Tensor-network formulation of QCD in the strong-coupling expansion

This paper presents a tensor-network formulation for the strong-coupling expansion of QCD with staggered quarks at nonzero chemical potential, deriving analytical results on small lattices and introducing an enhanced method for future large-scale computations.

The Gist

Imagine you are trying to solve a massive, 4-dimensional jigsaw puzzle. This puzzle represents the universe of Quantum Chromodynamics (QCD), the theory that explains how quarks (the tiny building blocks of protons and neutrons) stick together.

The problem is that this puzzle is incredibly complex. When you try to simulate it on a computer using standard methods (like Monte Carlo simulations), you hit a "sign problem." It's like trying to balance a scale where some weights are positive and others are negative, but the negative ones keep flipping signs randomly, making the scale impossible to balance. This happens especially when you try to add "chemical potential" (think of it as adding more particles or increasing the density), which is crucial for understanding things like the inside of neutron stars.

This paper presents a new, clever way to solve this puzzle using Tensor Networks. Here is the breakdown of what they did, using simple analogies:

1. The Old Way vs. The New Way

• The Old Way (Monte Carlo): Imagine trying to guess the solution to the puzzle by randomly placing pieces and hoping they fit. When the "sign problem" hits, it's like the puzzle pieces are haunted; they change color randomly, and your computer gets confused and crashes.
• The New Way (Tensor Networks): Instead of guessing, the authors decided to break the puzzle down into a giant, interconnected web of small, local rules. Think of the entire universe as a giant Lego structure. Instead of looking at the whole thing at once, they look at one small brick (a "tensor") and how it connects to its neighbors.

2. The "Strong-Coupling" Expansion

The authors are looking at the universe when the forces between quarks are very strong (like when they are packed tightly together).

• The Analogy: Imagine you are trying to describe a very loud, chaotic party. Instead of trying to describe every single conversation at once, you break the noise down into layers.
  • Layer 1: The basic chatter (the infinite coupling limit).
  • Layer 2, 3, 4: You add more details about the music, the dancing, and the specific interactions.
• The authors are calculating these layers one by one. They are expanding their math like a series of steps, getting more precise with each step (up to the 4th step in this paper).

3. The "Ghost" Problem (Grassmann Variables)

Quarks are "fermions," which means they follow weird rules (the Pauli Exclusion Principle). In math, they are represented by "Grassmann variables."

• The Analogy: Imagine the quarks are ghosts. You can't see them, and if two ghosts try to occupy the same spot, they cancel each other out. This makes them very hard to track in a computer simulation.
• The Trick: The authors found a way to "integrate out" (remove) the original ghosts and replace them with new, invisible helper ghosts (auxiliary variables) that live on the connections (links) between the Lego bricks.
• By doing this, they turned a messy, non-local problem (where ghosts interact across the whole room) into a clean, local problem (where ghosts only interact with their immediate neighbors). This makes the math much easier to handle.

4. The Tensor Network: A Web of Local Rules

Once they simplified the ghosts and the gauge fields (the "glue" holding the quarks together), they rewrote the entire universe as a Tensor Network.

• The Analogy: Imagine a giant spreadsheet where every cell contains a small instruction manual.
  • Each cell (a "tensor") knows its own number (numerical part) and its ghost rules (Grassmann part).
  • To find the answer for the whole universe, you don't need to solve everything at once. You just need to "contract" (multiply and sum) these cells together, passing information from neighbor to neighbor.
• The paper shows how to build these cells for any number of dimensions, colors, and flavors of quarks.

5. The Results: Small Lattices and Big Promises

The authors tested their method on a tiny 2x2 grid (a very small piece of the puzzle).

• The Discovery: They found that how you calculate the final answer matters immensely.
  • Method A: Calculate the total probability first, then take the log. (Like weighing the whole pile of sand, then dividing).
  • Method B: Calculate the log of the probability at every step, then sum them up. (Like weighing each grain of sand individually and adding the logs).
  • Result: Method B was much more accurate and matched the "gold standard" Monte Carlo data much better. It's like realizing that if you want to know the average height of a crowd, it's better to measure everyone individually and average the logs than to try to measure the whole crowd's "average height" directly.

6. The Future: The "Order-Separated" Upgrade

The paper mentions that for bigger puzzles (larger lattices), they need a new tool called OS-GHOTRG (Order-Separated GHOTRG).

