Cubic Phase Gate with a Trapped Ion Oscillator

AI summary

Overview

Research area: Experimental continuous-variable (CV) quantum information processing with trapped ions; specifically, the realization of nonlinear bosonic unitaries using hybrid continuous-variable/discrete-variable (CV-DV) interactions.

Technical level: Intermediate. The paper assumes familiarity with phase-space formulation (Wigner functions, characteristic functions), spin-dependent forces, and the notion of a universal CV gate set, but the experimental logic is described clearly enough for a reader with graduate-level quantum optics background.

Scope (one sentence): The paper reports an experimental demonstration, on a single trapped ¹⁷¹Yb⁺ ion, of a numerically optimized sequence of spin-dependent-force pulses that approximates the cubic phase gate on a motional mode, benchmarked by Wigner-function overlaps and nonlinear squeezing.

What This Paper Is About

Universal CV quantum computation requires an anharmonic, nonlinear unitary on top of the linear Gaussian operations (displacement, phase rotation, squeezing, beamsplitter), and the cubic phase gate U_CPG = exp(iκX̂³) is the canonical example. Such nonlinearities are hard to realize experimentally because the underlying oscillators are harmonic, so the nonlinearity must be engineered rather than found in the hardware. Following a proposal by K. Park and R. Filip, the authors approximate the cubic phase gate by a finite sequence of spin-dependent-force (SDF) pulses on a trapped-ion motional mode, with pulse phases and amplitudes found by numerical optimization, and then verify the resulting state against the ideal cubic phase state.

Key Contributions

1. Experimental approximation of the cubic phase gate on a trapped-ion oscillator. The authors implement the finite-Fourier-series approach of Park and Filip on the R1 radial motional mode of a ¹⁷¹Yb⁺ ion, using only carrier rotations and bichromatic SDF pulses.

2. A parameter-compressed numerical optimization. Symmetry constraints on the spin phases φ_i and displacement amplitudes ξ_i reduce the optimization from 2N parameters to N, yielding working pulse sequences for N = 6, 10 and 14.

3. Full state characterization of the output. The characteristic function is measured following an existing trapped-ion protocol and converted into Wigner functions, then compared against both the ideal cubic-unitary evolution and the numerically simulated protocol.

4. A second, independent benchmark. Nonlinear squeezing of the nonlinear quadrature p̂ − 3λx̂² is computed from two-dimensional fits to the characteristic function, with uncertainties from Monte Carlo re-fitting, and compared against the vacuum-state value.

Main Findings

• Simulated overlaps improve monotonically with pulse number. For the vacuum input, the simulated protocol versus the cubic unitary (κ = −1/2) gives overlaps of 0.983 (N = 6), 0.995 (N = 10) and 0.998 (N = 14).

• Best experimental overlap at N = 14. Measured cubic-unitary-versus-experiment overlaps (F_UE) for vacuum input are 0.882(2) for N = 6, 0.864(2) for N = 10, and 0.955(2) for N = 14; the corresponding simulated-versus-experiment values (F_SE) are 0.899(2), 0.868(2) and 0.958(2).

• The 10-pulse sequence underperforms the 6-pulse sequence experimentally. The authors attribute this to the 10-pulse protocol reaching the highest maximum n̄ of the three, which makes higher-order Lamb-Dicke terms and motional dephasing more significant.

• Coherent-state inputs degrade the approximation away from the origin. Simulated overlaps for the N = 14 protocol stay above 0.95 for |α_initial| ≲ 0.5 but decay rapidly for |α_initial| ≳ 0.5 along the P̂ axis; displacements along X̂ leave the overlap unchanged because they commute with the SDFs in the protocol.

• Measured coherent-state overlaps follow the same pattern. F_UE values at N = 14 are 0.908(2) for α = 0.5, 0.844(2) for α = −0.5, 0.930(2) for α = 0.5i and 0.895(2) for α = −0.5i, all below the vacuum-input value.

• The qubit is left essentially unentangled from the oscillator. For N = 14 the qubit purity is 0.9992 with Bloch components {σ_x, σ_y, σ_z} = {−0.9992, −0.0014, 0}, consistent with the target of an odd function Φ(X̂) ≈ κX̂³.

• Wigner-function slices match the expected Airy-function shape. A single slice at Im(γ) = 0 agrees with the ideal cubic phase state, with error bars from projection noise over 200 experimental repetitions.

• Nonlinear squeezing is observed below the vacuum level. Using the N = 14 state and accounting for one standard deviation of Monte Carlo uncertainty, the cubic phase state's nonlinear squeezing is suppressed below the vacuum for λ ≲ −0.23.

• Protocol duration is short relative to motional coherence. Total durations including two 25 µs carrier pulses are typically 250 µs (N = 6), 400 µs (N = 10) and 400 µs (N = 14), against estimated 1/e motional coherence times of 49(7), 47(7) and 60(20) ms for the Ax, R1 and R2 modes.

