Can multi-level Monte-Carlo address the sign problem in a finite-density Bose gas?

arXiv:2611.04735 · hep-lat, nucl-th · Submitted 2026-11-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Lattice. The day in lattice field theory: what the simulations put on the table.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Can multi-level Monte-Carlo address the sign problem in a finite-density Bose gas?".

Mira: Here's the gist of the paper. Numerical simulation of field theories with complex actions is frustrated by the sign problem, which causes an exponential growth in statistical sampling noise.

Kai: In this work, a multi-level Monte-Carlo approach to reweighting may offer a resolution to the sign problem at high densities and low temperatures.

Mira: The Sign Problem in Field Theories Field theories with a complex action are extremely difficult to simulate because the imaginary part introduces a nontrivial complex phase into the exponential weight.

Kai: This phase prevents Monte-Carlo importance sampling of the path integral, which is known as the sign problem or complex action problem.

Title and authors: Kai: So we're looking at this paper called Can multi-level Monte-Carlo address the sign problem in a finite-density Bose gas? It addresses that big hurdle where complex actions in field theories cause the statistical noise to grow exponentially.

Mira: Yeah, it tackles that sign problem by proposing a specific way to handle it using a multi-level Monte Carlo approach combined with reweighting. It tests this on a simplified model of a relativistic Bose gas at finite chemical potential.

Lev: So what's the core idea here? Is it just another reweighting trick, or is there something fundamentally different about this multi-level method that helps reduce that noise?

Kai: The key is factorizing the action and the observable into parts that can be sampled independently, which means we get a huge statistical enhancement. They rearrange the path integral to look like something involving level zero sampling and then level one sampling.

Mira: That level zero part gives you snapshots of the full field distribution according to the full action, but it’s still subject to that uncontrolled systematic effect from neglecting the phase.

Lev: And then they use level one sampling by freezing parts of the field and sampling those domains independently, which is supposed to scale the variance better than just using level zero estimates.

Kai: They tested different ways to slice up the space-time lattice, and they found that a checkerboard decomposition was actually the best for getting a lot of independent domains.

Mira: That’s interesting because they also showed that for larger lattices, combining different types of diagonal slicing patterns worked best for keeping the number of frozen sites low while maximizing domain count.

Lev: From an engineering standpoint, if you were building this on real hardware, you'd have to worry about the overhead of those freezing steps and how they affect the overall computational cost before you even get to that statistical advantage.

Kai: The results showed that for smaller lattices, like a six by six by six by six lattice, the multi-level method didn't give as much of a benefit, favoring simpler slicing schemes.

Mira: But on larger volumes, the combination of four orthogonal diagonal slicing patterns provided the best behavior across all tested volumes up to where they hit the phase transition in this simplified model.

Lev: So what does this mean for the actual physics we care about? Does it solve the sign problem generally, or is it just a strong tool for specific regimes?

Kai: It seems like a powerful tool because they showed that reweighting alone gets unreliable at higher chemical potentials, and this multi-level approach shows promise at high densities and low temperatures.

Mira: The paper demonstrates the promise of this method by showing that the reweighted results agree precisely with the expected curve at low chemical potential, but then statistical noise grows significantly above a chemical potential of zero point three.

Lev: That noise growth is what we see in simulations; it means that as you push things to higher densities, your computational budget just disappears into unreliability unless you have a way to manage that variance.

Kai: The results show that the phase-quenched results sit significantly above the expected density value, but the reweighted results match it perfectly at low chemical potential, which is a good sign for accuracy there.

Mira: The paper shows that when they treat the phase as part of the observable in this reweighting approach, you get a much more reliable determination of that density we call mu over V divided by d log Z over mu.

Lev: It's helpful to see how much statistical uncertainty drops when you move from level zero estimates to these improved level one estimates, even if it’s not perfect yet.

Kai: Overall, the conclusion is that this multi-level Monte Carlo approach allows for practical simulations using reweighting at significantly higher chemical potentials than reweighting alone.

Mira: It identifies the best identified approach for larger volumes as that combination of four orthogonal slicings S0one thousand one hundred eleven(D) tensor S0one-one-one(D) tensor S0-one two.

Lev: That specific combination is what you'd want to focus on if you were trying to make this runnable on a real machine for high density physics.

Kai: They also suggest that future work could look at thicker frozen regions between blocks to try and suppress those problematic pion contributions that can cause noise.

