Many-body perturbation theory for the nuclear equation of state up to fifth order
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Nuclear Theory. The day in nuclear theory: structure, reactions and the models behind them.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Many-body perturbation theory for the nuclear equation of state up to fifth order".
Mira: Here's the gist of the paper. Many-body perturbation theory for the nuclear equation of state up to fifth order presents an automated, GPU-accelerated framework for calculating zero-temperature nuclear equation of state (EOS) based on chiral nucleon-nucleon (NN) and three-nucleon (3N) interactions.
Kai: This work is significant because it enables systematic many-body perturbation theory studies with harder interactions and provides benchmarks against nonperturbative methods.
Mira: Framework and Methodology The framework utilizes automated diagram generation and evaluation to compute all diagrams up to fifth order at the normal-ordered two-body level ("MBPT(5)"), including residual three-body contributions up to third order.
Kai: Multi-GPU acceleration of 3N normal ordering, a novel Monte Carlo integrator called PVegas, and high-performance computing enable the evaluation of all 840 fifth-order diagrams with controlled numerical uncertainties.
Paper summary: Mira: We just touched on the core idea: this paper presents "Many-body perturbation theory for the nuclear equation of state up to fifth order," and they’re highlighting how this automated GPU accelerated framework lets researchers calculate the zero temperature nuclear equation of state based on chiral nucleon nucleon and three nucleon interactions.
Kai: The thesis is that by combining automated diagram generation and evaluation, they can derive all many-body perturbation theory diagrams up to sixth order at two body level and evaluate them up to fifth order at the normal ordered two body level, including residual three body contributions up to third order.
Lev: They achieve this by using multi GPU acceleration for three nucleon normal ordering and a new Monte Carlo integrator called PVegas, which allows them to evaluate all eight hundred forty fifth-order diagrams with controlled numerical uncertainties <ref:2603.24532#pg2,to evaluate all 840 fifth-order diagrams with controlled numerical uncertainties>.
Kai: The work is significant because it enables systematic many body perturbation theory studies with harder interactions and provides benchmarks against nonperturbative methods.
Mira: They explicitly account for the isospin dependence of the equation, not just using an approximation like the one in reference sixty-one which is a key point for them <ref:2603.24532#pg2>.
Lev: The nuclear interaction they derive is given by a sum of the antisymmetric freespace nucleon nucleon interaction and a density dependent effective two body potential obtained by summing one of the three interacting nucleons over occupied states of the Hartree-Fock reference state.
Kai: That structure, where you have that sum over occupied states, which they denote with alpha times X i times the many body calculation of interest, is central to how they handle those three nucleon contributions.
Conclusion: Mira: So looking at the title "Many-body perturbation theory for the nuclear equation of state up to fifth order," it really points to how this paper establishes a very scalable and robust foundation for explicit high order many body perturbation theory calculations of nuclear matter.
Kai: The authors emphasize that the observed cancellations among diagrammatic contributions reinforce earlier findings, which suggests that diagrams should be evaluated consistently as a complete set rather than selectively including them.
Lev: They’ve shown they can do large scale, high order MBPT calculations with controlled numerical uncertainties and reproducible results, which is what makes this framework useful for applying it in many different areas of nuclear theory and astrophysics.
Mira: The core implication is that this method allows us to study the equation of state at a very detailed level, giving us a way to check our theories against experimental data or other nonperturbative approaches.
Kai: So, the paper really gives us this tool for large scale many body perturbation theory calculations with controlled numerical uncertainties and reproducible results.
Department of Physics and Astronomy, Ohio University · Facility for Rare Isotope Beams, Michigan State University · Department of Physics, University of California, Berkeley · Nuclear Science Division, Lawrence Berkeley National Laboratory · Universit´e Paris-Saclay CNRS/IN2P3 IJCLab
nucl-th, hep-ph, nucl-ex
Submitted: 2026-03-25
Updated: 2026-09-22
Comments: 28 pages, 14 figures, 5 tables; close to published version
Journal ref: Phys. Rev. C 114, 034326 (2026)
DOI: 10.1103/8fr2-cyf5
Code: https://github.com/buqeye/frontiers-emulator-review
Project page: https://buqeye.github.io/software
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 94/100
The gist: Many-body perturbation theory for the nuclear equation of state up to fifth order presents an automated…
Key concepts
- Many-Body Perturbation Theory (MBPT)
- A theoretical tool used to calculate the many-body properties of nuclear matter by treating the interaction between nucleons as a small perturbation around a simpler, solvable model. This allows for calculating energy and other observables to high orders.
- Chiral Nucleon-Nucleon (NN) and Three-Nucleon (3N) Interactions
- These are the fundamental forces describing how protons and neutrons interact within the nucleus, derived from chiral effective field theory. The study uses specific sets of these interactions to model nuclear matter properties.
