Tensor Decomposition for Energy-Momentum Correlation Functions

This paper establishes the general functional form of the Euclidean energy-momentum tensor two-point function at zero and finite temperatures by decomposing it into fundamental tensorial structures, deriving differential relations via energy-momentum conservation, and expressing the full correlator in terms of a reduced set of spectral functions to facilitate more efficient lattice investigations.

The Gist

Imagine you are trying to understand the "heartbeat" of a hot, chaotic soup of particles known as the quark-gluon plasma (the stuff that existed just after the Big Bang). Physicists study this by looking at how energy and momentum move around inside this soup. They use a mathematical tool called a correlation function, which is like a map showing how a "push" at one spot affects another spot.

However, this map is incredibly complicated. It's not just a simple line or a circle; it's a 4D shape (a rank-4 tensor) that changes depending on the direction you look, the distance between points, and the temperature. Trying to analyze this raw data is like trying to read a book written in a language with 100 different letters, most of which are just noise or repetitions.

This paper, by Guy D. Moore and Jonas Winter, is essentially a translation guide and a compression algorithm for this complex data. Here is how they break it down:

1. The Problem: Too Much Noise, Too Many Directions

Imagine you are in a dark room with a single lightbulb. If you look at the light from the North, it looks different than if you look from the East. The paper explains that the "energy-momentum" map behaves similarly. It has a strong "directional bias."

• The Old Way: Scientists used to take all the data, average it out, and look at the result. But this is like averaging the sound of a violin, a drum, and a siren together; you lose the unique character of each instrument.
• The New Way: The authors say, "Let's separate the instruments first." They want to break the complex map into its fundamental "building blocks" (tensor structures) so we can study the pure signal without the noise.

2. The Solution: Breaking the Map into Lego Bricks

The authors developed a method to decompose the complex 4D map into a set of simpler, fundamental "Lego bricks" (mathematical projectors).

• Zero Temperature (Vacuum): In a cold, empty space, the map can be broken down into just five types of bricks.
• Hot Temperature (The Soup): When the soup is hot, the rules change slightly. If you average the data over time, you get ten types of bricks. If you look at specific moments in time, you get fourteen types.

Think of it like a prism. White light (the raw data) looks messy, but when you pass it through a prism (the authors' decomposition), it splits into a clean rainbow of distinct colors (the fundamental components).

3. The Rules of the Game: Conservation Laws

The universe has strict rules: Energy and momentum cannot just disappear; they must be conserved. In the language of this paper, this is called Energy-Momentum Conservation (EMC).

• The Analogy: Imagine you have a jigsaw puzzle. You might think you have 100 unique pieces, but the picture on the box (the conservation law) tells you that 50 of those pieces are actually just copies of the other 50, or that they must fit together in a specific way.
• The Result: The authors used these rules to show that even though the map looks like it has many independent parts, the laws of physics force them to be connected.
  • In a vacuum, those 5 bricks are actually linked so tightly that only 2 are truly independent.
  • In the hot soup, the 10 or 14 bricks are linked down to a much smaller set of spectral functions (the "true" independent variables).

4. Why This Matters: Finding the Signal in the Noise

In computer simulations (lattice QCD), the data gets very "noisy" the further apart you look. It's like trying to hear a whisper in a stadium; the further you are from the speaker, the harder it is to hear.

• The Old Problem: When scientists tried to fit the data to understand the "viscosity" (how sticky the soup is), they included all the noisy, far-away data, which ruined their precision.
• The New Advantage: By using the authors' decomposition, scientists can now fit the "tail" of the data (the far-away, noisy part) using the spectral functions. Because these functions are mathematically linked and simpler, you can fit the whole complex map using just a few parameters.
• The Benefit: This allows for much more precise calculations of how the quark-gluon plasma flows, without being thrown off by statistical noise.

Summary

The paper doesn't invent new physics or discover a new particle. Instead, it provides a better way to organize the data we already have.

• It takes a messy, 100-component puzzle.
• It sorts the pieces into distinct categories based on symmetry.
• It uses the laws of conservation to show which pieces are actually the same.
• It reduces the problem to a small set of "spectral functions" that act as the true DNA of the system.

This allows physicists to extract the "viscosity" of the early universe with much higher precision, turning a blurry, noisy picture into a sharp, clear image.

Technical Summary

Technical Summary: Tensor Decomposition for Energy-Momentum Correlation Functions

Problem Statement
The paper addresses the challenge of efficiently extracting hydrodynamic transport coefficients, such as shear and bulk viscosity, from lattice Quantum Chromodynamics (QCD) calculations. These coefficients are derived from the Euclidean energy-momentum tensor (EMT) two-point correlation function, $G_{\mu\nu\rho\sigma}(\tau, \mathbf{r}) = \langle T_{\mu\nu}(\tau, \mathbf{r})T_{\rho\sigma}(0, 0) \rangle$.

Current lattice approaches often involve averaging the correlation function over spatial separations at fixed Euclidean time. However, this method suffers from a poor signal-to-noise ratio because the correlation function decays rapidly with distance while fluctuations do not, leading to the integration of large volumes of pure noise. A more precise approach involves fitting the long-distance tail of the full space- and time-dependent correlation function. The obstacle to this method is that $G_{\mu\nu\rho\sigma}$ is a rank-4 tensor with complex angular dependence. Without a rigorous understanding of its tensorial structure, fitting the tail is inefficient and prone to errors. The authors note that the correlation function is not a single scalar function but a linear combination of underlying component functions with distinct angular dependencies.