• The Analogy: Imagine you are building a skyscraper. The current method builds the whole floor at once, but it gets messy when you get to the 10th floor. The new method (OS-GHOTRG) is like a specialized crane that builds the 1st floor, then the 2nd, then the 3rd, keeping the layers strictly separated so they don't get mixed up. This allows them to calculate the "expansion coefficients" (the specific numbers for each layer) directly, which is essential for getting accurate results on large scales.

Summary

In short, this paper is a blueprint for a new way to simulate the strong nuclear force.

1. They turned a chaotic, ghost-filled problem into a clean, local Lego-like structure.
2. They showed that on small scales, a specific way of crunching the numbers gives much better results.
3. They are building a "super-crane" (OS-GHOTRG) to apply this method to the full, massive universe, potentially solving the "sign problem" that has plagued physicists for decades.

This is a major step forward in understanding how matter behaves under extreme conditions, like inside neutron stars or the early universe, without getting stuck in the mathematical "sign problem" swamp.

Technical Summary

1. Problem Statement

The paper addresses the challenge of studying Quantum Chromodynamics (QCD) at finite baryon density (nonzero chemical potential, $\mu$).

• The Sign Problem: Standard Monte Carlo simulations of lattice QCD fail at $\mu \neq 0$ because the determinant of the Dirac operator becomes complex, leading to a severe sign problem that prevents importance sampling.
• Limitations of Current Methods: Existing approaches (reweighting, Taylor expansion, complex Langevin, etc.) generally fail to reach the physically relevant regime where $\mu/T > 1$.
• Strong-Coupling Expansion: While dual variable formulations have successfully mitigated the sign problem in the infinite-coupling limit ($\beta \to 0$), extending these methods to the strong-coupling expansion (finite but large $\beta \sim 1/g^2$) is difficult. Previous attempts often relied on mean-field approximations or specific vertex models.
• Goal: The authors aim to formulate QCD with staggered quarks in the strong-coupling expansion as a tensor network, allowing for the calculation of the partition function and observables (like the chiral condensate) at nonzero chemical potential without the sign problem, for arbitrary dimensions, colors ($N_c$), and flavors ($N_f$).

2. Methodology

The authors develop a rigorous derivation to rewrite the QCD partition function as the complete trace of a tensor network containing both numerical and Grassmann components. The methodology proceeds through several key stages:

A. Taylor Expansion and Dual Variables

• The partition function is expanded by performing Taylor series expansions of the Boltzmann factors for both the fermion hopping terms and the gauge (plaquette) action.
• This introduces occupation numbers (dual fields): $k, \bar{k}$ for fermion hopping on links and $n$ for plaquettes.
• The gauge links ($U$) and their adjoints ($U^\dagger$) are grouped by link, preparing them for integration.

B. Gauge Link Integration

• The authors integrate out the $SU(N_c)$ gauge fields analytically using known results for Haar measure integrals (specifically the generalized Weingarten calculus).
• Key Simplification: Unlike previous dual formulations where non-local sign factors appeared, the presence of Grassmann variables at both ends of every link allows for a significant simplification. The sum over permutations and groupings of color indices in the gauge integral is reduced by exploiting the antisymmetry of Grassmann variables.
• This results in a "link weight" that depends on the occupation numbers and a reduced set of permutation representatives, eliminating the need for complex decoupling operator indices in the final tensor structure.

C. Grassmann Integration and Auxiliary Variables

• To integrate out the original site-based Grassmann variables ($\psi, \bar{\psi}$), the authors introduce colorless and flavorless auxiliary Grassmann variables ($\chi$) on the links.
• This step decouples the fermionic degrees of freedom, transforming the non-local hopping terms into local contributions.
• The integration yields local constraints (Kronecker deltas) ensuring conservation of fermion number at each site and introduces Levi-Civita tensors to handle the color contraction.

D. Tensor Network Construction

• The partition function is rewritten as a contraction of local tensors $T_x$ defined on lattice sites.
• Tensor Indices: The indices of these tensors correspond to the occupation numbers of the links and plaquettes connected to the site.
• Handling Infinite Sums: Since the plaquette occupation numbers $n$ range from 0 to $\infty$, the bond dimension is initially infinite. The authors truncate the expansion at a maximum order $n_{max}$ in $\beta$.
• Edge Variables: To apply standard Tensor Renormalization Group (TRG) methods, the shared plaquette occupation numbers are converted into "edge variables" on the links, enforcing equality around a plaquette via Kronecker deltas.

