⚛️ phenomenology

The shape of quark flavors

This paper proposes a model where quark Yukawa couplings arise from Gaussian wave function overlaps in extra dimensions, utilizing machine learning to identify field configurations that naturally reproduce the observed quark mass hierarchies and experimental constraints.

Original authors: Shinsuke Kawai, Nobuchika Okada

Published 2026-09-22

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Shinsuke Kawai, Nobuchika Okada

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

In the vast landscape of particle physics, there is a section of the Standard Model that has long puzzled scientists: the flavor sector. This is the part of the theory that explains why matter comes in different types, or "flavors," and why these types have vastly different weights. Imagine a family of siblings where one is a feather and another is a boulder; in the world of subatomic particles, the top quark is roughly 350,000 times heavier than the up quark. The current rules of physics can describe these differences, but they cannot explain why they exist. The theory simply lists the weights as arbitrary numbers, like a phone book with no pattern, leaving physicists without a satisfying reason for the hierarchy. This gap in understanding suggests that there is a deeper layer of reality waiting to be discovered, one that might reveal a hidden order behind the apparent chaos of particle masses.

A new study by researchers Shinsuke Kawai and Nobuchika Okada proposes a fresh way to look at this puzzle. Instead of treating the different weights of quarks as random inputs, they suggest these differences arise from the physical locations of particles in hidden, extra dimensions of space. In their model, the universe has more than the three dimensions of space we experience every day. They imagine that the different types of quarks are not just abstract points but are actually localized, or pinned down, at specific spots within these extra dimensions. The strength of the connection between a quark and the Higgs field—which gives particles their mass—depends on how close these particles are to one another in this hidden space. If two particles are far apart, their connection is weak, resulting in a light particle. If they are close together, the connection is strong, creating a heavy particle.

To test this idea, the researchers built a mathematical model where the quarks are represented by wave functions, which are like fuzzy clouds of probability centered at specific points. They assumed the Higgs field, responsible for generating mass, is spread out evenly across these extra dimensions. The mass of each quark is then determined by how much these fuzzy clouds overlap. The further apart the clouds are, the less they overlap, and the lighter the resulting particle. By arranging the positions of the three left-handed quarks and the six right-handed quarks in this extra space, the researchers could theoretically generate the exact mass differences observed in nature. The challenge was to find the specific arrangement of these points that would reproduce the known masses of all six quarks and the complex mixing patterns that govern how they change into one another.

Solving this problem was not a matter of simple calculation; it was a search through a vast, multi-dimensional landscape of possibilities. The researchers had to find the precise coordinates for twelve different variables that would satisfy ten distinct experimental constraints, including the masses of the quarks and the angles that describe how they mix. To navigate this complex space, they turned to techniques borrowed from machine learning, specifically numerical optimization algorithms. These tools are designed to find the best possible solution by iteratively adjusting parameters to minimize errors, much like how a computer learns to recognize patterns by trial and error. They ran thousands of simulations, starting with random arrangements of the particles in the extra dimensions and letting the algorithm refine the positions until the predicted masses matched the real-world data.

The results were striking. The researchers found multiple configurations of particle positions that could reproduce the observed properties of quarks with high precision. In the most successful models, the predicted values for the quark masses and their mixing angles fell within the range of experimental uncertainty, often matching the data to within a single standard deviation. This means the model is not just a rough approximation but a highly accurate description of the data. One particularly successful arrangement involved two extra dimensions, where the researchers imposed a symmetry condition to ensure the model did not violate known laws regarding the conservation of charge and parity. In this scenario, the particles settled into specific geometric patterns, with certain pairs of quarks clustering closely together to explain the heavy top quark, while others remained far apart to account for the light up and down quarks.

The study does not claim to have solved the mystery of flavor once and for all, but it demonstrates that a geometric explanation is viable. The researchers showed that the extreme differences in quark masses can arise naturally from the exponential sensitivity of wave function overlaps to small changes in distance. A tiny shift in position in the extra dimensions can lead to a massive change in the resulting particle weight. This provides a compelling alternative to previous theories that relied on complex symmetry breaking mechanisms or arbitrary mass matrices. The work suggests that the "flavor" of the universe might simply be a reflection of its shape in hidden dimensions. While the specific arrangement of these extra dimensions remains to be confirmed by future experiments, the study proves that such a geometric framework is capable of matching the precision of modern particle physics data.

The researchers also noted that their method leaves room for further refinement. Because the model has enough flexibility to fit the current data, it could potentially accommodate additional constraints, such as those from the lepton sector, which includes electrons and neutrinos. The fact that the algorithm found many different solutions that all worked well suggests that the landscape of possible configurations is rich and varied. Some solutions formed distinct clusters, hinting at underlying structures in the parameter space that could be explored further. This work opens a new avenue for understanding the fundamental building blocks of matter, shifting the focus from arbitrary numbers to the geometry of space itself. By treating the flavor puzzle as a problem of spatial arrangement, the researchers have provided a clear, testable framework that links the abstract world of particle physics to the tangible concept of distance and location.