Mira: And the paper suggests combining this multi-level sampling with sign reweighting, which would aim to make an approach free from both the sign problem and the overlap problem.

Lev: That would be a huge step if they could actually execute that combination on real hardware without the overhead killing everything.

Kai: So, to wrap up this discussion on Can multi-level Monte-Carlo address the sign problem in a finite-density Bose gas? This paper shows that this method offers an exponential enhancement of statistics to counteract the sign problem when reweighting is used at high densities and low temperatures.

Mira: It’s promising because it gives us a way to get results that agree with expected behavior up to a chemical potential of zero point three in this simplified model.

Lev: It highlights how crucial it is to have these clever sampling techniques when you're trying to push the limits of what numerical methods can handle for things like nuclear matter densities.

Kai: We’ll keep an eye on those future directions, especially combining this multi-level sampling with sign reweighting.

The paper's summary: Kai: So, this paper is looking at how to tackle that sign problem in field theories using a specific technique called multi-level Monte Carlo reweighting for a finite density Bose gas.

Mira: It’s trying to get around the exponential noise growth you see when you have an imaginary part in the action, which messes up standard path integral sampling.

Lev: The core idea is they factorize that action and observable into pieces that can be sampled separately, so they get a statistical boost.

Kai: They set up this level zero sampling to get snapshots of the full field distribution from the whole action, but that’s still subject to some systematic error because you're ignoring the phase information.

Mira: Then they introduce level one sampling by freezing parts of the field and sampling those domains independently, which is supposed to reduce that variance significantly.

Lev: That scaling improvement is what matters for running on real hardware, because if the variance scales better than you thought, it might actually be feasible for high densities.

Kai: They tested different ways to slice up the space-time lattice, and they found a checkerboard decomposition was actually the best way to get a high number of independent domains.

Mira: And for bigger lattices, they discovered that combining four specific orthogonal diagonal slicing patterns gave them the best results across all volumes tested up to where the system starts showing phase transitions in this model.

Lev: That combination is what you’d want to focus on if you were trying to make this runnable on a real machine for high density physics because it minimizes those frozen sites while maximizing domain count.

Kai: The results show that for smaller lattices, like a six by six by six by six lattice, the multi-level method didn't give as much of a benefit, favoring simpler slicing schemes.

Mira: But on larger volumes, the combination of those four orthogonal diagonal patterns gave them the best behavior for everything up to that phase transition point in this simplified model.

Lev: So what does this mean for us? It suggests that reweighting alone gets unreliable at higher chemical potentials, but this multi-level approach shows promise at high densities and low temperatures.

Kai: Specifically, they’ve shown that when you treat the phase as part of the observable in this reweighting setup, you get a much more reliable determination of the density we call mu over volume divided by d log Z over mu.

Mira: It means they can perform precise reweighted calculations at chemical potentials significantly higher than what standard Monte Carlo sampling can handle alone.

Lev: That’s a big deal for QCD simulations, because it opens up a route to studying nuclear matter densities that were previously unreachable due to the sign problem.

Kai: They also suggest that future work should look at thicker frozen regions between blocks to try and suppress those problematic pion contributions that can cause noise.

Mira: And they propose combining this multi-level sampling with sign reweighting, aiming for an approach that avoids both the sign problem and the overlap problem entirely.

Lev: If they can actually execute that combination on real hardware without the overhead killing everything, it would be a huge step for simulating QCD at nuclear matter densities or beyond.

Kai: We’ll keep an eye on those future directions, especially combining this multi-level sampling with sign reweighting.

The paper's improvements: Tom: So, this paper is looking at ways to improve their multi-level Monte Carlo reweighting approach to tackle that sign problem in a finite density Bose gas for more reliable results.

Kai: They suggest they should look at making those frozen regions between the blocks thicker, which is supposed to help suppress those problematic pion contributions that create statistical noise.

Mira: That makes sense from a theory standpoint because those pions are what cause the biggest headaches in the calculation, so reducing their impact should clean up the measurement record.

Lev: From an engineering side, if you make those regions thicker, you increase the computational cost per step but hopefully reduce the overall variance of that estimate.

Kai: It’s a trade-off, right? You spend more time computing each snapshot to get a cleaner final result for your density measurement.

Mira: Exactly. The goal is to find that sweet spot where you get enough statistical cleanliness without making the simulation take an unreasonably long time to run on real hardware.

