Table of Contents
Fetching ...

Quantum Noise Suppression at Scale with Crosstalk-Robust Gate Sets

Andy J. Goldschmidt, Emilio Peláez Cisneros, Ryan Sitler, Kevin Olsson, Kaitlin N. Smith, Gregory Quiroz

Abstract

We introduce crosstalk-robust gate sets, which are obtained using a novel, scalable optimal control problem exploiting locality. Through the suppression of pairwise quantum crosstalk, the gate sets enable robustness that extends to multi-qubit circuits. The IBM Quantum Platform devices provide a testbed for our gate sets, where we study their efficacy via error suppression protocols and randomized parallel single-qubit circuits of up to eight qubits. Furthermore, we provide the first known assessment of the impact of complete optimal control gate sets on quantum algorithms. Using a Hamiltonian simulation of a four-qubit transverse field Ising model, we show that noise-informed gates enhance median algorithmic performance by a factor of four over baseline Gaussian gates using the same calibration procedures. Lastly, we provide numerical evidence that optimized gate sets enable larger qubit-qubit coupling strengths that can cut two-qubit gate times in half. This result confirms that hardware-software co-design using quantum optimal control can create new opportunities for quantum computing architectures.

Quantum Noise Suppression at Scale with Crosstalk-Robust Gate Sets

Abstract

We introduce crosstalk-robust gate sets, which are obtained using a novel, scalable optimal control problem exploiting locality. Through the suppression of pairwise quantum crosstalk, the gate sets enable robustness that extends to multi-qubit circuits. The IBM Quantum Platform devices provide a testbed for our gate sets, where we study their efficacy via error suppression protocols and randomized parallel single-qubit circuits of up to eight qubits. Furthermore, we provide the first known assessment of the impact of complete optimal control gate sets on quantum algorithms. Using a Hamiltonian simulation of a four-qubit transverse field Ising model, we show that noise-informed gates enhance median algorithmic performance by a factor of four over baseline Gaussian gates using the same calibration procedures. Lastly, we provide numerical evidence that optimized gate sets enable larger qubit-qubit coupling strengths that can cut two-qubit gate times in half. This result confirms that hardware-software co-design using quantum optimal control can create new opportunities for quantum computing architectures.
Paper Structure (30 sections, 30 equations, 12 figures, 3 tables)

This paper contains 30 sections, 30 equations, 12 figures, 3 tables.

Figures (12)

  • Figure 1: (a) Identical $X$ gates are harmed by $ZZ$ crosstalk when played in parallel, but distinct (optimized) $X$ gates can operate unaffected. (b) Optimal control finds crosstalk-robust gate sets by solving for gates (vertices) subject to $ZZ$ crosstalk errors (edges). Solid edges connect shaped gates, and dashed edges connect to the preserved idle identity. (c) Two-coloring a heavy-hex lattice of qubits enables device-wide crosstalk mitigation. (d) The pulse shapes of the gate set are shaped $X$ rotations equipped with standard calibration routines.
  • Figure 2: A crosstalk robust gateset (CRGS) is a single qubit gate set, $\mathcal{G}$, repeated in a set of colors, $\mathcal{C}$. (a) A two-colorable square lattice (top) and a three-colorable triangular lattice (bottom). (b) i. A two-element gateset and (b) ii. a three-element gateset are colored according to (a) and assembled into a graph $(V, E)$ for Eq. \ref{['eqn:robust-gate set-problem']}.
  • Figure 3: Crosstalk-robust gate sets optimized via Eq. \ref{['eqn:robust-gate set-problem']}. Each individual gate is constrained to have a minimum fidelity of $0.9999$. The log of crosstalk susceptibility is plotted over various amplitude and curvature bounds (note the suppressed zero on the $x$ and $y$ axis). Insets (b)-(d) show control amplitudes (GHz) and times (ns) for selected cases.
  • Figure 4: Gate sets generated and executed on IBM Brisbane. Pulse profiles shown for (a) Gaussian, (b) detuning-robust, and two-colorable, crosstalk-robust gate sets denoted as (c) blue and (d) red.
  • Figure 5: Illustration of a transpile pass to decompose the four-qubit circuit fragment into one-qubit, two-qubit, and virtual moments. The transpilation leverages the cancellation features afforded by the crosstalk-robust gate set to schedule overlapping single-qubit gates without accruing spatially-correlated errors.
  • ...and 7 more figures