Quantum Computers Efficiently Model More Complex Systems with New Mathematical Technique

Ke Wang and colleagues at University of Michigan have developed a pivot-shifted Carleman linearization framework that extends the range of nonlinear systems suitable for efficient quantum simulation. The framework exhibits logarithmic scaling of truncation order with simulation time and precision, key to removing conventional limitations on initial conditions for both stable and unstable systems. Numerical results using established equations, including the logistic and Lotka-Volterra models, confirm improved stability, accuracy, and exponential error decay, indicating a sharp advancement in applying Carleman-based quantum algorithms to complex dynamical systems

Logarithmic scaling enhances quantum simulation of dynamical systems

The truncation order now scales logarithmically with simulation time and target precision, a substantial improvement over prior methods lacking this efficiency. This unlocks the possibility of simulating systems for significantly longer durations and with greater accuracy, something previously unattainable due to exponential growth in computational cost. A pivot-shifted Carleman linearization framework was introduced, effectively reshaping the dynamics before applying standard quantum techniques.

The new approach broadens the range of stable and unstable systems amenable to quantum simulation by removing the conventional requirement for a lower bound on initial conditions. Experiments using the logistic and Lotka-Volterra equations revealed exponential error decay as the truncation order increased, signifying improved simulation precision. Theoretical predictions of sustained accuracy were validated by observing long-time convergence of the truncated Carleman embedding for stable systems within the shifted coordinate system. Furthermore, the method provided short-time convergence guarantees for unstable systems following the coordinate shift, expanding the range of solvable problems. Careful selection of a ‘pivot’ state, a key element of the new approach, was also confirmed by experiments to optimise both stability and accuracy during the simulation process.

Quantum simulation of nonlinear dynamics via linear mappings

Quantum algorithms are of interest because of their potential for speedups over classical methods. Quantum simulation, particularly of quantum systems, is a key algorithmic primitive, though it suffers from the curse of dimensionality which limits classical approaches. The inherent linearity of quantum mechanics makes it well-suited for implementation on quantum computers, leading to advances in Hamiltonian and Lindblad simulation. Recent work explores quantum advantages for classical tasks, including linear system solvers and algorithms for linear differential equations, by exploiting quantum mechanics’ linearity.

However, many scientific and practical phenomena are governed by nonlinear differential equations, posing substantial simulation challenges even for classical high-performance computing platforms. The fundamental difficulty arises because quantum evolution is intrinsically linear, while target dynamics are nonlinear, necessitating a linear framework or intermediate operations to reproduce nonlinear behaviour. This mismatch between linear quantum evolution and nonlinear target dynamics is a central obstacle in designing quantum algorithms for nonlinear differential equations.

Various strategies have been developed to address this mismatch in existing quantum approaches. The Koopman and von Neumann frameworks embed dynamics into a linear operator evolution acting on observables, while variational and hybrid quantum-classical methods recast nonlinear dynamics as optimisation problems using parameterised quantum circuits. Linearization or approximation schemes are also employed to mimic nonlinear evolution. Carleman linearization has emerged as a promising framework, lifting the nonlinear system into a higher dimensional linear system by enlarging the state space and encoding higher order monomials of the solution.

This construction is systematic, compatible with quantum linear system techniques, and admits rigorous error analysis, yielding efficient quantum simulation algorithms under dissipativity assumptions and a condition analogous to the Reynolds number in fluid dynamics. Extensions beyond the dissipative regime have also been established for nonresonant and stable systems. Despite this progress, current Carleman linearization applicability relies on restrictive assumptions, narrowing the class of treatable nonlinear systems and imposing conditions on initial values to guarantee convergence of the truncated Carleman expansion. For example, consider the n-dimensional quadratic ordinary differential equation ∂tx = F2x⊗2 + F1x + F0, x = x0, where x ∈Rn, x⊗2 ∈Rn2, F2 ∈Rn×n2, F1 ∈Rn×n, and F0 ∈Rn. Existing analyses typically require the linear part to be dissipative, meaning its spectral abscissa satisfies α(F1). Many stable systems of practical interest fall outside this framework, making establishing a quantum advantage challenging. The logistic equation ∂tx = x−x2, which remains bounded and converges to a stable equilibrium for every initial condition x > 0, is a simple example that fails the condition α(F1).

Quantum simulation of nonlinear systems via pivot-shifted Carleman linearization

Pioneering methods to simulate complex systems on quantum computers are offering potential breakthroughs in fields from materials science to finance. This new framework, leveraging Carleman linearization, expands the range of solvable nonlinear equations. While the logistic and Lotka-Volterra equations serve as compelling proof-of-concept examples, a critical question remains: can this pivot-shifted approach scale to tackle the significantly more intricate, higher-order nonlinearities prevalent in real-world phenomena.

This work represents a significant advance in quantum algorithm design, relaxing the demanding initial condition requirements historically limiting Carleman linearization. By employing a pivot-shifted Carleman linearization, a longstanding requirement for specific initial conditions has been removed, enabling the simulation of a broader range of unstable and stable dynamics. This technique circumvents previous restrictions and unlocks access to previously intractable problems by reshaping the system’s behaviour before applying quantum methods. The framework’s efficiency stems from a logarithmic scaling of truncation order with both simulation time and desired precision.

The researchers developed a new framework using pivot-shifted Carleman linearization to simulate a wider range of nonlinear ordinary differential equations on quantum computers. This method removes a previous restriction on initial conditions, allowing for the simulation of both stable and unstable systems like the logistic and Lotka-Volterra equations. The technique achieves this by reshaping the system’s dynamics prior to quantum processing, improving stability and accuracy. The authors demonstrated that the computational effort required scales logarithmically with simulation time and target precision, suggesting a potentially efficient approach.