Lev: If they can manage that, it could make their method viable for studying nuclear matter densities that were previously out of reach because of the sign problem’s exponential growth.

Kai: They also propose a way forward by combining this multi-level sampling with sign reweighting itself, which would aim to eliminate both the sign problem and the overlap problem at once.

Mira: That sounds like they're trying to build a method that handles both of those major issues in one go, which is a very ambitious goal for field theory simulations.

Lev: If they can pull off combining them, it would be a massive step toward simulating Quantum Chromodynamics at the densities we expect inside neutron stars.

Kai: It’s interesting because they’re trying to find that middle ground between just fixing the noise and actually solving the underlying sign problem issue.

Mira: So, they're not just stopping at fixing the variance in one way, but proposing a deeper structural change to how you sample the path integral itself.

Lev: That level of complexity suggests they’re aiming for something that could actually provide a more robust way to determine that density mu over volume divided by d log Z over mu.

Conclusion: Kai: So, to wrap up, this paper on "Can multi-level Monte Carlo address the sign problem in a finite-density Bose gas?" shows that using reweighting with multi-level sampling can give us results at much higher chemical potentials than before.

Mira: It’s promising because they get results that match expected behavior up to a chemical potential of zero point three in this simplified model.

Lev: That means we can get a more reliable determination of that density mu over volume divided by d log Z over mu, which is crucial for pushing QCD simulations.

Kai: It highlights how much statistical uncertainty drops when you move from level zero estimates to these improved level one estimates, even if it’s not perfect yet.

Mira: They also point out that the best identified approach for larger volumes was that combination of four orthogonal slicing patterns, which really minimizes those frozen sites while maximizing domain count.

Lev: That specific combination is what you'd want to focus on if you were trying to make this runnable on a real machine for high density physics because it keeps the overhead down.

Kai: They also suggest that future work should look at making those frozen regions between blocks thicker to try and suppress those problematic pion contributions that cause noise.

Mira: And they propose combining this multi-level sampling with sign reweighting itself, aiming for an approach that avoids both the sign problem and the overlap problem at once.

Lev: If they can actually pull off combining them, it would be a massive step toward simulating QCD at nuclear matter densities or beyond.

Kai: So, overall this paper demonstrates that this multi-level Monte Carlo approach offers an exponential enhancement of statistics to counteract the sign problem when reweighting is used at high densities and low temperatures.

Mira: It’s promising because it gives us a way to get results that agree with expected behavior up to a chemical potential of zero point three in this simplified model.

Lev: It highlights how crucial it is to have these clever sampling techniques when you're trying to push the limits of what numerical methods can handle for things like nuclear matter densities.

Rajatha Ranpatige, Finn M. Stokes

ARC Special Research Centre for the Subatomic Structure of Matter (CSSM) · Adelaide University

hep-lat, nucl-th

Submitted: 2026-11-03

Updated: 2026-11-03

Comments: 14 pages, 6 figures

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

Importance score: 82/100

The gist: Numerical simulation of field theories with complex actions is frustrated by the sign problem, which causes an exponential growth in statistical sampling noise.

Key concepts

Sign Problem
This difficulty arises in field theories with complex actions because the imaginary part of the action introduces a complex phase into the statistical weights. This prevents standard Monte Carlo methods from accurately sampling the path integral, leading to exponential noise that makes simulations intractable for many physical systems.
Reweighting Approach
This method attempts to handle the sign problem by treating the complex phase as part of an observable. Instead of sampling directly, one samples a related quantity and then uses a reweighting factor to correct for the missing phase information. However, this introduces statistical uncertainties that can become very large due to cancellations between different phases.
Multi-level Monte Carlo Strategy
This technique solves the noise issue by factoring the path integral into independent components. Level-0 sampling provides a full field snapshot, and subsequent levels sample smaller, independently weighted domains. This factorization allows the variance of the estimate to scale much better than simpler methods.
Checkerboard Decomposition
This is one specific way to decompose the simulation lattice into independent regions. The checkerboard scheme was found to be particularly effective for maximizing the number of independent domains that can be sampled, which is crucial for improving the statistical quality of the multi-level sampling method.

Terminology

Summary

Numerical simulation of field theories with complex actions is frustrated by the sign problem, which causes an exponential growth in statistical sampling noise. p. 0

Where the paper says it

au Abstract. Numerical simulation of field theories with complex actions is frustrated by the sign problem, which causes an exponential growth in statistical sampling noise.

