WorksPhysics and Astronomy
Gluon Entanglement Entropy inside a Nucleon: A Toy Model
Abstract
We construct a toy model of a nucleon, in which three static quarks interact via a SU(3) gauge field on a planar honeycomb lattice. The dynamics of the gauge field is described by the Kogut-Susskind Hamiltonian, truncated to the lowest three SU(3) irreducible representations. We show that the internal structure of the toy nucleon reflects salient features of the physical nucleon state. We then find the entanglement entropy of the gauge field within the nucleon state and compute its time evolution after a quench, in which all three valence quarks are suddenly removed. We show that the entanglement entropy in the final state is dominated by the dynamically generated contribution rather than the initial state entropy.
At a fundamental level, all high-energy reactions involving elementary particles are governed by a unitary S-matrix. However, as it is impossible to experimentally measure all details of the final state in a reaction that produces many particles, the coarse-grained final state is often described as a sample from a statistical particle distribution that carries a non-vanishing entropy. This interpretation dates back to Fermi’s statistical model [1] of multiparticle production, which posits that the entropy of the final particle distribution is microcanonically maximal. At its most inclusive level, the statistical model has been found to give a good description of hadron yields in reactions ranging from heavy-ion collisions to electron-positron annihilation [2, 3].
The origin of this coarse-grained entropy of the multiparticle final states has long been debated. Various arguments have invoked the decoherence of a high-density semi-classical gluon state in nuclear collisions [4, 5], the QCD analogue of Hawking-Unruh radiation [6], and the pre-formation of the entanglement entropy of partons inside the participating hadrons [7, 8, 9, 10, 11, 12], which is released by the reaction. In particular, it has been argued that the coarse-grained entropy characterizing the final state of a deep inelastic scattering reaction , which is usually obtained from the single-particle momentum distribution, is proportional to the proton’s gluon distribution function [7, 10] and can be interpreted as the entanglement entropy of low- gluons contained in the proton’s wave function. On the other hand, in the QCD-string breaking model of multi-particle production in high-energy annihilation, the flux-tube between the quark-antiquark pair is modeled as a coherent gluon field, and the final-state entropy is generated by string breaking [13, 14, 15, 16, 17, 18, 19, 20, 21].
In a deep inelastic scattering reaction, a single quark is suddenly ejected from the ground state nucleon after it has absorbed a high-energy, highly virtual photon. This process cannot be easily modeled on a small lattice, but we can create a setup that allows us to study the consequences of an even more drastic process, the simultaneous removal of all three color charges representing the valence quarks. This process, which cannot be easily realized experimentally, can be thought of as an idealized version of a high-energy reaction, where three valence quarks are suddenly ejected from the nucleon rather than a single one. Using a simplified model of a nucleon as a bound state of three static quarks, we compute the entanglement entropy of a localized region of gauge field in the toy nucleon ground state and trace its time evolution after the quench-like removal of the three color charges.
The toy model. We now construct a numerically exactly solvable toy model of a nucleon that allows us to explore the relative importance of initial-state entanglement entropy residing in the gluon field inside a nucleon and the entropy production by decoherence among the different components of the excited nucleon state after the onset of the reaction. In our setup a fast-moving nucleon is modeled by three static fundamental SU(3) color charges (the valence quarks) attached to the corners of a two-dimensional triangular lattice as shown in Figure 1. One can think of the planar lattice as a slice of the nucleon state at a given rapidity, where the valence quarks appear frozen because of time dilation. The dynamical gauge field is described by the Kogut-Susskind Hamiltonian in the minimal truncation of the SU(3) gauge group, which includes the irreps (0,0), (1,0) and (0,1) in Dynkin notation. A similar configuration has recently been used to compute the string tension associated with three static color charges from the energy contained in the gauge field on a hexagonal lattice [22]. Here we choose a triangular lattice to better represent the geometrical symmetry in our toy nucleon.
Figure 1. Triangular lattice configuration with three static fundamental SU(3) charges at its corners. The colored lines with arrows indicate the chromoelectric flux boundary condition that ensures that the overall configuration is a color singlet. We consider triangular lattices with 10 (shown here) and 15 plaquettes.
