⚛️ high-energy theory

Additional constraints for the tensor bootstrap

This paper introduces two new positivity constraints based on "open bubbles" and "color matrices" to derive sharp bounds on unitary tensor integrals at finite $N$ and investigate deviations from Gaussian universality.

Original authors: Samuel Laliberte, Reiko Toriumi

Published 2026-06-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Samuel Laliberte, Reiko Toriumi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to understand a massive, complex machine made of billions of tiny, interconnected gears. In the world of theoretical physics, these "gears" are called tensors, and the machine is a tensor model. Scientists use these models to study everything from the fabric of space-time to how particles interact.

For a long time, physicists have been great at understanding this machine when it has an infinite number of gears (a concept called the "large N limit"). In this infinite world, the machine behaves in a very predictable, boring way: it acts like a simple, smooth wave. This is called Gaussian universality. It's like saying that no matter how you mix a giant pot of soup, if you have enough ingredients, it will always taste exactly the same.

However, the real world doesn't have infinite gears; it has a finite number. When the number of gears is small or medium-sized, the machine behaves chaotically. The simple rules break down, and the soup might taste different depending on exactly how many carrots you put in. Until now, figuring out exactly how this finite machine behaves has been incredibly difficult, often requiring brute-force computer simulations that can get stuck or take forever.

The New Tool: The "Tensor Bootstrap"

The authors of this paper, Samuel Laliberte and Reiko Toriumi, have developed a new mathematical "flashlight" to shine on this finite machine. They call it the Tensor Bootstrap.

Think of the bootstrap not as pulling yourself up by your bootstraps, but as using a set of strict rules to narrow down the possibilities. Imagine you are trying to guess the weight of a mystery box.

You know it can't be negative.
You know it can't be heavier than a car.
You know it must fit inside a specific room.

By combining these rules, you can get a very precise estimate of the weight without ever opening the box. The authors use similar "positivity constraints" (rules that say certain things must be positive or zero) to squeeze the possible answers for the tensor machine until only the correct ones remain.

Two New Ways to Look at the Machine

To make their flashlight brighter, the authors invented two new ways to look at the machine's gears:

Open Bubbles: Imagine a "bubble" is a complete, closed loop of gears that represents a specific pattern in the machine. An "open bubble" is what happens if you carefully cut one gear out of that loop. The remaining piece is still a valid object, but now it has an "open end" where the gear used to be. By studying these open ends, the authors can see how the rest of the machine reacts.
Color Matrices: In these models, the connections between gears have different "colors" (like red, green, and blue wires). A "color matrix" is what you get if you snip a single colored wire. This creates a flat, two-dimensional map (a matrix) that shows how the machine behaves specifically along that color.

By turning these "open bubbles" and "color matrices" into giant grids of numbers (called Gram matrices) and demanding that these grids follow strict mathematical rules (they must be "positive semi-definite," which is a fancy way of saying they can't contain impossible negative values), the authors can trap the true behavior of the system.

What They Found

The authors tested their method on two specific types of machines:

The Three-Pillow Machine: A system where three different types of interactions are treated equally.
The One-Pillow Machine: A system where only one type of interaction is active, breaking the symmetry.

The Results:

Sharp Bounds: They were able to draw very tight "fences" around the possible values of the machine's behavior. Before this, the fences were wide and vague; now, they are narrow and precise.
The Finite vs. Infinite Gap: They showed that for small numbers of gears (low N), the machine behaves very differently from the infinite version. It doesn't follow the simple "Gaussian" rules.
Convergence: As they increased the number of gears (N) in their calculations, they watched the machine's behavior slowly "melt" into the simple, predictable infinite behavior. This confirmed that their method works and correctly bridges the gap between the chaotic finite world and the smooth infinite world.
Symmetry Breaking: In the "One-Pillow" machine, they proved that at finite sizes, the machine remembers that one interaction is special and the others are not. But as the machine gets huge, it forgets this difference and treats them all the same again.

Why This Matters

This paper doesn't claim to cure diseases or build new engines directly. Instead, it provides a new mathematical toolkit. It allows physicists to study complex systems at finite sizes with much higher precision than before, without needing to run endless, expensive computer simulations. It's like upgrading from a blurry map to a high-definition GPS for navigating the complex landscape of quantum physics.

In short, the authors found a clever way to use logic and strict rules to pin down the behavior of complex systems when they are small, showing us exactly how they differ from the idealized, infinite versions we usually study.