In this work, a multi-level Monte-Carlo approach to reweighting may offer a resolution to the sign problem at high densities and low temperatures. p. 1
Where the paper says it

2 In this work we demonstrate that a multi-level Monte-Carlo approach to reweighting may offer a resolution to the sign problem at high densities and low temperatures. This is demonstrated in a simplified scenario of a relativistic Bose gas that displays the same “Silver Blaze” phenomenon and sign problem.

The Sign Problem in Field Theories

Field theories with a complex action are extremely difficult to simulate because the imaginary part introduces a nontrivial complex phase into the exponential weight. This phase prevents Monte-Carlo importance sampling of the path integral, which is known as the sign problem or complex action problem. p. 0

Where the paper says it

This phase prevents us from treating the weight as a probability for Monte-Carlo importance sampling of the path integral. This is known as the sign problem or complex action problem.

The sign problem has frustrated numerical efforts to understand the physics of large numbers of interacting particles, such as in highly correlated condensed matter systems, atomic nuclei, neutron stars, and quark-gluon plasma. p. 0
Where the paper says it

This is known as the sign problem or complex action problem. It has frustrated numerical efforts to understand the physics of large numbers of interacting particles, such as in highly correlated condensed matter systems, atomic nuclei, neutron stars, and quark-gluon plasma.

Many approaches have been proposed to solve this problem over 40 years, including the complex Langevin approach [1–9], imaginary chemical potential [10–27], reweighting [28–34], Taylor series expansion [35–49], and Lefschetz thimble methods [50–54]. p. 0
Where the paper says it

Much effort has been invested to search for solutions to the sign problem that would unlock these important areas of study for over 40 years. Many approaches have been proposed, for example the complex Langevin approach [1–9], imaginary chemical potential [10–27], reweighting [28–34], Taylor series expansion [35–49], and Lefschetz thimble methods [50–54].

It has been proven that this problem is NP-hard, meaning a general solution is unlikely to exist. p. 0
Where the paper says it

Many approaches have been proposed, for example the complex Langevin approach [1–9], imaginary chemical potential [10–27], reweighting [28–34], Taylor series expansion [35–49], and Lefschetz thimble methods [50–54]. However, it has been proven that this problem is NP-hard [55], and so a general solution is unlikely to exist.

The Simplified Model and Initial Results

The approach is tested on a simpler model of a relativistic Bose gas that displays the same sign problem at finite chemical potential. The action is split into real and imaginary parts, resulting in the path integral being written as ⟨O⟩ = 1/Z Z Dϕ O[ϕ]e−SR[ϕ] e−iSI[ϕ]. p. 1

Where the paper says it

Since the action S = P x L(x) is complex, the e −S factor is not purely real and thus can’t be interpreted as a probability. To resolve this, we start by splitting the resulting action into the real and imaginary parts S = SR + iSI , where SR = X x h (2d + m2 )ϕ ∗ xϕx − X 3 i=1 ϕ ∗ xϕx+ˆi + ϕ ∗ x+ˆi ϕx (3) − cosh(µ) ϕ ∗ xϕx+ˆ4 + ϕ ∗ x+ˆ4 ϕx + λ(ϕ ∗ xϕx) 2 i , and (4) SI = −i X x sinh(µ) ϕ ∗ xϕx+ˆ4 − ϕ ∗ x+ˆ4 ϕx .

The phase-quenched expectation value is numerically straightforward but introduces an uncontrolled systematic effect due to the neglected phase. p. 1
Where the paper says it

One approach is to neglect it and obtain what is known as the phase quenched expectation value ⟨O⟩P Q = 1 ZP Q Z Dϕ O[ϕ]e −SR[ϕ] , where ZP Q = R Dϕ e−SR[ϕ] . This approach is numerically straightforward, but introduces an uncontrolled systematic effect due to the neglected phase.

The reweighting approach treats the phase as part of the observable, where ⟨O⟩ = ⟨Oe−iSI[ϕ]⟩PQ ⟨e−iSI[ϕ]⟩PQ. p. 2
Where the paper says it

Another approach is to treat the phase as part of the observable [28], in an approach known as reweighting. Then the expectation value becomes ⟨O⟩ = ⟨Oe −iSI[ϕ] ⟩P Q ⟨e−iSI[ϕ]⟩P Q , where the ⟨e −iSI[ϕ] ⟩P Q accounts for the difference between the phase-quenched and full partition functions.