Technical Summary: The Shape of Quark Flavors

Problem Statement
The flavor sector of the Standard Model (SM) is characterized by a large number of free parameters and striking mass hierarchies (e.g., $m_t/m_u \sim 10^5$ ) that lack a compelling underlying explanation, contrasting with the gauge sector which is dictated by gauge principles. While Grand Unified Theories (GUTs) and phenomenological approaches like the Froggatt-Nielsen mechanism or democratic mass matrix models have been proposed to address this "flavor puzzle," they often struggle with incompatibility with data or a loss of predictive power. This paper proposes a data-driven heuristic model where Yukawa couplings are not fundamental constants but arise from the geometric arrangement of fermion fields in extra spatial dimensions.

Methodology
The authors construct a model where the wave functions of chiral fermions are localized as Gaussian packets in $d$ extra dimensions ( $d=1$ or $d=2$ ), while the Higgs doublet wave function is flat. The effective 4D Yukawa couplings are determined by the overlap integrals of these Gaussian wave functions. Consequently, the Yukawa matrix elements depend exponentially on the relative separations of the fermion fields in the extra-dimensional space.

The study focuses on two specific scenarios:

One Extra Dimension: The positions of three left-handed quark doublets ( $q_i$ ) and six right-handed singlets ( $u_i, d_i$ ) are parametrized by real numbers. The model allows for non-trivial phases in the down-type Yukawa matrix to satisfy experimental constraints.
Two Extra Dimensions: To address the Strong CP problem without invoking the Peccei-Quinn mechanism, the authors impose Hermiticity on the Yukawa matrices ( $(y)^\dagger = y$ ). This requires two extra dimensions to provide sufficient freedom. The Hermiticity conditions constrain the field positions, reducing the independent parameters.

Numerical Optimization
Given the high-dimensional parameter space (12 parameters for both models) and the non-linear relationship between field positions and observables, the authors employ numerical optimization techniques from machine learning.

Loss Function: A total loss function is defined as the sum of weighted terms for quark masses ( $L_{qm}$ ), CKM angles ( $L_{CKM}$ ), the Jarlskog invariant ( $L_{Jar}$ ), and, in the 1D model, a term to suppress CP violation ( $L_{CP}$ ).
Weights: The loss function is normalized such that a $1\sigma$ deviation from the experimental central value contributes unity to the total loss.
Algorithm: The authors use the Adam optimizer with random initial configurations to minimize the loss function against 10 experimental constraints (6 quark masses and 4 CKM parameters) at the 1 TeV scale.

Key Results
The numerical optimization successfully identified field configurations that reproduce the observed SM flavor parameters with high precision:

One Extra Dimension: Out of 4096 initial samples, the best-fit sample achieved a total loss of 0.942. This indicates that the model reproduces the 10 observables within approximately $1\sigma$ of their experimental values. The resulting field configurations show that the large top Yukawa coupling arises from the close proximity of the third-generation left-handed and right-handed up-type fields.
Two Extra Dimensions: In this Hermitian model, the CP is automatically conserved in the Yukawa sector, though the topological CP phase needs to vanish separately. The best-fit sample achieved a loss of 0.457, representing an even tighter fit. The analysis revealed 16 distinct solution types based on sign choices in the Hermiticity constraints.
Geometric Patterns: Visualizations of the optimized field configurations (Figs. 1 and 2) reveal recurring geometric features. In both models, a pair of left-handed ( $q$ ) and right-handed up-type ( $u$ ) fields are consistently found in close proximity, naturally generating the $O(1)$ top Yukawa coupling. In the 2D model, the third left-handed quark field ( $q_3$ ) often aligns with the vertical axis defined by $q_2$ .
Parameter Space Structure: Using t-SNE dimensionality reduction on 502 viable samples in the 2D model, the authors observed clustering in the 12-dimensional parameter space, though the specific structure of these clusters remains to be fully understood.

Significance and Claims
The paper claims to provide a systematic, geometric framework for generating quark mass matrices and CKM parameters. By treating flavor parameters as emergent properties of field locations in extra dimensions, the model naturally explains the exponential hierarchy of quark masses through $O(1)$ separations in the extra-dimensional space.

The authors position this work as an alternative to traditional flavor models like Froggatt-Nielsen or democratic mass matrices. They emphasize that the models are relatively simple and generic, yet they yield numerous parameter configurations that fit experimental data at the $\lesssim 1\sigma$ level. The authors modestly suggest that while the current 12-parameter models are flexible, imposing one or two additional theoretical constraints could potentially reduce the viable configuration space to a unique solution. They note that the lepton sector, including neutrino masses, is not addressed in this work but could be analyzed within the same framework in future studies.

More like this