Spectral decomposition and high-accuracy Greens functions: Overcoming the Nyquist-Shannon limit via complex-time Krylov expansion

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

Kai: We're looking at the paper: "Spectral decomposition and high-accuracy Greens functions".

Mira: Here's the gist of the paper. The accurate computation of low-energy spectra of strongly correlated quantum many-body systems, typically accessed via Green’s-functions, is a long-standing problem posing enormous challenges to numerical methods.

Kai: The paper demonstrates how augmenting the time series with complex-time Krylov states can overcome the Nyquist-Shannon limit and significantly improve the frequency resolution of Green’s functions, even for two-dimensional systems How it works The method addresses the limitation imposed by Fourier transforming a time series, where the Nyquist-Shannon theorem limits the frequency resolution according to ∆ω = 2π/T.

Mira: This is overcome by augmenting the time series with complex-time Krylov states.

Kai: The general frequency-dependent Green’s function GAB(ω) is regularized by introducing a finite broadening η > 0.

Paper summary: Kai: So we were just talking about how this paper aims to solve a long-standing problem in computing low-energy spectra for strongly correlated systems using Green’s functions because standard Fourier transforms hit a Nyquist-Shannon limit.

Mira: The thesis is that augmenting the time series with complex-time Krylov states can overcome this sampling limit and significantly improve frequency resolution of Green’s functions, even for two-dimensional systems.

Lev: It claims they do this by recasting the time integration using an operator (T, omega) that boosts evolution onto all intervals

pT, (p+one)T), and then decomposing the Green's function into a zeroth order part TAB and a correction term K TAB <ref:2411.09680#pg1>. [Kai: They argue that harnessing additional information about spectral properties encoded in this correction term K TAB(omega) allows for the recovery of faint signals, contrasting this approach with compressed sensing schemes.

Mira: The real point is that they show how to approximate this contribution K TAB efficiently using tensor network methods, which is where the computational advantage comes from.

Lev: They spend a lot of effort on constructing a complex-time Krylov space generated by evolving the initial state with step size tau along a complex contour of segments.

Kai: This construction creates states that have large overlap with low-energy eigenstates, and they approximate the full Green’s function using an expression involving Y AB(t, omega + i eta) and the boost operator.

Mira: The analysis then bounds the approximation error using Arnoldi iterations, leading to a practical noise threshold R two delta t eta D h r,r-one where h r,r-one is the Arnoldi residual <ref:2411.09680#pg1>.

Lev: Crucially, that Arnoldi residual can be exponentially suppressed as long as delta t < -one which means you get exponential convergence towards the exact Green’s function for large rank r and small time steps delta t <ref:2411.09680#pg3,exponential convergence towards the exact Green’s function>.

Kai: The paper shows that increasing the complex angle alpha is also beneficial, though they have to avoid alpha > zero point one to prevent numerical instabilities <ref:2411.09680#pg3>.

Mira: So, to summarize, the paper demonstrates how this technique overcomes the Nyquist-Shannon limit by using Krylov states for efficient spectral decomposition and correction terms.

Lev: It’s a method that shows how you can get better frequency resolution than standard sampling limits allow.

Conclusion: Kai: The title itself is quite descriptive, focusing on spectral decomposition and overcoming that specific sampling limitation using complex-time Krylov expansion. It points to the core technical mechanism of the work.

Mira: What this means for us is that we can now compute these low-energy spectra with much higher accuracy than what was possible before, pushing the boundaries on what's numerically feasible in quantum many-body physics.

Lev: For someone building real hardware, it’s important to know that this method involves managing the overhead of those complex time steps and ensuring you have enough resources to handle the TN operations needed for evaluating overlaps.

Kai: The authors showed that for two-dimensional models, this technique moves beyond what was previously possible with simpler tensor network approaches which were mostly stuck in one dimension.

Mira: They are showing that if you can get good results from this, it opens up possibilities for studying more complex quantum systems where correlations are everywhere.

Lev: But the paper does have its limitations, and they state that increasing the frequency of those complex-time Krylov space steps is beneficial for convergence analysis rather than just reaching large final evolution times T.

Kai: They also flagged that you should prefer checking the Arnoldi residual to decide if convergence is sufficient, because it gives a threshold to discriminate spectral signals from errors like replica features.

Mira: So, the main implication is that we have a tool now for high-accuracy spectral information in strongly correlated systems.

Lev: It's a tool that lets us probe physics in ways that previous numerical methods couldn't reach due to those sampling constraints.