With enough Monte-Carlo samples, this approach will give the correct result but suffers from statistical uncertainties that blow up due to cancellations between different phases. p. 2
Where the paper says it

With enough Monte-Carlo samples, this approach will give the correct result. However, due to cancellations from the different phases of different ϕ configurations, the statistical uncertainties from the Monte-Carlo sampling blow up and the result becomes extremely noisy, as can be seen in Figure 1.

The Multi-level Monte-Carlo Strategy

To resolve the statistical noise problem, a multi-level algorithm is employed, which relies on finding a factorisation of the action and observable into independently sampleable components. The path integral is rearranged as ⟨O⟩ = Z Dϕomega Xj D Y(1) (9). p. 0

Where the paper says it

Introduction Field theories with a complex action are extremely difficult to simulate, as the imaginary part introduces a nontrivial complex phase into the exponential weight. This phase prevents us from treating the weight as a probability for Monte-Carlo importance sampling of the path integral.

Level-0 sampling provides snapshots ϕ(0),kΓ of the full field distributed according to the full action. p. 3
Where the paper says it

(9) We can evaluate this nested path integral through a multi-level Monte-Carlo integration approach. Starting with a conventional Monte Carlo sampling which we call level 0, we obtain snapshots ϕ (0),k Γ of the full field, distributed according to the full action p (0)[ϕΓ] = exp(−S[ϕΓ]) = D Y (1) i=1 exp(−Si [ϕΓi , ϕΩ]).

Level-1 sampling involves freezing the field within a boundary region and sampling individual domains independently, with fields weighted by p(1)[ϕ(1),k(0),k(1)iΓ]. p. 3
Where the paper says it

(11) The statistical variance of such an estimate scales as N (0)−1 , but a multi-level approach can do much better. If we consider a single level-0 sample, ϕ (0),k(0) Γ , we can perform a level-1 sampling in the following way: We “freeze” the field within the boundary region ϕ (0),k(0) Ω , and sample the individual Γi domains

This allows the variance of the estimate to potentially scale as N(1)-D(1), a significant improvement over level-0 estimates. p. 4
Where the paper says it

(12) Then, taking N(1) of these level-1 samples for each domain and for each k (0) gives a set of estimates ⟨O⟩(1),k(0) = X j D Y (1) i=1 1 N(1) N X (1) k (1) i =1 Oji[ϕ (0),k(0),k(1) i Γi , ϕ(0),k(0) Ω ] , (13) which approximate the desired expectation value. Because the domains are sampled independently, and the product of their contributions is taken, we effectively consider all possible combinations of the individual samples, so the variance of the estimate has the potential to scale as N (1)−D(1) , a significant improvement over the level-1 estimate.

Decomposition Schemes and Optimal Strategy

Various decompositions are tested, including axis-aligned parallel slices in time, space, or diagonal slices. The checkerboard decomposition Cδ is identified as the best option for achieving the largest number of domains. p. 6

Where the paper says it

When combining slicings, more diagonal decompositions become advantageous due to the way they wrap around the periodic boundaries and intersect with one another. To achieve the largest number of domains the checkerboard is by far the best option.

For a 4x4x4x4 lattice, the checkerboard decomposition achieves 128 independent domains with half of the lattice sites frozen. p. 6
Where the paper says it

7 Figure 3: A checkerboard decomposition C 0 = S 0 v (2), illustrated on a two-dimensional slice through a 4 × 4 × 4 × 4 lattice. Black squares indicate frozen sites, and white squares the non-frozen sites, forming eight independent domains on each slice, for a total of 128 domains.

Testing on larger lattices showed that combinations of orthogonal diagonal slicing patterns are optimal for maximizing domain count and minimizing frozen sites. p. 7
Where the paper says it

On the middle (8 × 8 × 8 × 8) lattice, there is enough volume available that more complicated decompositions become optimal. The best approach we identified was a combination of four orthogonal diagonal slicing patterns: S 0 [1,1,1,1](2) ⊗ S 0 [1,1,−1,−1](2) ⊗ S 0 [1,−1,1,−1](2) ⊗ S 0 [1,−1,−1,1](2), with N(0) = 100, 000 and N(1) = 10.