The SU(3) Kogut-Susskind Hamiltonian for a planar hexagonal lattice is given by
where is the gauge coupling, , and denotes the plaquette operator. We denote the state of the plaquette in the qutrit representation: corresponds to the electric flux ground state. Applying the plaquette operator gives , , and , where denotes the modulo 3 operation. For details of the calculation of the Hamiltonian matrix elements in the qutrit basis we refer to Chen et al. [22]. Here we consider gauge couplings and , which lie in the ergodic domain for this lattice Hamiltonian. The Hilbert space on which the Hamiltonian (1) acts depends on the state of the external links of the lattice, which represent external static charges and serve as fixed boundary conditions. Only the internal and edge links and plaquettes, shown as dashed lines or blue lines with arrowheads in Figure 1, are dynamical; the brown links injecting chromoelectric flux at the triangle corners are static.
The ground state of the lattice Hamiltonian accordingly depends on the external color charges. We denote the vacuum, i.e., the ground state in the absence of external charges (all external links are in the singlet representation), as , and the ground state in the presence of three external color charges in the fundamental representation at the locations indicated in Figure 1 (only external links in the location of the static quarks are chosen in the triplet representation) as . This state serves as our model for the fast moving nucleon.
Before we compute the entanglement entropy of the gauge field and its time evolution after the removal of the quarks, we explore the energy density and planar pressure per plaquette inside our toy nucleon. These can be obtained from the stress-energy tensor and written as
where denotes a plaquette and the sum is over all links forming the perimeter of the plaquette. The factor differs for contributions from edge links and internal links, the latter being shared between two neighboring plaquettes. Specifically, we assign to edge links and to internal links.
The (vacuum subtracted) energy densities and planar pressures for the lattice are given in Figure 2 and Figure 3 for a coupling of . We use color coding to indicate the magnitude of the energy density and pressure for the different types of plaquettes, with darker colors indicating larger values. Table I shows these values for and a weaker coupling for the three different types of plaquettes that exist in the lattice (body-centered, edge-centered and off-center) labeled by color in Figure 2 and Figure 3. As Table I shows, the energy density distribution inside the nucleon undergoes a structural change between the two couplings. At the larger coupling () the energy density reflects a Y-shaped, junction-like configuration shown in Figure 2, whereas at the lower coupling the gauge field configuration changes to a triangular shape. It is known from quenched lattice-QCD simulations [23] that the gauge field distribution in the nucleon exhibits such a junction shape when the valence quarks are well separated (see Fig. 6 in [23]) and the running QCD coupling is large.

Figure 2. Vacuum subtracted energy densities of the gauge field per plaquette for the lattice at a coupling of . Dark red: body-centered plaquettes; pale yellow: edge-centered plaquettes, light red: off-center plaquettes.

Figure 3. Vacuum subtracted planar pressure of the gauge field per plaquette for the lattice at a coupling of . Note the different color coding – Pale yellow: body-centered, dark red: edge-centered, orange: off-center.
| 0.3 | 0.5 | 0.3 | 0.5 | |
| body-centered | 0.425 | 0.714 | 0.166 | 0.083 |
| edge-centered | 0.619 | 0.561 | 0.334 | 0.138 |
| off-center | 0.493 | 0.690 | 0.143 | 0.116 |
Table I. Vacuum subtracted energy density and planar pressure values for the lattice at the couplings and . For the definition of the different plaquette types, see Figure 2 and Figure 3.
The pressure distribution inside the proton is experimentally known to be positive (repulsive) at short radii and negative (attractive) at large radii [24]. A theoretical analysis [25] shows that the dynamical pressure component is positive everywhere (see Figure 2 in [25]). The pressure component due to the trace anomaly, which is absent in our toy model, is negative and dominates at large radii.
Before computing the entanglement entropy of the gauge field, we list some of the shortcomings of our toy model: (1) Our valence quarks are static and not dynamical and thus do not carry entropy themselves. (2) Our toy nucleon is two-dimensional, not three-dimensional. (3) Our nucleon is not moving so it has much lower energy and thus potentially lower entanglement entropy than a fast-moving nucleon. (4) As level densities grow rapidly with the space dimension and local Hilbert space size (such as the number of irreps included in the basis) of a quantum system, our model can be expected to underestimate the absolute magnitude of the entanglement entropies of the initial and final states. (5) Being constrained by the fixed lattice, the excitation energy created by the sudden removal of the quarks cannot disperse in space. These deficiencies can only be removed by a full dynamical quantum calculation of QCD with dynamical quarks in (3+1) dimensions, which is currently impossible.
Entanglement entropy
We now turn to discuss the entanglement entropy of gluons inside our toy nucleon and its time dependence. Although the toy nucleon represents a pure quantum state, the gauge field on a geometrically restricted subregion of the lattice is given by a density matrix:
where denotes the complement of the region . As is well known, even the vacuum state has a nonzero entanglement entropy when restricted to a finite region . For theories with a mass gap, is proportional to the surface area of the domain . The vacuum subtracted entanglement entropy of the gauge field in our toy nucleon state for the domain is given by
We interpret this quantity as the entanglement entropy of the gauge field with respect to a localized domain within our toy nucleon.
When we suddenly remove the external charges by setting all external links to the singlet state, the nucleon state is no longer an eigenstate of the Hamiltonian, but rather corresponds to a coherent superposition of excited states of the SU(3) vacuum. We can track the time evolution of this state by evolving the gauge field configuration of the state in the eigenbasis of the vacuum sector of the Hamiltonian (1):
with the initial condition . The entanglement entropy of the state is denoted as .
We are now ready to discuss the time evolution of the gluon entanglement entropy in our toy nucleon, beginning with the lattice shown in Figure 1. Figure 4 shows the vacuum subtracted entanglement entropy for four different subdomains of the toy nucleon shown in Figure 1 comprising plaquettes, respectively. Using the labeling of the plaquettes in Figure 1, these are given by the subregions , , , and . For those regions that are not symmetric under rotations, we average our results over the plaquettes obtained by rotation by multiples of .
Figure 4. Vacuum subtracted entanglement entropies of subsystems of size of the toy nucleon as a function of time. The top panel is for ; the bottom panel is for . The values for represent the gluon entanglement entropies in the original toy nucleon ground state. Note that the entropy is shown on a logarithmic scale.
We see that only a fraction of the final equilibrated entropy can be attributed to the entanglement entropy of the gauge field in the toy nucleon, shown by the value of for . Numerical values are shown in Table II. We also note that the relative initial entropy increase is faster for , but the initial and final values of the entanglement entropy are larger for due to the increased energy stored in the gauge field for the larger coupling.
| \multicolumn{4}{c | }{0.3} | \multicolumn{4}{c}{0.5} | ||||||
| 1 | 3 | 5 | 7 | 1 | 3 | 5 | 7 | |
| 0.499 | 0.535 | 0.479 | 0.308 | 0.558 | 0.692 | 0.727 | 0.642 | |
| 0.861 | 2.184 | 2.607 | 1.642 | 1.007 | 2.770 | 3.732 | 2.554 | |
| 0.064 | 0.140 | 0.175 | 0.169 | 0.026 | 0.048 | 0.056 | 0.050 |
Table II. Vacuum subtracted entanglement entropies for different subregions of the toy nucleon in Figure 1 for and at and . The entanglement entropies of the subregions in the vacuum state are also shown. As can be seen, only a small fraction of the final entropy can be attributed to the entanglement entropy of the gauge field inside the unperturbed nucleon state.
In Figure 5 we compare the vacuum subtracted entanglement entropies of a three-plaquette subsystem () for two lattice sizes and two different coupling constants (solid curves) and (dashed curves). Here denotes the side length of the triangular lattice. In particular, lines (red) refer to the 10-plaquette lattice shown in Figure 1 while lines (black) indicate a triangular lattice composed of 15 hexagonal plaquettes. Unsurprisingly, the initial-state entanglement entropy is larger for than because the gauge field configuration in the probe region is less constrained by the requirement that the overall field configuration is a pure quantum state.
