fiber superatom setting, a superadiabatic iterative technique reduces the dynamics of four Rydberg superatoms to an effective three-level system, with the second counterdiabatic Hamiltonian having the same structure as the effective Hamiltonian. The reported ideal fidelity is $W$31, it remains above $W$32 when spontaneous emission and photon leakage are included, and the construction extends formally to $W$33 superatoms (Yang et al., 21 Sep 2025).

Many-body analog preparation has now reached experimental scale. On QuEra’s Aquila device, an odd antiferromagnetic Rydberg ring realizes topological ring frustration, and the symmetric single-kink state

$W$34

emerges as the target. The protocol has been implemented for $W$35 Rubidium atoms. The paper emphasizes that $W$36 is not exactly the standard $W$37 state at finite $W$38, but serves as its frustrated-ring analogue; a lower-bound fidelity of around $W$39 is certified for the experimentally prepared 11-qubit state, and ideal scaling studies report optimized preparation time

$W$40

The associated minimum gap scales as $W$41, which explains both the possibility of the construction and its size-dependent adiabatic cost (Catalano et al., 20 Oct 2025).

6. Determinism, resource tradeoffs, and recurrent limitations

The literature uses “deterministic” in several distinct senses. Remote $W$42-basis measurement from ideal EPR sources is deterministic because every measurement outcome is correctable (Moreno et al., 2015). Entanglement swapping with the asymmetric three-qubit state is deterministic because every Charlie outcome is accepted and corrected by a single-qubit Pauli (Harraz et al., 2024). Local optical expansion becomes deterministic only when all input qubits are processed in parallel (Yesilyurt et al., 2016), and ancilla-mediated spin-photon doubling is deterministic in principle because it avoids probabilistic fusion, post-selection, and post-processing (Ozaydin et al., 2021). By contrast, heralded Rydberg protocols are intrinsically conditional on ion detection (Gärttner, 2015), and purification layers can turn a deterministic underlying protocol into a postselected one by discarding measurement branches (Harraz et al., 2024).

A second recurrent distinction concerns what is being prepared. Some protocols target the standard symmetric $W$43, others an asymmetric $W$44-class state tailored to a measurement algebra, others a generalized phased $W$45, others one of several chiral three-qubit $W$46 states, and others a finite-size many-body analogue such as the single-kink frustrated-ring state (Harraz et al., 2024, Cole et al., 2021, Wampler et al., 2024, Catalano et al., 20 Oct 2025). This distinction matters operationally: correction tables, dark-state conditions, and validation procedures depend on the exact target manifold, not merely on the label “W state.”

Resource tradeoffs are equally pronounced. Approximate NISQ synthesis can outperform exact arbitrary-state initialization because it lowers depth and CNOT count (Creevey et al., 2023). Dissipative all-to-all preparation can be constant-time in the ideal limit but becomes thermodynamically fragile under weak dark-state-preserving errors (Wampler et al., 2024). Frustration-based analog preparation avoids the local-gapped-ground-state obstruction but pays a polynomial adiabatic-time cost set by a gap that closes as $W$47 (Catalano et al., 20 Oct 2025). The abstract of the low-depth measurement-based amplitude-encoding scheme reports $W$48 depth, $W$49 ancilla qubits, and a worked W-state example, indicating yet another point in the depth-space tradeoff landscape (Nmaju et al., 9 Oct 2025).

Verification is often as nontrivial as preparation. One remote protocol requires a full three-qubit $W$50-basis analyzer in its ideal deterministic form, though imperfect-source postselection only needs discrimination among $W$51, $W$52, and the rest (Moreno et al., 2015). The deterministic entanglement-swapping manuscript specifies the target state exactly but omits the elementary decomposition of its initial W-state preparation block (Harraz et al., 2024). In the Aquila experiment, always-on interactions during global rotations obstruct direct $W$53-basis measurements, forcing the use of a two-setting fidelity estimator together with simulation-informed Bayesian tomography (Catalano et al., 20 Oct 2025). This suggests that, for W-state preparation, state certification is frequently limited by the same architectural constraints that shape the preparation protocol itself.