Tu Delft Team Introduces Reduced Basis Algorithm for Polynomial Nonlinear Equations

Scientists at TU Delft, in collaboration with the Delft Institute of Applied Mathematics and Delft University of Technology, have developed a new method for addressing nonlinear differential equations using quantum computers, despite the inherent linearity of quantum evolution. Monica Lăcătuş and colleagues present a reduced basis algorithm (RBA) that efficiently solves these equations by composing polynomial updates and constructing a linear operator suitable for implementation on a quantum computer. This approach strategically shifts the primary computational load to a classical preprocessing stage, focused on determining the reduced monomial basis, while crucially maintaining accuracy equivalent to the initial time discretization scheme employed. The algorithm’s qubit requirement scales logarithmically with grid size for partial differential equations, offering a potential pathway towards managing the substantial resources needed for complex scientific simulations, a long-standing challenge in the field. Reduced qubit scaling enables simulation of complex partial differential equations A qubit requirement of O(DlogN + nmD+1 log p) now exists for simulating partial differential equations, representing a substantial improvement over previous methods. Historically, simulating nonlinear partial differential equations on quantum computers has been severely limited by the exponential scaling of qubit requirements or the necessity for numerous copies of the quantum state, both of which quickly render accurate modelling of complex phenomena intractable, even with moderately sized computational grids. The reduced basis algorithm circumvents these limitations by strategically shifting the computational burden to classical computers, where essential components are pre-calculated and a streamlined representation of the nonlinear dynamics is constructed. This allows for simulations with grid sizes previously inaccessible due to prohibitive resource demands. The significance of this lies in the ability to model more realistic and complex systems, moving beyond simplified representations often necessitated by computational constraints. For polynomial ordinary differential equations, a lifted register requires at most O(nm log p) qubits, where ‘n’ represents the system’s dimensionality, ‘m’ the number of timesteps used in the time-discretisation scheme, and ‘p’ the polynomial degree defining the nonlinearity. Achieving a qubit requirement of O(D log N + nmD+1 log p) for partial differential equations discretised across N grid points was possible, with ‘D’ denoting the number of spatial dimensions. The logarithmic dependence on grid size is particularly noteworthy, as it implies that the computational cost increases much more slowly with increasing resolution, a critical factor for practical applications. Numerical tests utilising the Lorenz system, a benchmark for chaotic systems, and the one-dimensional Burgers equation, a model for fluid dynamics, confirmed that the algorithm accurately reproduces the discrete time nonlinear dynamics. These tests validated the algorithm’s ability to capture the essential behaviour of these systems without introducing significant errors. A trade-off between timestep composition, reduced basis growth (the size of the pre-calculated basis), and locality of calculations was revealed during testing. However, these qubit counts do not yet account for the substantial overhead required for state preparation and measurement, representing a significant hurdle to near-term implementation on existing, noisy intermediate-scale quantum (NISQ) hardware. The Lorenz system, frequently used for testing chaotic systems, presents a particular challenge due to its sensitivity to initial conditions and complex dynamics. The successful reproduction of its behaviour by the RBA demonstrates the algorithm’s robustness and accuracy in handling such systems. Similarly, the Burgers equation, a fundamental equation in fluid dynamics, requires accurate modelling of nonlinear advection terms, which the RBA effectively addresses. The choice of these two systems as test cases highlights the algorithm’s versatility and potential for application in diverse scientific domains. Furthermore, the algorithm’s performance is influenced by the choice of time discretisation scheme and the size of the reduced basis; careful optimisation of these parameters is crucial for achieving optimal performance. Easing quantum simulation through pre-processed data and reduced qubit requirements Scientists at TU Delft and collaborating institutions have devised a method to tackle nonlinear equations, a persistent obstacle to utilising quantum computers for realistic scientific modelling. The reduced basis algorithm demonstrably reduces the quantum resources needed, but a delicate balancing act exists between the complexity of the time steps, the growth of the simplified basis, and the locality of calculations. Reducing the number of qubits, the fundamental building blocks of quantum computing power, is vital for practical application of these systems, as qubit coherence times and gate fidelities remain limited in current quantum hardware. The development of algorithms with lower qubit requirements is therefore a crucial step towards realising the full potential of quantum computing for scientific discovery. The new reduced basis algorithm shifts the heaviest computational load to conventional computers, pre-processing data before it reaches the quantum processor. This strategy is based on the observation that many of the computational operations involved in solving nonlinear differential equations can be performed efficiently on classical computers, leaving the quantum computer to focus on the tasks that it excels at, such as simulating quantum phenomena. A new technique for solving polynomial nonlinear differential equations on quantum computers has been established by researchers at TU Delft and collaborating institutions, moving beyond the limitations of earlier methods. Intensive pre-processing on conventional computers streamlines complex calculations by identifying essential components before engaging the quantum processor. Consequently, the number of qubits required scales logarithmically with grid size, a significant improvement over approaches demanding exponential growth or numerous quantum state copies. This logarithmic scaling is a key advantage, as it allows the algorithm to handle increasingly complex problems with a manageable increase in computational resources. The pre-processing stage involves constructing a reduced basis, which is a set of functions that can accurately represent the solution of the differential equation with a limited number of terms. This reduces the complexity of the quantum computation and lowers the qubit requirement. The implications of this work extend beyond the specific equations tested. The reduced basis algorithm provides a general framework for tackling nonlinear differential equations on quantum computers, potentially applicable to a wide range of scientific and engineering problems, including climate modelling, materials science, and drug discovery. While challenges remain in terms of state preparation, measurement overhead, and the limitations of current quantum hardware, this research represents a significant step forward in harnessing the power of quantum computing for solving complex scientific problems. Future research will focus on optimising the algorithm for specific applications and exploring techniques for reducing the overhead associated with state preparation and measurement. Researchers developed a reduced basis algorithm to solve polynomial nonlinear differential equations on quantum computers. This method shifts the computational burden to classical computers during a pre-processing stage, identifying essential components before quantum processing begins. Consequently, the number of qubits needed grows logarithmically with grid size, offering a more efficient scaling than previous approaches. The authors intend to optimise the algorithm for specific applications and reduce state preparation overhead in future work. 👉 More information🗞 Reduced basis algorithm for solving nonlinear differential equations on quantum computers🧠 ArXiv: https://arxiv.org/abs/2606.13457 Stay current. See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals. Tags: