Demonstration of tripartite cat states in two distinct classes of entanglement

Technical Summary: Demonstration of Tripartite Cat States in Two Distinct Classes of Entanglement

Problem Statement
While bipartite entanglement is well-established, genuine multipartite entangled states in the macroscopic continuous-variable (CV) domain remain largely inaccessible. Existing experimental demonstrations of multipartite entanglement, such as GHZ and W states, have been confined primarily to discrete-variable (DV) systems (e.g., trapped ions, superconducting qubits). Extending these states to the CV regime, particularly at macroscopic scales, is challenging due to the need for long coherence times and precise nonlocal control over many-body interactions in a large Hilbert space. Furthermore, the landscape of tripartite entanglement comprises two inequivalent classes—GHZ-type (lacking pairwise entanglement) and W-type (retaining residual bipartite entanglement)—which exhibit distinct physical properties and applications. Generating and characterizing both classes in a single macroscopic CV platform has not been previously achieved.

Methodology
The authors experimentally realized macroscopic tripartite entangled "cat" states encoded in three quantum harmonic oscillators (microwave resonators) coupled to a single ancillary superconducting transmon. The hardware consists of three tantalum hairpin resonators (Alice, Bob, and Charlie) arranged above and below a central chip hosting the transmon, readout resonator, and Purcell filter.

To generate the states, the team developed a "one-to-all" conditional control strategy utilizing the transmon's energy levels:

1. GHZ-cat Generation: The GHZ-cat state, defined as $|GHZ_{cat}\rangle \propto |\alpha, \alpha, \alpha\rangle + |-\alpha, -\alpha, -\alpha\rangle$, was generated using an Echoed Conditional Displacement (ECD) gate. This gate entangles the three oscillators with the transmon's ground ($|g\rangle$) and excited ($|e\rangle$) states. By preparing the transmon in a superposition, applying a three-mode ECD, and performing post-selection on the transmon state $|g\rangle$, the target GHZ-cat state is obtained.
2. W-cat Generation: The W-cat state, defined as $|W_{cat}\rangle \propto |\alpha, -\alpha, -\alpha\rangle + |-\alpha, \alpha, -\alpha\rangle + |-\alpha, -\alpha, \alpha\rangle$, requires mapping three superposition components to three distinct transmon levels ($|g\rangle, |e\rangle, |f\rangle$). Since standard ECD gates operate only in the $g-e$ subspace, the authors engineered an "Uneven Echoed Conditional Displacement" (UECD) gate. This gate operates in the $g-f$ subspace while leaving the $|e\rangle$ component undisturbed, compensating for asymmetric dispersive shifts through optimized pulse parameters.
3. State Characterization: Instead of full-state tomography, which is resource-intensive for multimode CV systems, the authors employed an efficient subspace tomography protocol. They mapped the expectation values of 64 Pauli operators (treating $|\alpha\rangle$ and $|-\alpha\rangle$ as logical qubit states) to specific points in the joint characteristic function (CF). This reduced the measurement requirement from $\sim 10^5$ points to 172 unique CF points.

Key Results

• State Fidelity: The experiment achieved fidelities of $0.83 \pm 0.02$ for the GHZ-cat state and $0.70 \pm 0.02$ for the W-cat state (after tomography-error correction). Both values exceed the theoretical thresholds required to certify genuine multipartite entanglement (GME) for their respective classes ($0.75$ for GHZ and $2/3$ for W).
• Entanglement Verification:
  • GHZ-cat: The state violated Mermin's inequality ($M_s = 2.7 \pm 0.1$), confirming nonlocality and GME. Crucially, the reduced density matrices of any two modes showed zero entanglement, consistent with the GHZ class.
  • W-cat: The state retained bipartite entanglement. After tracing out one mode, the remaining two-mode state exhibited a negativity of $0.13 \pm 0.01$, corresponding to 64% of the ideal value, confirming the W-class structure.
• Error Analysis: The primary limitation on fidelity was identified as photon loss in the oscillators, which causes the decay of coherence in the superposition of coherent states. The authors developed a control experiment (three-mode "cat-and-back") to independently calibrate these decoherence factors, allowing for accurate error correction of the measured Pauli operators.

Significance and Claims
The paper claims to provide the first experimental synthesis and characterization of two distinct classes of macroscopic tripartite entangled states (GHZ-cat and W-cat) within a single continuous-variable hardware platform. The work demonstrates a scalable method for generating inequivalent classes of multipartite CV entanglement using one-to-all conditional controls.

The authors position this achievement as a significant step toward:

1. Fault-Tolerant Quantum Processing: Providing a testbed for multipartite bosonic codes used in quantum error correction.
2. Fundamental Physics: Enabling studies on the structure of entanglement, its robustness against particle loss, and the quantum-to-classical transition in macroscopic systems.
3. Hardware Advancement: Showcasing the capability of bosonic cQED architectures to perform complex, nonlocal control over multiple modes without requiring stringent hardware matching of coupling strengths.

The authors remain modest regarding the current fidelity limits, attributing them to the specific coherence times of their tantalum resonators ($50-100 \mu s$). They note that theoretical analysis suggests that with state-of-the-art resonator lifetimes ($1 ms$) and improved transmon coherence, fidelities could reach $0.93$ and $0.88$ for GHZ and W states, respectively. The work is presented as a foundational demonstration rather than a fully fault-tolerant system, serving as a valuable testbed for future investigations into macroscopic many-body entanglement.