Quantum optimal control of steady orbits

arXiv:2606.15383 · quant-ph, math-ph, math.MP, physics.chem-ph · Submitted 2026-06-13 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Quantum optimal control of steady orbits".

Mira: Here's the gist of the paper. Periodically driven dissipative systems can settle into steady orbits – fixed loops on their dynamical manifolds The formalism proposed here designs a periodic control sequence that makes the system approach a steady orbit passing through user-specified waypoints This work proposes an efficient quantum control strategy for stroboscopic steady states and limit cycles Problem Statement The primary challenge addressed is controlling steady orbits with the same care as quantum gates and destination states, which involves engineering an entire limit cycle The difficulty lies in the infinite loop problem that defeats gradient ascent pulse engineering (GRAPE) because it relies on explicit numerical propagation in the time domain Unlike prior art, this work targets asymptotic limit cycles rather than just effective Hamiltonians or unitary gates Steady-orbit GRAPE Formulation The standard Liouville space formulation is used to describe the system's evolution, which is a dissipative generalization of the Liouville–von Neumann equation The evolution generator L can include terms such as kinetics, diffusion, hydrodynamics, and magic angle spinning The fidelity in the GRAPE framework is defined as a function of the overlap between the system state at the end of a one-off control sequence and some desired destination state Newton-Raphson Optimization Route The authors propose using a recently designed Newton-Raphson type method for computing steady states, which delivers the steady orbit in just a few propagator-vector multiplications This approach simplifies the optimization target by making the asymptotic fidelity gradient much easier to differentiate because matrix powers are now absent and waypoint projectors are fixed The asymptotic fidelity gradient is obtained by applying a local patch to the usual GRAPE expression, which is identical except for one local patch involving applying-1 P− where P P P P = N N−1 2 1  Implementation and Distortions The steady orbit version of GRAPE (SO-GRAPE) is a localized patch in any existing Liouville space GRAPE implementation The algorithm is computationally efficient, scaling linearly with the number of control sequence slices because the required operation can be computed using only sparse matrix-vector operations Instrumental distortions, such as restricted amplifier bandwidth, are integrated into the optimization algorithm to account for real-life hardware effects DNP Pulse Sequence Optimisation The method is illustrated using a dynamic nuclear polarisation experiment where the SO-GRAPE algorithm optimises the content of the DNP block to achieve maximum nuclear magnetisation at n → ∞ The optimal pulse depends on the initial guess, and one notable improvement when starting from known highly optimal DNP pulse sequences is to return the electron magnetisation precisely on the Z axis at the end of the pulse The results show that greater attenuation in higher frequency lobes compels the optimal pulse waveform to spend longer irradiating those frequencies to achieve a comparable magnetisation tilt angle The proposed method is distinct from prior art because it targets through waypoint projectors a user-specified limit cycle, not just a generator or a stroboscopic eigenstate of the control sequence period map The utility of the method is demonstrated for any dissipative dynamics that is representable by an action of an evolution generator exponential on a state vector The algorithm is available, along with its examples, in versions 2.12 and later of Spinach library.

Kai: Acknowledgements This work was funded by the Deutsche Forschungsgemeinschaft through SFB 1527 (Project No.

Mira: 454252029), by EPSRC (EP/Y035267/1) The work at Weizmann Institute was supported by a research grant from the Center for New Scientists and a research grant from Anton Rabie, Stanley and Tanya Rossby Endowment Fund, Danielle Bitton & Raphy Benbaron, and two benefactors who have chosen to remain anonymous Computational work was carried out on the Weizmann Faculty of Chemistry's high-performance computing facility CHEMFARM, which is supported in part by the Ben May Center for Chemical Theory and Computation Matlab technical support from the awesome engineering team at MathWorks is also gratefully acknowledged Author Contributions SAJ, GM: DNP pulse sequence exploration and analysis; YZ: HiPER instrument characterisation; RH, HEM, GS: HiPER instrument design and characterisation, DNP experiments; MK, CA: Spinach helper functions; IK: SO-GRAPE theory, algorithms, and Spinach implementation The paper is 2606.15383.

Paper summary: Kai: So looking at "Quantum optimal control of steady orbits," what they’ve done is design a method to find a sequence that gets your quantum system into a specific fixed loop when it's driven and damped.

