A better boundary for simulating quantum systems and their surroundings

Where should a quantum environment end?

A quantum system interacting with its surroundings is called an open quantum system. Its environment can absorb an excitation, shift the system’s characteristic frequencies and influence its later evolution. Representing those effects accurately is a central challenge in simulation.

In an arXiv preprint submitted on November 9, 2026, Pengju Chen, Da-Wei Luo and Ting Yu of Stevens Institute of Technology develop a framework for describing that environmental response. Their work combines mathematical derivations with numerical tests, rather than a laboratory experiment.

The starting point is a familiar representation: a qubit, a quantum system with two levels, coupled to a chain of harmonic oscillators. These oscillators are mathematical models of environmental modes that exchange excitations with their neighbours. A suitable change of basis can turn an environment whose modes connect directly to the qubit into this chain without changing the response seen by the qubit.

An indefinitely long chain cannot simply be retained in a finite calculation. But abruptly cutting it creates an artificial end that can reflect an excitation back towards the qubit.

Replacing the missing chain with its response

The authors describe the omitted environment through its self-energy. Here, this is a frequency-dependent mathematical response that captures both absorption and environment-induced frequency shifts.

Moving the boundary between the explicitly simulated system and its environment changes the self-energy through a recursive rule. Each step removes one oscillator from the explicit description while incorporating its influence into the response of the remaining environment.

For chains whose frequencies and neighbour-to-neighbour couplings approach constant values far from the qubit, this recursion has a fixed point: a response unchanged by adding another identical step. The authors call it the asymptotic boundary self-energy, or BSE.

Its absorption spectrum has the Wigner semicircle shape, rising and falling across a finite frequency band. The paper connects this established semicircular tail to the self-energy construction; it does not claim to discover the semicircle itself.

A second version, the chain-extended BSE, adjusts the tail parameters using the trend in the available chain coefficients. For the class of spectra considered, the authors fit corrections that decrease approximately as the inverse square of the site index. This incorporates information that the constant-parameter tail discards.

Measuring error without moving the goalposts

Another contribution is a distance between two self-energies, called a pseudo-hyperbolic distance. It compares their complex responses, including both absorption and frequency shifts.

On the real-frequency axis, this distance stays unchanged when both responses are propagated through the same lossless chain segment with nonzero couplings. Its applicability also requires suitable finite boundary values with positive real parts. Under those conditions, comparing two environmental models does not depend on whether the comparison is made at the chain’s termination or at the qubit.

This separates two sources of approximation error. Depth error arises because the chosen tail model does not yet represent the actual environment accurately at the selected chain depth. Closure error arises when a finite replacement fails to reproduce that chosen tail model.

The distinction matters: retaining more exact chain sites cannot necessarily overcome limitations in the finite replacement attached to them.

What the benchmark showed

The numerical demonstration used an initially excited qubit and an initially unexcited, uncorrelated environment. Its spectral density—the distribution of coupling strength across frequencies—combined a frequency-proportional background with a Gaussian enhancement inside a finite band.

The authors compared two analytic BSE closures with three finite replacements, each containing ten damped oscillator modes. Reference dynamics were calculated numerically from the original continuous spectrum, with retained chain lengths ranging from 10 to 60 sites.

Fitting the finite replacements using the new distance reduced their reported root-mean-square closure error from about 0.011 for the earlier time-correlation fitting approach to about 0.00029: roughly a 38-fold reduction in that particular error measure.

For at least 18 retained sites, the chain-extended BSE had the smallest population, spectral and cut-invariant errors among the tested closures. At 60 sites, its relative population error was about 40 times smaller than the next-best closure.

Accuracy is not the whole cost

These results concern one structured environment and dynamics restricted to at most one excitation. The chain-extended method also depends on chain coefficients following its assumed trend. Broader dynamical tests remain future work.

The new distance is not, by itself, a guarantee of accurate dynamics over a finite time interval. Even a hard cutoff can appear harmless before a reflected excitation returns.

Nor is the most accurate analytic closure automatically the cheapest: in the authors’ implementation, it still requires continuum integration. Finite damped-mode replacements remain useful for computational efficiency. The preprint’s contribution is a more systematic way to construct and assess these boundaries, not a demonstrated universal solution for quantum simulation.