Technical Summary: Additional Constraints for the Tensor Bootstrap

Problem Statement
While tensor models have been extensively studied in the large- $N$ limit, where observables satisfy "Gaussian universality" (factorizing according to Wick's theorem via melonic dominance), the behavior of these models at finite $N$ remains poorly understood. In the finite $N$ regime, subleading contributions are not suppressed, the melonic description breaks down, and exact solutions to the Schwinger-Dyson equations are generally unknown. Existing numerical methods struggle in the strong coupling regime, necessitating non-perturbative analytical tools to constrain tensor observables directly at finite $N$ . Although the "matrix bootstrap" has successfully constrained matrix models using positivity constraints, extending these techniques to arbitrary unitary tensor integrals has been a recent development.

Methodology
The authors extend the tensor bootstrap program by introducing two new classes of positivity constraints derived from the requirement that the probability measure is positive. These constraints are constructed as Gram matrices that must be positive semi-definite ( $M \succeq 0$ ).

Open Bubbles: The authors define "open bubbles" as tensor-like objects obtained by removing a single vertex (white or black) from a bubble invariant. A Gram matrix is constructed by taking the expectation values of the inner products of these open bubbles.
Color Matrices: The authors define "color matrices" as matrix-like quantities obtained by cutting a single edge (color contraction) from a bubble invariant. A Gram matrix is constructed from the expectation values of these matrices, which transform as $U(N)$ matrices for the specific color of the cut edge.

These positivity constraints are combined with the Schwinger-Dyson equations (consistency conditions derived from the invariance of the partition function under infinitesimal field transformations). The authors employ a semi-definite programming approach (using the SDPA algorithm) to:

Generate Gram matrices from bubble invariants up to a specific truncation level $n$ (number of vertex pairs).
Generate Schwinger-Dyson equations involving open bubbles up to a corresponding vertex count.
Partially solve the Schwinger-Dyson equations to express higher-order bubble expectations in terms of lower-order ones.
Substitute these relations into the Gram matrices.
Optimize the expectation value of a target bubble observable $\langle B \rangle$ by maximizing and minimizing it subject to the constraint that all Gram matrices remain positive semi-definite.

Key Contributions

Novel Positivity Constraints: The paper introduces and formalizes the use of "open bubbles" and "color matrices" to generate additional positive semi-definite conditions for tensor models, distinct from previous matrix-based approaches.
Finite $N$ Bounds: The authors demonstrate that these constraints yield sharp upper and lower bounds on unitary tensor integrals at finite $N$ , specifically for rank-3 tensors with quartic interactions.
Quantitative Probe of Universality: The method provides a quantitative framework to measure deviations from Gaussian universality as a function of $N$ and coupling strength.

Results
The authors applied their method to two specific rank-3 tensor models:

Three-Pillow Interaction: A model with a potential invariant under the permutation of three quartic "pillow" interactions.
- The method produced sharp bounds on the expectation values of the elementary melon and the three quartic pillows for various coupling constants $g$ at $N=1, 5,$ and $50$.
- As $N$ increases, the finite $N$ bounds converge to the exact large $N$ solution derived from Gaussian universality.
- The bounds on the relationship between pillow expectation values and the elementary melon lie above the large $N$ Gaussian universality curve at finite $N$ , converging to it as $N \to \infty$ .
One-Pillow Interaction: A model where the potential explicitly breaks the permutation symmetry between the three possible quartic pillows, leaving only the second and third pillows symmetric.
- At finite $N$ , the bootstrap bounds correctly reflect the symmetry breaking: the expectation values of the two symmetric pillows are identical, while the third (present in the potential) deviates.
- The difference between the expectation values of the symmetric pillows is zero, while the difference involving the asymmetric pillow is non-zero.
- This confirms that the bootstrap can detect the re-emergence of specific permutation symmetries at finite $N$ that are washed out by Gaussian universality in the large $N$ limit.

Significance and Claims
The paper claims that the tensor bootstrap, utilizing these new positivity constraints, is a powerful non-perturbative tool for studying tensor models at finite $N$ . The authors emphasize that they did not assume large $N$ factorization properties; rather, the large $N$ limit (and Gaussian universality) naturally emerges from the finite $N$ constraints as $N$ becomes sufficiently large (e.g., $N=50$ in their examples).

The work highlights that the bootstrap can probe the breakdown of Gaussian universality and the specific structure of finite $N$ symmetries. However, the authors are modest regarding the scalability of their approach. They note that the size of Gram matrices and the number of Schwinger-Dyson equations grow combinatorially with tensor rank and truncation level. While level-3 constraints were sufficient for rank-3 models, extending this to higher-rank tensors or higher-order interactions requires significantly more computational resources, and finding minimal sets of constraints remains a critical challenge for future work. The study was restricted to melonic interactions and observables, leaving the application to non-melonic observables as an open direction.

The New Tool: The "Tensor Bootstrap"

Two New Ways to Look at the Machine

What They Found

Why This Matters

More like this