E. Expansion Strategies for Observables

The paper investigates two distinct ways to compute observables (specifically the chiral condensate $\Sigma$) from the truncated partition function $Z(n_{max})$:

1. Expansion A: Compute $\Sigma = \frac{1}{V} \frac{\partial \ln Z(n_{max})}{\partial m}$ directly using the truncated polynomial $Z(n_{max})$.
2. Expansion B: Expand $\ln Z$ (the free energy) in powers of $\beta$ first, truncate the series, and then differentiate.

• Finding: Theoretical analysis and numerical tests show that Expansion B is superior. Expansion A suffers from incorrect volume dependence at large $\beta V$, leading to unphysical plateaus, whereas Expansion B correctly captures the thermodynamic behavior.

F. Order-Separated GHOTRG (OS-GHOTRG)

• Standard Higher-Order Tensor Renormalization Group (HOTRG) methods compute the partition function $Z$ as a numerical value for a given $\beta$, making it impossible to extract the individual expansion coefficients $Z_n$ needed for Expansion B on large lattices.
• The authors introduce a new method, Order-Separated GHOTRG (OS-GHOTRG), which explicitly tracks and computes the coefficients $Z_n$ of the $\beta$-expansion during the tensor contraction. This allows for the application of Expansion B on larger lattices.

3. Key Contributions

1. General Formulation: A complete tensor-network formulation for QCD with staggered quarks in the strong-coupling expansion for arbitrary $d$, $N_c$, and $N_f$, including nonzero chemical potential.
2. Sign Problem Elimination: The derivation demonstrates that the sign problem arising from $\mu \neq 0$ is naturally handled within the tensor network framework without numerical instabilities, as the complex phases are absorbed into the local tensor structure.
3. Analytical Derivation: A detailed step-by-step derivation of the gauge integration and Grassmann integration, showing how color indices are disentangled and how auxiliary variables facilitate local tensor construction.
4. Methodological Insight: The identification that expanding the free energy ($\ln Z$) rather than the partition function ($Z$) is crucial for obtaining reliable results in the strong-coupling regime.
5. New Algorithm: The proposal and preliminary demonstration of the OS-GHOTRG method to extract expansion coefficients, overcoming a limitation of standard tensor network methods.

4. Results

• 2x2 Lattice (Analytical): The authors computed the partition function coefficients $Z_n$ up to order $n_{max}=4$ analytically.
  • Comparison with Monte Carlo (MC) data for the chiral condensate showed that Expansion B agrees well with MC data over a wide range of $\beta$ and $\mu$.
  • Expansion A deviated significantly from MC data as $\beta$ increased, confirming the theoretical prediction of volume-dependent errors.
• 8x8 Lattice (Numerical): Using the new OS-GHOTRG method, results were obtained for an $8 \times 8$ lattice.
  • The OS-GHOTRG results using Expansion B matched Monte Carlo data significantly better than the standard GHOTRG method (which effectively uses Expansion A logic due to numerical fitting limitations).
  • Standard GHOTRG results exhibited a plateau at large $\beta$, consistent with the theoretical failure of Expansion A in the large volume limit.

5. Significance

• Bridging the Gap: This work provides a robust framework to study the QCD phase diagram at finite density beyond the infinite-coupling limit, a regime previously inaccessible to dual formulations.
• Overcoming the Sign Problem: It confirms that tensor network methods are a viable alternative to Monte Carlo for QCD at finite density, as they do not suffer from the sign problem.
• Algorithmic Advancement: The development of OS-GHOTRG is a critical step toward applying tensor networks to realistic lattice sizes where extracting expansion coefficients is necessary for physical accuracy.
• Future Outlook: The paper sets the stage for the next paper in the series, which will detail the OS-GHOTRG algorithm and apply it to larger lattices to map out the phase diagram of QCD in the strong-coupling regime.

In summary, the paper successfully translates the strong-coupling expansion of QCD into a tensor network language, solves the technical challenges of color and Grassmann integration, and demonstrates that a specific expansion strategy (Expansion B) combined with a new algorithm (OS-GHOTRG) yields accurate results that agree with Monte Carlo simulations, even in the presence of a chemical potential.