Quantum Simulation Accurately Models Complex Systems

Zhen Huang and colleagues at the University of California, Berkeley, in a collaboration between the University of California, Berkeley, University of Michigan, Flatiron Institute, and Lawrence Berkeley National Laboratory, found that accurately representing Gaussian baths over extended time periods does not necessarily increase with simulation duration. Their thorough analysis reveals that the key limitation for long-time simulation lies not in the length of the simulation itself, but in the presence of non-analytic features within the bath’s spectral density. The study establishes bounds on the number of complex exponentials needed to represent bath correlation functions, offering valuable insight for both quantum and classical simulations of systems with memory.

Logarithmic scaling of computational cost for non-Markovian bath simulations

A computational complexity bound of O(log²(1/(ω_cε))) has been found for simulating non-Markovian Gaussian baths at zero temperature, applicable to super-Ohmic bosonic and gapped fermionic baths, irrespective of simulation time. Previously, simulating these systems was thought to require computational effort scaling polynomially with time. This new bound demonstrates that, for many scenarios, the number of calculations remains constant as simulation duration increases. This represents a significant advancement in computational efficiency, particularly for systems where the environment possesses ‘memory’, meaning its influence on the system extends beyond instantaneous interactions. The bath’s spectral density, denoted by J(ω), describes the distribution of frequencies present in the environment and is central to understanding its behaviour. Non-analytic features in J(ω), such as step discontinuities or singularities, dramatically impact the computational cost. The parameter ω_c represents a characteristic frequency of the bath, while ε defines the desired accuracy of the simulation. The logarithmic scaling implies that increasing the simulation time does not lead to a proportional increase in computational resources, provided the spectral density remains relatively smooth.

The findings are applicable to both quantum and classical simulations, offering a pathway to model more complex systems with memory effects efficiently. This has implications for diverse fields including quantum chemistry, materials science, and condensed matter physics, where accurately modelling environmental interactions is crucial. For sub-Ohmic bosonic baths at zero temperature, computational complexity remains independent of time, meaning the computational cost is constant regardless of how long the simulation runs. However, finite temperature and gapless fermionic baths introduce only polylogarithmic dependencies, indicating a slower increase in computational cost with time than previously anticipated. These bounds currently assume idealised conditions and do not yet account for the practical challenges of implementing these algorithms on real-world hardware, such as finite precision arithmetic and communication overhead between processors. The analysis considered the impact of maximal simulation time, inverse temperature, and the characteristics of the bath’s spectral density on computational cost, revealing that sharp features in the spectrum, rather than simulation duration, present the primary challenge. Specifically, the number of complex exponentials required to accurately represent the bath correlation function scales with the sharpness of these features.

Optimised Complex Exponential Sums for Efficient Quantum Environmental Simulations

Hierarchical equations of motion methods were central to this analysis, representing the interactions between a quantum system and its environment using ‘bath correlation functions’ to map how the environment influences the system over time. These correlation functions describe the statistical relationship between the environmental degrees of freedom at different times. The team employed recent advances within these methods, specifically focusing on representing these correlation functions as optimised sums of complex exponentials. This optimisation process involves carefully selecting the frequencies and amplitudes of the complex exponentials to minimise the error in representing the true bath correlation function. By overcoming limitations of methods like TEDOPA (Time-Dependent Perturbation Theory with Optimised Accumulation) and standard hierarchical equations of motion, which scale poorly with simulation time, this allows for efficient long-time simulations across diverse physical systems. TEDOPA and traditional hierarchical methods often suffer from the ‘curse of dimensionality’, where the computational cost grows exponentially with the number of environmental modes. This approach enables efficient modelling of environmental influences, and the analysis highlighted that sharp features in the bath’s spectral density, not simulation duration, present the primary challenge. The use of complex exponentials allows for a compact and efficient representation of the bath correlation function, reducing the number of parameters that need to be stored and manipulated during the simulation.

Simulation efficiency persists despite complex energy level representation

Tools used to simulate complex systems, particularly those exhibiting non-Markovian behaviour, are undergoing refinement. Accurately modelling these ‘environments’, or Gaussian baths, does not necessarily become more demanding with longer simulations. Representing the bath’s spectral density, a measure of its energy levels, as a sum of complex exponentials is crucial for this efficiency. The efficiency arises because the complex exponential representation allows for a systematic reduction in the number of terms needed to achieve a desired level of accuracy. Despite the acknowledged issue of ‘sharp’ energy levels within these systems, which can complicate calculations, the finding that simulation complexity isn’t always tied to simulation length is significant for many realistic models. These sharp energy levels correspond to the non-analytic features in the spectral density, and their presence necessitates a more careful treatment to avoid inaccuracies in the simulation.

Fundamental limits on simulating complex systems interacting with their environments have been established, revealing a surprising result about computational cost. For many realistic scenarios, the difficulty of these simulations does not automatically increase as the simulation runs for longer periods, challenging previous assumptions of escalating computational demands. Instead, the primary obstacle to modelling these ‘open quantum systems’, systems exchanging energy and information with their surroundings, lies in the characteristics of the environment itself, specifically sharp, abrupt changes in its energy levels, known as non-analytic features. Understanding and mitigating the impact of these non-analytic features is therefore paramount for developing efficient and accurate simulation methods. This work provides a rigorous framework for analysing the computational cost of simulating non-Markovian environments and offers valuable guidance for choosing appropriate simulation techniques.

The research demonstrated that simulating complex quantum systems interacting with their environments does not necessarily become more computationally demanding as the simulation time increases. This is important because it suggests that longer, more realistic simulations are achievable without prohibitive computational cost, provided the environment’s energy levels are relatively smooth. The primary limitation, however, was identified as sharp, non-analytic features within the environment’s spectral density, which require more complex calculations. Future work could focus on developing algorithms specifically designed to efficiently handle these problematic spectral features, potentially using techniques to smooth or approximate them without sacrificing accuracy.