⚛️ quantum physics

Filon Methods for Highly Oscillatory Controlled Quantum Systems

This paper introduces Filon and Controlled Filon numerical methods that leverage Filon quadrature to efficiently simulate highly oscillatory controlled quantum systems, demonstrating significant computational speedups—up to 6x over the best Hermite methods—by drastically reducing the number of timesteps required for accurate simulations of superconducting transmon qubits.

Original authors: Spencer Lee, Daniel Appelö

Published 2026-07-16

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Spencer Lee, Daniel Appelö

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to listen to a song played on a violin, but the violin is being played at a speed so fast that the sound waves blur into a single, high-pitched scream. Now, imagine you need to record every single vibration of that string to understand the music perfectly. If you try to take a photo of the string with a camera that snaps one picture every second, you will miss everything; the string will have moved thousands of times between your shots. This is the daily struggle of scientists trying to build quantum computers. These machines use tiny particles called qubits that vibrate incredibly fast—billions of times per second. To control them, scientists must write "control pulses," which are like musical scores telling the qubits when to dance. But because the qubits vibrate so wildly, simulating their behavior on a computer is like trying to count every grain of sand on a beach while running a marathon. Standard computer methods are forced to take tiny, microscopic steps to keep up with the vibration, which makes the calculations take forever and use up massive amounts of energy.

The paper you are about to read tackles this exact headache. The authors, Spencer Lee and Daniel Appelö, are mathematicians who specialize in quantum computing. They are trying to solve a problem called "quantum optimal control," which is essentially the art of designing the perfect control pulses to make quantum computers work without errors. The core challenge is that the math describing these particles is "highly oscillatory," meaning it wiggles back and forth with extreme speed. The authors introduce a new set of tools, inspired by an old mathematical trick called "Filon quadrature," to simulate these wiggly systems much faster. Instead of taking tiny, painful steps like a snail, their new methods take giant, confident strides, knowing exactly where the wiggles are going. They tested these methods on a realistic simulation of a quantum logic gate (a basic building block of a quantum computer) and found that their approach could be hundreds of times faster than the current best methods, all while keeping the same level of accuracy.

The Problem: Counting the Unseeable

To understand why this is such a big deal, let's look at how quantum computers work. They rely on qubits, which can be in a state of "0," "1," or a spooky mix of both. To make a qubit do something useful, like flipping from 0 to 1, scientists hit it with microwave pulses. These pulses are like a conductor waving a baton, telling the qubit how to move. The problem is that the qubit is already vibrating naturally at a frequency in the Gigahertz range (billions of cycles per second). When you add the control pulse on top of that, the math describing the system becomes a chaotic, super-fast dance.

To simulate this dance on a computer, you have to break time into tiny slices, called "timesteps." If the dance moves too fast, your slices have to be incredibly small to catch every move. If you make a slice too big, you miss the step, and your simulation becomes garbage. For a typical quantum gate that lasts only a few hundred nanoseconds (billionths of a second), standard methods might need to take thousands of these tiny slices. It's like trying to watch a hummingbird's wings by taking a photo every second; you'd just see a blur.

For years, scientists have used a shortcut called the "Rotating Wave Approximation" (RWA). This is like putting on sunglasses that filter out the fastest, most annoying vibrations so you can see the slower, more important movements. It makes the math easier, but it's an approximation. It throws away some of the real physics, which can lead to errors in the final design of the quantum computer. The authors of this paper wanted to simulate the real thing—the full, unfiltered, high-speed vibration—without waiting for the computer to age a thousand years.

The Solution: The "Filon" Magic Trick

The authors decided to stop fighting the vibrations and start riding them. They adapted a mathematical technique called Filon quadrature. Imagine you are trying to calculate the area under a curve that is wiggling like a snake. A standard method would measure the height of the snake at many, many points to get a good guess. Filon's method is smarter: it asks, "What is the frequency of the snake's wiggle?" Once you know the snake wiggles at a specific rhythm, you don't need to measure every single point. You can measure just a few key points and use the known rhythm to predict the rest of the curve with incredible accuracy.

In this paper, the authors applied this idea to the equations that govern quantum systems. They realized that the "wiggles" in the quantum system come from the natural energy of the qubits (the drift Hamiltonian). By telling the computer, "Hey, we know this part of the system wiggles at frequency X," the new method can take much larger timesteps. It's the difference between trying to count every step of a runner by looking at their feet (standard method) versus knowing their stride length and just counting their strides (Filon method).

They developed two specific versions of this method:

The Filon Method: This handles the natural, fast vibrations of the quantum system.
The Controlled Filon Method: This is the super-charged version. It also accounts for the "control pulses" (the microwave signals) which have their own specific rhythms. Since these pulses are often designed to match the qubit's natural frequency, this method is perfectly tuned to the problem.

The Results: Speeding Up the Quantum Race

The authors put their new methods to the test in two ways. First, they used a simple model called a "Rabi oscillator," which is like a single qubit being pushed back and forth. They compared their new methods against the current gold standard, known as the "Hermite method." The results were striking. To reach the same level of accuracy, the Filon methods could use timesteps that were 160 times larger than the Hermite method in some cases. In the world of computer simulations, being able to take 160 times bigger steps means you finish the job 160 times faster.

Then, they moved to a much harder, real-world scenario: simulating a CNOT gate on two superconducting transmon qubits. This is a fundamental operation for a quantum computer, where one qubit controls the state of another. They simulated this in two different "frames of reference": the "Lab Frame" (the real, messy, high-speed world) and the "RWA Frame" (the simplified, filtered world).

In the Lab Frame, where the vibrations are at their most intense, the Controlled Filon method was the clear winner.

For a target accuracy of $10^{-4}$ (a very precise simulation), the Controlled Filon method was 4.7 times faster than the best Hermite method and over 500 times faster than the basic Hermite method.
For an even higher accuracy of $10^{-7}$ , the Controlled Filon method was 6.1 times faster than the best Hermite method.

Even in the simplified RWA frame, where the problem is easier, the Controlled Filon method was still the fastest, beating the best Hermite method by about 2 times and the basic one by 10 times.

Why This Matters

The most exciting part of this paper isn't just that the math is faster; it's that it allows scientists to stop using the "sunglasses" (the RWA approximation). Because the Filon methods are so efficient, they can simulate the real, unfiltered Lab Frame dynamics without the computer crashing. This means engineers can design quantum gates that are more accurate and robust, knowing exactly how the system behaves in the real world, not just in a simplified model.

The authors showed that by understanding the rhythm of the quantum dance, we can stop trying to count every single step and start predicting the whole performance. Their "Controlled Filon" method is a new, powerful tool that could help speed up the development of the quantum computers of the future, making the dream of a working quantum machine a little bit more real.

Technical Summary: Filon Methods for Simulating Highly Oscillatory Controlled Quantum Systems

Problem Statement
The design and control of quantum computers, particularly those based on superconducting transmon qubits, rely heavily on the precise simulation of quantum dynamics governed by the time-dependent Schrödinger equation. A central bottleneck in quantum optimal control is the computational cost of evaluating the objective function and its gradient thousands of times. The primary obstacle to efficiency is the highly oscillatory nature of these systems: transition frequencies are in the GHz range, while gate durations are in the nanosecond range, resulting in hundreds to thousands of oscillations per gate. Standard numerical integrators require timesteps small enough to resolve the shortest period of the solution, leading to prohibitive computational costs. While the rotating frame transformation (RFT) combined with the rotating wave approximation (RWA) reduces oscillation, it introduces modeling errors and, in the case of coupled qudits, does not eliminate oscillation entirely.

Methodology
The authors adapt Filon quadrature, a technique for computing highly oscillatory integrals, into two new timestepping methods for linear systems of ordinary differential equations (ODEs): the Filon method and the Controlled Filon method.

