A diagrammatic method to compute the effective Hamiltonian of driven nonlinear oscillators

Xu Xiao, Jayameenakshi Venkatraman, Rodrigo G. Cortiñas, Shoumik Chowdhury, Michel DevoretarXiv 2023 · DOI · arXiv

In the world of quantum computing, specifically circuit quantum electrodynamics (cQED), the nonlinear oscillator is king. Whether it's a transmon qubit or a SNAIL-based parametric amplifier, these systems rely on the delicate interplay between nonlinearity and microwave drives. However, calculating exactly how these systems behave when driven hard has historically been a mathematical nightmare.

A new paper from researchers at Yale University introduces a diagrammatic method that brings "Feynman-like" simplicity to these complex calculations, offering both a powerful computational tool and a new intuitive language for quantum engineers.

TL;DR

Researchers have developed a diagrammatic perturbation method to calculate the effective behavior of driven nonlinear oscillators. By translating complex algebra into intuitive "mixer" and "propagator" diagrams, they can predict high-order quantum effects—like multiphoton resonances—with high precision, matching experimental data where traditional methods fail.

What Problem Does This Paper Address?

When we drive a nonlinear oscillator (like a superconducting qubit) with microwaves, the system’s Hamiltonian becomes time-dependent. To understand the "slow" physics we actually care about—like how a qubit's frequency shifts or how it transitions between states—we usually look for a "static effective Hamiltonian."

The standard tool for this is the Rotating Wave Approximation (RWA), but it’s only a first-order guess. As drives get stronger or we look for more complex behaviors (like "three-legged" Schrödinger cat states), we need higher-order corrections. Existing methods like Schrieffer-Wolff perturbation theory become exponentially difficult to calculate by hand and offer little physical insight into why certain terms appear. There was a clear need for a systematic, intuitive, and computationally efficient way to handle these "beyond-RWA" effects.

Methods

The authors introduce Quantum Harmonic Balance (QHB), a framework developed in phase space (using the Husimi Q representation). This approach allows them to treat classical and quantum dynamics on equal footing.

The core innovation is the Diagrammatic Representation:

  • Vertices (Mixers): Represent nonlinear interactions (e.g., a 3-wave or 4-wave mixing process).
  • External Edges (Excitations): Represent either the drive photons or the resonant oscillations of the system.
  • Internal Edges (Propagators): Represent "off-resonant" intermediate states that the system briefly visits.

Unlike iterative algebraic methods, this diagrammatic approach is non-iterative. If you want to know a specific 10th-order term, you don't have to calculate orders 1 through 9 first; you simply draw all the topologically unique diagrams that fit the 10th-order criteria and sum them up. The authors even provided an open-source computer program to automate this process.

Datasets

As a theoretical and methodological paper, it does not use "training" datasets in the machine-learning sense. Instead, it validates the method using:

  1. Experimental Data: They compared their model against previously published ac-Stark shift measurements from a transmon-cavity system, showing that their expansion converges perfectly with real-world results at high orders.
  2. Numerical Benchmarks: They used exact Floquet numerical diagonalization—a computationally expensive but "perfect" simulation—to verify their analytical predictions for multiphoton resonances.

Key Insights and Ideas

  • Quantum Bonds: In their "expanded" diagrams, quantum corrections to classical physics appear as "quantum bonds" between resonant excitations. This visually separates the classical "harmonic balance" from purely quantum effects like the Lamb shift.
  • The Metapotential: The paper identifies a shared semiclassical object called a "metapotential." This energy surface governs both classical bifurcations (like period-doubling) and quantum phenomena (like the formation of Schrödinger cat states), unifying two fields that often speak different languages.
  • Bookkeeping Advantage: Because the method "book-keeps" different types of terms, it can handle extremely high-order processes (up to the 14th order and beyond) that would crash standard symbolic algebra software due to memory overflow.
  • Interference and Suppression: They discovered that at certain "sweet spots" in circuit design, different off-resonant paths can interfere destructively, completely suppressing unwanted parasitic resonances.

Key Results

  • High-Order Precision: The method successfully modeled the ac-Stark shift in a driven transmon up to the 10th order, providing a parameter-free fit to experimental data.
  • Multiphoton Characterization: They provided the first comprehensive analytical map of multiphoton resonances (e.g., a 5:3 resonance where 5 oscillator quanta couple to 3 drive photons). These processes are critical for understanding qubit "ionization" and readout errors.
  • Circuit Design (IST): They proposed the Inductively Shunted Transmon (IST). By adding a linear inductor, they showed they could "dilute" the nonlinearity, effectively mitigating dangerous multiphoton resonances and preventing the system from entering a chaotic "quantum diffusion" regime.

Limitations and Future Work

The authors acknowledge that their method is a perturbative expansion, meaning it works best in the "weakly excited" regime. If the drive becomes so violent that the system enters a fully chaotic state (quantum diffusion), the expansion may fail to converge.

Future work includes extending the diagrammatic language to open quantum systems (incorporating decay and decoherence directly into the diagrams) and applying the method to more complex multi-mode systems, such as large-scale quantum processors where "frequency crowding" makes parasitic resonances a major hurdle.

Summary updated

Abstract

In this work, we present a new diagrammatic method for computing the effective Hamiltonian of driven nonlinear oscillators. At the heart of our method is a self-consistent perturbation expansion developed in phase space, which establishes a direct correspondence between the diagram and algebra. Each diagram corresponds to a Hamiltonian term, the prefactor of which, like those in Feynman diagrams, involves a simple counting of topologically equivalent diagrams. Leveraging the algorithmic simplicity of our diagrammatic method, we provide a readily available computer program that generates the effective Hamiltonian to arbitrary order. We show the consistency of our schemes with existing perturbation methods such as the Schrieffer-Wolff method. Furthermore, we recover the classical harmonic balance scheme from our result in the limit of $\hbar\rightarrow0$. Our method contributes to the understanding of dynamic control within quantum systems and achieves precision essential for advancing future quantum information processors. To demonstrate its value and versatility, we analyze five examples from the field of superconducting circuits. These include an experimental proposal for the Hamiltonian stabilization of a three-legged Schrödinger cat, modeling of energy renormalization phenomena in superconducting circuits experiments, a comprehensive characterization of multiphoton resonances in a driven transmon, a proposal for an inductively shunted transmon circuit, and a characterization of classical ultra-subharmonic bifurcation in driven oscillators. Lastly, we benchmark the performance of our method by comparing it with experimental data and exact Floquet numerical diagonalization.
BibTeX
@misc{xiao2023diagrammatic,
  author = {Xu Xiao and Jayameenakshi Venkatraman and Rodrigo G. Cortiñas and Shoumik Chowdhury and Michel Devoret},
  title = {{A diagrammatic method to compute the effective Hamiltonian of driven nonlinear oscillators}},
  year = {2023},
  eprint = {2304.13656},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2304.13656},
}

Loading related content...

Figure