Performance Across Lattice Volumes

The results show that different schemes perform better for different lattice volumes, but the trends are consistent across the simulations. On the smallest (6x6x6x6) lattice, the benefits of multi-level were limited, favoring minimal slices. p. 6

Where the paper says it

On the smallest (6 × 6 × 6 × 6) lattice, the potential benefits of multi-level were limited, and the optimal approach we found was the simplest decomposition into two regions. Because the lattice volume is small, the surface-to-volume ratio is large, and so any slices freeze a notable fraction of the sites, so decompositions with a minimal number of slices are best.

For larger volumes, the combination of four orthogonal diagonal slicing patterns provided the best-behaved values for all considered volumes up to the phase transition. p. 8
Where the paper says it

On the largest (16×16×16×16) lattice, the best approach we identified was again a combination of four orthogonal diagonal slicing patterns, with twice as many slices as on the smaller lattice: S 0 [1,1,1,1](4) ⊗ S 0 [1,1,−1,−1](4) ⊗ S 0 [1,−1,1,−1](4) ⊗ S 0 [1,−1,−1,1](4), with N(0) = 1, 000 and N(1) = 1, 000. 6 shows that this approach has the best-behaved values of all the volumes considered so far, all the way out to the phase transition.

This approach achieved results consistent with previous complex Langevin calculations. p. 9
Where the paper says it

results all the way out to the phase transition, for the same or better compute cost. Our approach achieved results consistent with previous complex Langevin calculations [60,68].

Conclusion and Future Directions

A multi-level Monte-Carlo approach allows practical simulations using reweighting at significantly higher chemical potentials than reweighting alone, providing a huge statistical advantage. The best identified approach for larger volumes was the combination of orthogonal slicings S01,1,1,1 ⊗ S01,−1,−1 ⊗ S0[−1, 2]. Future work will investigate thicker frozen regions between blocks to suppress problematic pion contributions. p. 9

Where the paper says it

The frozen regions between blocks need to be much thicker than the single lattice sites used here. Future work will investigate these decomposition approaches with thicker frozen regions between the domains, capable of suppressing the problematic pion contributions.

It is also valuable to explore combining multi-level sampling with sign reweighting to produce an approach free from both the sign problem and the overlap problem. p. 9
Where the paper says it

Sign reweighting [34, 69, 70] is a proposed solution to this problem that eliminates the overlap problem and only leaves the sign problem, that this approach seems to resolve. It will be valuable to explore how these two techniques can be combined to produce an approach free from both problems.

This offers a promising pathway to simulating QCD at nuclear matter densities or beyond. p. 9
Where the paper says it

It will also be valuable to explore how the situation changes if multi-level sampling with more than two levels is used, as described at the end of Sec. In conclusion, this is a promising result, and a sign that this approach may offer a pathway to simulating QCD at nuclear matter densities or beyond, providing key theoretical insight into nuclear and neutron star physics.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved ability to perform first-principles numerical calculations for QCD at finite baryon chemical potential by applying a reweighting procedure coupled with a multi-level Monte Carlo sampling algorithm, which offers an exponential enhancement of statistics to counteract the sign problem.

  2. The improved AI system can accurately determine physical observables up to the phase transition in simplified models like the relativistic Bose gas, significantly increasing the range over which precise calculations are possible.

  3. The system can resolve statistical noise problems by factorizing actions and observables into components that can be Monte Carlo sampled independently, allowing for a dramatic reduction in uncertainty for a given computational cost.

  4. The AI can choose optimal lattice decomposition schemes based on specific lattice geometries to maximize the number of independent domains while minimizing frozen sites, leading to improved precision across different volumes.

  5. The system can perform precise reweighted calculations up to chemical potentials significantly higher than those attainable by conventional Monte Carlo sampling, providing a more reliable determination of the density n = (1/V)d lnZ/dµ.

Abstract

Numerical simulation of field theories with complex actions is frustrated by the sign problem, which causes an exponential growth in statistical sampling noise. A key example of such a theory is QCD at finite baryon chemical potential, which has important applications to the structure of atomic nuclei and neutron stars. Until now, the sign problem has precluded the possibility of first-principles numerical calculations that could provide key insight into these important physical systems. In this work, we propose an approach to avoiding this limitation by applying a reweighting procedure coupled with a multi-level Monte-Carlo sampling algorithm that offers an exponential enhancement of statistics to counteract the sign problem. We test the approach on a simpler model that displays the same sign problem at finite chemical potential. Our results demonstrate the promise of the proposed approach, significantly increasing the range over which precise calculations are possible, and providing improvements all the way up to the phase transition.

Sources

Related papers