Methodology
The authors perform a systematic tensor decomposition of the EMT two-point function based on the remaining rotational symmetries in three distinct scenarios:

1. Zero Temperature (Vacuum): Full $SO(D-1)$ rotational symmetry around the separation axis.
2. Finite Temperature (Time-Averaged): Symmetry reduced to $SO(D-2)$ due to the periodic temporal direction, with averaging over the temporal separation $\tau$.
3. Finite Temperature (General Time): No averaging over $\tau$, retaining full dependence on both spatial separation $\mathbf{r}$ and Euclidean time $\tau$.

The methodology proceeds in three main stages:

1. Tensor Decomposition: The authors construct a minimal basis of orthonormal projectors ($P^X_{\mu\nu\rho\sigma}$) that respect the relevant symmetries (rotational invariance, index symmetries of the EMT, and time-reflection properties). They decompose the full correlator into a sum of these projectors weighted by scalar component functions $G_X(\tau, r^2)$:
   $$G_{\mu\nu\rho\sigma} = \sum_X G_X(\tau, r^2) P^X_{\mu\nu\rho\sigma}(\hat{r})$$
   This reduces the 100 components of the rank-4 tensor to a small set of independent scalar functions (5 in vacuum, 10 for time-averaged finite temperature, and 14 for general finite temperature).

2. Energy-Momentum Conservation (EMC) Constraints: The authors derive the implications of the conservation law $\partial_\mu T^{\mu\nu} = 0$ in coordinate space. Unlike momentum space, where conservation imposes algebraic constraints, in coordinate space it yields differential relations (ordinary differential equations for time-averaged cases, partial differential equations for general cases) linking the component functions. These relations significantly reduce the number of truly independent functions.

3. Spectral Decomposition: For the vacuum and time-averaged finite temperature cases, the authors utilize the differential constraints to express the component functions in terms of a smaller set of spectral functions ($\rho_S(m)$). They derive explicit integral representations (Lehmann-type) where the component functions are integrals over mass $m$ of spectral functions weighted by specific kernel functions $f_n(m, r)$ and coefficients $c_{S,X}^i$.

Key Results

• Vacuum ($T=0$): The correlation function is decomposed into 5 component functions ($G_{TT}, G_{RT}, G_{ss}, G_{\ell\ell}, G_{s\ell}$). EMC imposes 3 differential constraints, reducing the system to 2 independent degrees of freedom. The authors provide a spectral representation involving two spectral functions ($\rho_{TT}, \rho_{tt}$) corresponding to the tensor and trace channels.
• Finite Temperature (Time-Averaged): The decomposition yields 10 component functions. EMC provides 5 differential constraints, leaving 5 independent degrees of freedom. The spectral representation involves 5 spectral functions ($\rho_{TT}, \rho_{UT}, \rho_{tt}, \rho_{uu}, \rho_{tu}$).
• Finite Temperature (General Time): The decomposition yields 14 component functions. EMC results in 10 coupled partial differential equations. While the authors note that a simple spectral decomposition is not immediately clear for the general case due to the PDE nature, they show that Fourier transforming to Matsubara frequencies reduces the problem to ordinary differential equations. For non-zero Matsubara modes, this leaves 4 independent degrees of freedom.
• Zero Separation Limits: The paper analyzes the behavior of the projectors and component functions as $r \to 0$. It establishes that several projectors become degenerate or vanish, leading to specific relations between component functions at zero separation (e.g., $G_{TT}(0) = G_{RT}(0) = G_{\ell\ell}(0)$).
• Application to Transport Coefficients: The authors explicitly map standard transport coefficient definitions (shear viscosity $\eta$, bulk viscosity $\zeta$, and second-order thermodynamic coefficients) to integrals over the decomposed component functions. For instance, the shear viscosity is shown to be a weighted integral of $G_{TT}, G_{RT},$ and $G_{\ell\ell}$.

Significance and Claims
The paper claims to provide the "general functional form" of the EMT two-point function in Euclidean space, establishing a rigorous framework for lattice QCD analyses. The primary significance lies in enabling "more efficient future lattice investigations."

By reducing the raw lattice data to a small set of component functions dependent only on $r^2$ (and $\tau$), the authors argue that:

1. Data Reduction: The massive amount of tensor data can be losslessly compressed into a handful of scalar functions.
2. Improved Fitting: The angular dependence is analytically known via the projectors. This allows for simultaneous fitting of all component functions using a small set of spectral parameters (representing the lightest states), rather than fitting noisy, direction-dependent data points individually.
3. Constraint Testing: The differential relations derived from EMC and the positivity constraints (reflection positivity) provide rigorous checks for lattice artifacts, gradient flow effects, and statistical uncertainties.

The authors modestly state that the next step is the actual application of these tools to lattice data, noting that work in this direction is already ongoing. They do not present new lattice results or experimental predictions, but rather the theoretical infrastructure required to extract transport coefficients with higher precision from existing and future lattice simulations.