Mathematical Framework: The methods operate on the ansatz that the solution $\psi(t)$ can be decomposed into slowly varying envelopes $f(t)$ and highly oscillatory carrier waves $e^{i\omega t}$ . Specifically, $\psi(t) = R^\dagger(t)f(t)$ , where $R(t) = e^{-i \text{diag}(\omega)t}$ and the frequencies $\omega$ are inferred from the system's drift Hamiltonian.
Filon Quadrature Adaptation: Instead of approximating the full solution directly, the methods approximate the integral of the product of the Hamiltonian and the solution by replacing the slowly varying envelope with a Hermite interpolating polynomial. The oscillatory part is integrated analytically.
- Filon Method: Designed for linear ODEs where the oscillation arises from the system's natural dynamics (drift Hamiltonian).
- Controlled Filon Method: An extension specifically tailored for controlled quantum systems where the Hamiltonian includes time-dependent control pulses. This method accounts for explicit oscillations in the control terms (carrier waves multiplying smooth envelopes) by shifting the ansatz frequencies accordingly.
Implementation: Both methods result in a linear system of equations of the form $S_I \psi_{n+1} = S_E \psi_n$ at each timestep. The authors utilize iterative solvers (GMRES) to solve these systems. The methods are designed to be compatible with discrete adjoint frameworks, allowing for the efficient computation of gradients required for pulse optimization.
Stability and Accuracy: Theoretical analysis establishes that the local truncation error decreases as the oscillation frequency increases. The methods are proven to be A-stable for orders $s=0$ and $s=1$ (corresponding to global orders 2 and 4, respectively) when applied to systems with purely imaginary eigenvalues.

Key Contributions

Novel Algorithms: The introduction of the Filon and Controlled Filon methods for solving linear ODEs with highly oscillatory solutions, specifically tailored for quantum control applications.
Error Analysis: Derivation of error bounds showing that the methods become more accurate as the oscillation frequency increases, a property that allows for fixed accuracy with significantly fewer timesteps than standard methods.
Control Pulse Integration: The development of the Controlled Filon method, which explicitly incorporates the oscillatory structure of control pulses (carrier waves) into the quadrature scheme, rather than treating them as part of a generic forcing term.
Efficiency in Lab Frame: Demonstration that these methods make direct simulation in the laboratory frame (without RWA) computationally affordable, thereby avoiding the approximation errors associated with the RWA.

Results
The authors validate the methods through numerical experiments on two systems: a Rabi oscillator and a CNOT gate implementation on two transmon qudits coupled to a resonator bus.

Rabi Oscillator: In both lab and RWA frames, the Filon and Controlled Filon methods achieved the same accuracy as Hermite methods (the current state-of-the-art for this application) using substantially larger timesteps. For example, in the lab frame at order $s=0$ , the Filon methods used timesteps roughly 17 to 160 times larger than the Hermite method to reach an error of $10^{-2}$ .
CNOT Gate Simulation: In a realistic simulation of a CNOT gate (550 ns duration), the Controlled Filon method was the most efficient method tested across all target accuracies.
- At a target error of $10^{-4}$ in the lab frame, the $s=0$ Controlled Filon method was 4.7 times faster than the best Hermite method ( $s=2$ ) and over 500 times faster than the $s=0$ Hermite method.
- At a target error of $10^{-7}$ , the $s=1$ Controlled Filon method was 6.1 times faster than the $s=2$ Hermite method.
- The methods maintained their efficiency advantage even when simulating in the RWA frame, though the gains were most pronounced in the lab frame where standard methods struggle with high-frequency resolution.

Significance
The paper claims that these methods represent a significant advancement in the computational efficiency of quantum optimal control. By leveraging the known oscillatory structure of the problem, the Filon methods drastically reduce the number of timesteps required to achieve high precision. This reduction allows for the direct simulation of quantum dynamics in the laboratory frame, bypassing the modeling errors inherent in the rotating wave approximation. The authors note that while high-order Hermite methods are already faster than low-order standard integrators, the Filon methods offer a further order-of-magnitude improvement (up to 500x over same-order Hermite methods), making high-fidelity simulation of complex quantum gates more feasible. The work also highlights that the methods are generalizable to any linear ODE system with oscillatory solutions, though the primary focus remains on quantum control.

The Problem: Counting the Unseeable

The Solution: The "Filon" Magic Trick

The Results: Speeding Up the Quantum Race

Why This Matters

More like this