Methodology in Plain English

The gate is built out of a repeated elementary operation: a pair of laser tones (red and blue sidebands applied together) that push the ion's motion in a direction that depends on the state of its internal qubit. Each such pulse is described by two knobs, a spin phase φ and a displacement amplitude ξ. A single pulse is just a displacement; a cleverly chosen train of them, interleaved with the qubit's state, can synthesize a nonlinear phase that depends on position as X̂³.

To find the knob settings, the authors simulate the pulse train on a vacuum state with the qubit prepared in the |−⟩ state and numerically minimize the infidelity 1 − |⟨ψ_tar|ψ_sim⟩| against the state produced by the true cubic unitary. Initial guesses were random (φ_i between 0 and 2π, ξ_i between 0 and 1), and symmetry constraints halve the number of free parameters. A final carrier rotation returns the qubit to the computational basis.

Experimentally, the ion is laser-cooled to n̄ < 0.05, optically pumped to |g⟩, rotated to |−⟩, subjected to the optimized pulse train, and rotated back. Because the whole sequence is stitched into one uninterrupted arbitrary-waveform-generator pulse, the experiment avoids unnecessary latency. The output is characterized not by full tomography but by measuring the characteristic function χ(β) = ⟨D̂(β)⟩: a carrier pulse selects the real or imaginary component, and an SDF pulse selects the phase-space coordinate β. The characteristic function is then Fourier-transformed into a Wigner function on a discrete grid with spacing Δa = Δb = 0.18, and overlaps are computed from the discrete Wigner functions with Δx = Δp = 0.18 × √2. For the nonlinear-squeezing benchmark, the characteristic function is fitted with a 12-parameter two-dimensional model, and higher derivatives at the origin give the required moments.

The hardware is a single ¹⁷¹Yb⁺ ion in a four-rod, two-needle linear Paul trap driven at 17.7 MHz, with qubit states encoded in the ²S₁/₂ hyperfine clock states separated by 12.6 GHz, Raman transitions driven by a 355 nm mode-locked pulsed laser with beams crossing at 135°, and detection via a 369.5 nm beam and photomultiplier tube. Simulations and optimizations used QuTiP with a Hilbert-space dimension of 100.

Why This Matters

The result shows that a hard nonlinear CV gate can be assembled from operations that already exist on essentially every trapped-ion quantum processor: carrier rotations and bichromatic sideband pulses. That lowers the barrier to entry for CV quantum computing on ion traps and provides an experimental data point for the bosonic quantum signal processing family of protocols, which the paper links to applications in quantum simulation, CV computing, metrology, and device control/characterization.

Real-world applications (potential, as framed by the paper's cited literature):

• Continuous-variable quantum computing. Supplying the non-Gaussian element needed for universality alongside displacement, rotation, squeezing and beamsplitter gates.
• Quantum simulation. Emulating anharmonic potentials, including the double-well potential, in a controllable oscillator.
• Quantum metrology. Generating non-Gaussian probe states whose nonlinear squeezing can outperform vacuum as a resource.
• Device control and characterization. Using programmable bosonic QSP sequences to calibrate and control oscillator hardware.

Industry relevance: The protocol requires no specialized hardware beyond standard Raman laser systems and an arbitrary waveform generator, so it is directly transferable to existing trapped-ion platforms being developed for quantum computing; the characteristic-function measurement and nonlinear-squeezing benchmark also offer practical verification tools for CV hardware.

Future Directions

• Better optimization. The authors state that the numerical optimization could be improved with more sophisticated minimization protocols and by adding weights to each non-unique solution set.
• Extension to other nonlinear gates. The same workflow is proposed for the quartic phase gate and the double-well potential.
• Pushing to larger coherent-state inputs. The rapid decay of overlap for |α_initial| ≳ 0.5 along the P̂ axis is an open limitation that would need addressing for the gate to act on encoded, displaced states.
• Suppressing the n̄-dependent error. The anomalous N = 10 result points to higher-order Lamb-Dicke terms and motional dephasing at large occupation, which would need mitigation for longer or more demanding sequences.

Target Audience

This paper is most useful to experimentalists working on trapped-ion quantum computing and to researchers in continuous-variable quantum information who want a concrete, reproducible recipe for engineering non-Gaussian operations. It is also relevant to theorists developing bosonic quantum signal processing and other hybrid CV-DV protocols, and to readers interested in benchmarks such as Wigner-function overlap and nonlinear squeezing for non-Gaussian state verification.

Authors’ abstract

The motional modes of trapped ions are promising candidates for continuous-variable (CV) quantum computation and simulation. However, to achieve a universal gate set for CV quantum computation, anharmonic nonlinear unitaries are required. Here, we experimentally demonstrate a protocol that allows us to approximate the cubic phase gate. By numerically optimizing the phase angles and amplitudes of a set of state-dependent force pulses, we demonstrate the cubic phase gate for both vacuum and low-$α$ coherent state inputs and verify its performance. Our protocol can also be extended to approximate other nonlinear unitaries.