Figure 5. Comparison of the vacuum subtracted entanglement entropies of a three-plaquette subsystem () for two different coupling constants (solid curves) and (dashed curves) and two lattice sizes (red curves) and (black curves), where indicates the side length of the triangular lattice. The subsystem is the same as shown in Figure 4 for and corresponds to the central three plaquettes that do not overlap with the edges for . The values for represent the gluon entanglement entropies in the original toy nucleon ground state. Note that the entropy is shown on a logarithmic scale.
Conclusions. We have constructed a simplified model
of a nucleon with three static quarks that allows us to numerically compute the entanglement entropy of the gauge field in the interior of the nucleon and its time evolution after a sudden perturbation. The gauge field binding the quarks has features that mirror those known from Euclidean lattice QCD simulations and experiment. We have found that the final-state entropy in the gauge field after the sudden removal of the valence quarks far exceeds the entanglement entropy of the gauge field in the unperturbed nucleon state. We conclude that attempts to “measure” the gluon entanglement entropy in deep-inelastic scattering, where one quark is ejected violently from the nucleon, suffer from severe contamination by the statistical entropy created in the far out-of-equilibrium evolution of the final state.
Finally, while it is possible to compute the entanglement entropy of a region within the nucleon using the tools of Euclidean lattice gauge theory [26, 27, 28], it is more challenging to measure this entropy experimentally. As our model study has shown, measurements of entanglement entropy based on processes that destroy the nucleon, such as inelastic scattering, are at least contaminated, if not dominated by statistical entropy created by the measurement process. It is an open question whether experimental processes using nondestructive measurements can be devised that can reliably determine the entanglement entropy of regions within the nucleon ground state.
Acknowledgments: Our study was motivated by remarks by K. Rajagopal in a lecture at the Kavli Institute for Theoretical Physics (https://online.kitp.ucsb.edu/online/quarkgluon-c26/rajagopal/). We thank D. Kharzeev and K. Rajagopal for comments on a near-final version of the manuscript. D.H. and B.M. are supported by the National Science Foundation (Project PHY-2434506). B.M. also acknowledges support by the U.S. Department of Energy, Office of Science (Grant DE-FG02-05ER41367). X.Y. is supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics, InQubator for Quantum Simulation (IQuS) under DOE (NP) Award Number DE-SC0020970 via the program on Quantum Horizons: QIS Research and Innovation for Nuclear Science. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a Department of Energy Office of Science User Facility using NERSC award NP-ERCAP0032083. This work was enabled, in part, by the use of advanced computational, storage and networking infrastructure provided by the Hyak supercomputer system at the University of Washington. This research was supported in part by grant NSF PHY-2309135 to the Kavli Institute for Theoretical Physics (KITP).
References
References
- [1]E. Fermi, Prog. Theor. Phys. 5, 570 (1950).