Mira: It’s important because it addresses the difficulty of controlling these entire limit cycles, which is usually much trickier than just targeting a single final state or an effective Hamiltonian, like the work done in magnetic resonance >

Lev: They are proposing a Newton-Raphson type method that computes this steady orbit quickly, using only a few propagator multiplications because they simplify the gradient calculation significantly >

Kai: The implication is that we can now tackle dissipative systems where we want to settle into these long-term periodic attractors, which are common in things like quantum cooling engines or precision metrology devices >

Mira: It changes the focus from just finding a good path to a final destination state to actively engineering the entire trajectory around a specific, repeating loop defined by you >

Lev: And as an engineer, I see that it's efficient because it uses sparse operations, which means on real hardware, we might be able to handle these control sequences without getting bogged down in massive computational overhead >

Kai: So for the listener driving or cooking, this just means that if your system is settling into a repeating pattern instead of just stopping at one point, there's now a more direct way to program that pattern >

Conclusion: Kai: So we've been looking at how this paper, "Quantum optimal control of steady orbits," tackles getting a quantum system into a specific repeating loop when it's being driven and damped.

Mira: Right, so the authors are proposing this new way to find those steady orbits, these fixed loops that the system settles into over time.

Lev: What they’re focusing on is this whole "limit cycle" thing, which is harder than just finding one final destination state or a stable point in space.

Kai: They say they've designed a periodic control sequence to make the system approach these steady orbits, and it passes through waypoints you specify.

Mira: The paper claims they're using a Newton-Raphson type method for this, which supposedly lets them find that steady orbit in just a few propagator multiplications.

Lev: That sounds promising because it simplifies the math; they say the gradient calculation becomes much easier to differentiate because they don't have those matrix powers floating around.

Kai: And they introduce this thing called SO-GRAPE, which is basically just a localized patch you can put into any existing Liouville space GRAPE implementation.

Mira: So, it’s not inventing an entirely new optimization framework but rather a clever way to patch onto what we already have to target these specific cycles.

Lev: And they even factored in real-world hardware issues like restricted amplifier bandwidth directly into the optimization algorithm, which is a big deal for practical use.

Kai: It's interesting because it’s applied to dissipative dynamics, which is where most of our real-world physics lives, and the way they handle those constraints makes it relevant.

Mira: The authors state this method targets a user-specified limit cycle rather than just some abstract generator or a stroboscopic eigenstate of the control period map.

Lev: So, the actual payoff here is that this works for any dissipative dynamics that can be described by an evolution generator exponential on a state vector.

Kai: It’s available in version twelve point one two and later of the Spinach library, which means it's already in use for some experiments.

Mira: It shows that we can control these repeating patterns with this approach, not just single events or final states.

Lev: So, what does this mean for us? It suggests a more robust way to design control pulses for systems that naturally settle into cycles instead of just stopping somewhere.

Kai: We need to see how well the fidelity holds up when we actually run these sequences on our physical hardware with real noise and dissipation.

Shebha Anandhi Jegadeesan, Yujie Zhao, Robert Hunter, Hassane El Mkami, Maximilian Keitel, Callum Musselwhite, Graham Smith, *Guinevere Mathies*, *Ilya Kuprov

Department of Chemistry, University of Konstanz, Germany. · School of Physics and Astronomy, University of St Andrews, United Kingdom. · School of Chemistry and Chemical Engineering, University of Southampton, United Kingdom. · Department of Chemical and Biological Physics, Weizmann Institute for Science

quant-ph, math-ph, math.MP, physics.chem-ph

Submitted: 2026-06-13

Updated: 2026-11-13

Code: https://github.com/IlyaKuprov/Spinach

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: Periodically driven dissipative systems can settle into steady orbits – fixed loops on their dynamical manifolds The formalism proposed here designs a periodic control sequence that makes the…

Key concepts

Steady Orbits
These are fixed loops or stable periodic solutions that a dissipative system settles into over time. The paper focuses on controlling these specific, long-term behaviors rather than just short control sequences or static states.
Limit Cycle Engineering
The goal is to design a complete sequence of controls that forces the system onto a desired steady orbit, which is essentially an entire closed loop in the system's state space. This contrasts with methods that only target specific points or fixed periodic states.
SO-GRAPE
This stands for Steady-Orbit Gradient Ascent Pulse Engineering. It is a localized patch applied to standard GRAPE methods, optimized using a Newton-Raphson approach. This method simplifies optimization by eliminating the need for explicit time propagation, making it computationally efficient for finding steady orbits.
Liouville Space Formulation
This mathematical framework describes the evolution of the system's state in a way that accounts for dissipation (energy loss) alongside standard quantum evolution. It allows researchers to define control fidelity based on how well the system state matches a desired target after a sequence of operations.

Terminology

Summary

Periodically driven dissipative systems can settle into steady orbits – fixed loops on their dynamical manifolds The formalism proposed here designs a periodic control sequence that makes the system approach a steady orbit passing through user-specified waypoints This work proposes an efficient quantum control strategy for stroboscopic steady states and limit cycles

Problem Statement

The primary challenge addressed is controlling steady orbits with the same care as quantum gates and destination states, which involves engineering an entire limit cycle The difficulty lies in the infinite loop problem that defeats gradient ascent pulse engineering (GRAPE) because it relies on explicit numerical propagation in the time domain Unlike prior art, this work targets asymptotic limit cycles rather than just effective Hamiltonians or unitary gates

Steady-orbit GRAPE Formulation

The standard Liouville space formulation is used to describe the system's evolution, which is a dissipative generalization of the Liouville–von Neumann equation The evolution generator L can include terms such as kinetics, diffusion, hydrodynamics, and magic angle spinning The fidelity in the GRAPE framework is defined as a function of the overlap between the system state at the end of a one-off control sequence and some desired destination state

Newton-Raphson Optimization Route

The authors propose using a recently designed Newton-Raphson type method for computing steady states, which delivers the steady orbit in just a few propagator-vector multiplications This approach simplifies the optimization target by making the asymptotic fidelity gradient much easier to differentiate because matrix powers are now absent and waypoint projectors are fixed The asymptotic fidelity gradient is obtained by applying a local patch to the usual GRAPE expression, which is identical except for one local patch involving applying-1 P− where P P P P = N N−1 2 1 

Implementation and Distortions

The steady orbit version of GRAPE (SO-GRAPE) is a localized patch in any existing Liouville space GRAPE implementation The algorithm is computationally efficient, scaling linearly with the number of control sequence slices because the required operation can be computed using only sparse matrix-vector operations Instrumental distortions, such as restricted amplifier bandwidth, are integrated into the optimization algorithm to account for real-life hardware effects

DNP Pulse Sequence Optimisation

The method is illustrated using a dynamic nuclear polarisation experiment where the SO-GRAPE algorithm optimises the content of the DNP block to achieve maximum nuclear magnetisation at n → ∞ The optimal pulse depends on the initial guess, and one notable improvement when starting from known highly optimal DNP pulse sequences is to return the electron magnetisation precisely on the Z axis at the end of the pulse The results show that greater attenuation in higher frequency lobes compels the optimal pulse waveform to spend longer irradiating those frequencies to achieve a comparable magnetisation tilt angle

Conclusions and Outlook

The proposed method is distinct from prior art because it targets through waypoint projectors a user-specified limit cycle, not just a generator or a stroboscopic eigenstate of the control sequence period map The utility of the method is demonstrated for any dissipative dynamics that is representable by an action of an evolution generator exponential on a state vector The algorithm is available, along with its examples, in versions 2.12 and later of Spinach library. p. 6

Acknowledgements

This work was funded by the Deutsche Forschungsgemeinschaft through SFB 1527 (Project No. 454252029), by EPSRC (EP/Y035267/1) The work at Weizmann Institute was supported by a research grant from the Center for New Scientists and a research grant from Anton Rabie, Stanley and Tanya Rossby Endowment Fund, Danielle Bitton & Raphy Benbaron, and two benefactors who have chosen to remain anonymous Computational work was carried out on the Weizmann Faculty of Chemistry's high-performance computing facility CHEMFARM, which is supported in part by the Ben May Center for Chemical Theory and Computation Matlab technical support from the awesome engineering team at MathWorks is also gratefully acknowledged

Author Contributions

SAJ, GM: DNP pulse sequence exploration and analysis; YZ: HiPER instrument characterisation; RH, HEM, GS: HiPER instrument design and characterisation, DNP experiments; MK, CA: Spinach helper functions; IK: SO-GRAPE theory, algorithms, and Spinach implementation The paper is 2606.15383.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Development of a steady-orbit GRAPE method that designs a periodic control sequence driving a dissipative quantum system towards a steady orbit passing through user-specified waypoints. This allows for engineering an entire limit cycle rather than just a single destination state, which is crucial for applications like coherent oscillators and precision metrology devices.

  2. Integration of instrumental distortions, such as restricted amplifier bandwidth, directly into the optimization algorithm by calculating a filter function that can be used as a differentiable distortion handle. This enables the design of instrument adapted control waveforms, accounting for real-life hardware limitations like those in the HiPER spectrometer described in Section 4.

  3. Implementation of a Newton-Raphson type method to compute stroboscopic steady states, which simplifies the calculation of the asymptotic fidelity gradient by eliminating expensive matrix power expansions and enabling efficient sparse matrix-vector operations. This improves computational efficiency while maintaining the ability to find attractors in dissipative dynamics where prior methods struggled due to numerical instability.

  4. A patch for existing Liouville-space GRAPE implementations (SO-GRAPE) that modifies the standard algorithm to first compute the steady orbit before running the standard GRAPE gradient calculation, ensuring fidelity and gradient calculations are based on the correct attractor state. This allows leveraging existing high-performance numerical optimization tools while targeting limit cycles instead of fixed points.

Related papers