- [2]J. Cleymans and H. Satz, Z. Phys. C 57, 135 (1993), arXiv:hep-ph/9207204.DOI
- [3]F. Becattini, P. Castorina, J. Manninen, and H. Satz, Eur. Phys. J. C 56, 493 (2008), arXiv:0805.0964 [hep-ph].DOI
- [4]A. Kovner, L. D. McLerran, and H. Weigert, Phys. Rev. D 52, 6231 (1995), arXiv:hep-ph/9502289.DOI
- [5]R. J. Fries, B. Müller, and A. Schäfer, Phys. Rev. C 79, 034904 (2009), arXiv:0807.1093 [nucl-th].DOI
- [6]P. Castorina, D. E. Kharzeev, and H. Satz, Eur. Phys. J. C 52, 187 (2007), arXiv:0704.1426 [hep-ph].DOI
- [7]D. E. Kharzeev and E. M. Levin, Phys. Rev. D 95, 114008 (2017), arXiv:1702.03489 [hep-ph].DOI
- [8]O. K. Baker and D. E. Kharzeev, Phys. Rev. D 98, 054007 (2018), arXiv:1712.04558 [hep-ph].DOI
- [9]Z. Tu, D. E. Kharzeev, and T. Ullrich, Phys. Rev. Lett. 124, 062001 (2020), arXiv:1904.11974 [hep-ph].DOI
- [10]D. E. Kharzeev and E. M. Levin, Phys. Rev. D 104, L031503 (2021), arXiv:2102.09773 [hep-ph].DOI
- [11]K. Zhang, K. Hao, D. Kharzeev, and V. Korepin, Phys. Rev. D 105, 014002 (2022), arXiv:2110.04881 [quant-ph].DOI
- [12]U. Gürsoy, D. E. Kharzeev, and J. F. Pedraza, Phys. Rev. D 110, 074008 (2024), arXiv:2306.16145 [hep-th].DOI
- [13]A. Florio, D. Frenklakh, K. Ikeda, D. Kharzeev, V. Korepin, S. Shi, and K. Yu, Phys. Rev. Lett. 131, 021902 (2023), arXiv:2301.11991 [hep-ph].DOI
- [14]S. Grieninger, K. Ikeda, D. E. Kharzeev, and I. Zahed, Phys. Rev. D 109, 016023 (2024), arXiv:2312.03172 [hep-th].DOI
- [15]K. Lee, J. Mulligan, F. Ringer, and X. Yao, Phys. Rev. D 108, 094518 (2023), arXiv:2308.03878 [quant-ph].DOI
- [16]A. Florio, D. Frenklakh, K. Ikeda, D. E. Kharzeev, V. Korepin, S. Shi, and K. Yu, Phys. Rev. D 110, 094029 (2024), arXiv:2404.00087 [hep-ph].DOI
- [17]Y. Liu, W.-Y. Zhang, Z.-H. Zhu, M.-G. He, Z.-S. Yuan, and J.-W. Pan, Phys. Rev. Lett. 135, 101902 (2025), arXiv:2411.15443 [cond-mat.quant-gas].arxiv.org/abs/2411.15443
- [18]K. Ikeda, Z.-B. Kang, D. E. Kharzeev, and W. Qian, (2025), arXiv:2512.18062 [hep-ph].arxiv.org/abs/2512.18062
- [19]A. Florio, D. Frenklakh, K. Ikeda, D. Kharzeev, V. Korepin, S. Shi, and K. Yu, PoS QCHSC24, 056 (2025).
- [20]A. Florio, D. Frenklakh, S. Grieninger, D. E. Kharzeev, A. Palermo, and S. Shi, Phys. Rev. D 112, 094502 (2025), arXiv:2506.14983 [hep-ph].arxiv.org/abs/2506.14983
- [21]S. Grieninger, M. J. Savage, and N. A. Zemlevsky, (2026), arXiv:2601.08825 [hep-ph].arxiv.org/abs/2601.08825
- [22]V. Chen, B. Müller, and X. Yao, (2026), arXiv:2601.10065 [hep-lat].arxiv.org/abs/2601.10065
- [23]F. Bissey, F.-G. Cao, A. Kitson, B. G. Lasscock, D. B. Leinweber, A. I. Signal, A. G. Williams, and J. M. Zanotti, Nucl. Phys. B Proc. Suppl. 141, 22 (2005), arXiv:hep-lat/0501004.DOI
- [24]V. D. Burkert, L. Elouadrhiri, and F. X. Girod, Nature 557, 396 (2018).
- [25]D. Fujii, M. Kawaguchi, and M. Tanaka, Phys. Lett. B 866, 139559 (2025), arXiv:2503.09686 [hep-ph].DOI
- [26]P. V. Buividovich and M. I. Polikarpov, Nucl. Phys. B 802, 458 (2008), arXiv:0802.4247 [hep-lat].DOI
- [27]E. Ito, K. Nagata, Y. Nakagawa, A. Nakamura, and V. I. Zakharov, PTEP 2016, 061B01 (2016), arXiv:1512.01334 [hep-th].
- [28]A. Rabenstein, N. Bodendorfer, P. Buividovich, and A. Schäfer, Phys. Rev. D 100, 034504 (2019), arXiv:1812.04279 [hep-lat].DOI