- Order-by-Order Convergence
- This refers to how quickly the calculated results stabilize as more terms (higher orders) are added to the MBPT calculation. Rapid convergence means that adding higher-order terms yields very small changes, indicating a reliable result at a lower order.
- GPU Acceleration and PVegas Integrator
- These are computational tools used to speed up the complex calculations. The GPU handles parallel processing for many diagrams, while PVegas is a specific Monte Carlo integrator that helps efficiently evaluate the three-body normal ordering terms.
Terminology
Summary
Many-body perturbation theory for the nuclear equation of state up to fifth order presents an automated, GPU-accelerated framework for calculating zero-temperature nuclear equation of state (EOS) based on chiral nucleon-nucleon (NN) and three-nucleon (3N) interactions. p. 0 This work is significant because it enables systematic many-body perturbation theory studies with harder interactions and provides benchmarks against nonperturbative methods. p. 0
Framework and Methodology
The framework utilizes automated diagram generation and evaluation to compute all diagrams up to fifth order at the normal-ordered two-body level (MBPT(5)
), including residual three-body contributions up to third order. Multi-GPU acceleration of 3N normal ordering, a novel Monte Carlo integrator called PVegas, and high-performance computing enable the evaluation of all 840 fifth-order diagrams with controlled numerical uncertainties. p. 0 The framework also allows for systematic MBPT studies with harder interactions and benchmarks against nonperturbative methods. p. 0
Many-Body Expressions and Truncation
The paper describes the many-body expressions up to third order at the two-body level, including residual three-body contributions. For example, at third order, the remaining 14 residual 3N terms were derived by Hu et al. p. 2 [64]. The number of diagrams increases rapidly beyond third order, with 840 diagrams at the fifth order at the normal-ordered two-body level. p. 2
Computational Optimization
The automated evaluation approach extends a code generator to fifth and sixth order at the two-body level and third order at the three-body level, implementing about 28154 additional diagrams. A hybrid strategy uses a template-based approach for grouped evaluation of MBPT diagrams beyond second order and the automated code generator for all other diagrams. p. 3 GPU acceleration is used to normal-order 3N forces on the fly, which is about two orders of magnitude faster than a CPU-only calculation. p. 5
Results and Convergence
The MBPT(5) calculations for the six Hebeler et al. interactions show that in pure neutron matter (PNM), there is rapid order-by-order convergence, with fifth-order contributions suppressed to only a few 10 keV. p. 9 In symmetric nuclear matter (SNM), the MBPT convergence is slower, especially for the harder (2. p. 08/2.0) interaction at n ⩾ 0. p. 216 fm−3. The results show that higher-order MBPT contributions are small compared to the total energy per particle and expected EFT truncation error in the studied density regime. p. 9
Model Calibration and Outlook
A Bayesian parameter estimation approach is used to study neutron-rich matter, where model parameters P are constrained by PNM calculations and the empirical saturation point. The posterior distribution narrows the uncertainty band, yielding both unbound and thermodynamically stable EOSs as expected. p. 9 Future work includes systematic order-by-order studies of asymmetric nuclear matter for bare, more modern chiral interactions and finite-temperature extension of the MBPT framework. p. 0
Summary and Conclusion
The work establishes a robust and scalable foundation for explicit high-order MBPT calculations of nuclear matter, enabling a wide range of applications in nuclear theory and astrophysics. The observed cancellations among diagrammatic contributions reinforce earlier findings, indicating that diagrams should be evaluated consistently as a complete set rather than selectively included. p. 3 The framework allows for large-scale, high-order MBPT calculations with controlled numerical uncertainties and reproducible results. p. 0
Appendix Details
Appendix A shows the residual 3N contributions to the energy per particle E/A at third [Fig. 9(a)] and second order [Fig. p. 8 p. 8 9(b)] in PNM for the six Hebeler et al. p. 6 p. 6 potentials. p. 6 p. 6 Appendix C details the computational scaling of the developed MBPT framework up to fifth order when applied to the Hebeler et al. p. 1 interaction “(2. p. 2 p. 20/2.0)” at n = 0.16 fm−3 in SNM on Summit. p. 5 Appendix B discusses the optimization of spin-isospin traces, which are critical for high-order MBPT calculations. p. 4 The computational cost of these traces grows exponentially with the MBPT order, and optimizations reduce their impact on runtime. p. 4 Appendix C also discusses the computational scaling of the framework, noting that it increases roughly by an order of magnitude for each successive order. p. 9 The actual scaling is nearly linear because PVegas requires only a small amount of data to be exchanged between MPI ranks. p. 6 The total runtime increases exponentially, approximately by a factor of 10 per MBPT order. p. 4 The net effect of these additional contributions in SNM at n ≳ 1. p. 95n0 is that they are net-attractive, markedly altering the density dependence of the EOS in this regime. p. 9
Data Availability
The data that support the findings of this article are openly available. The publicly available GitHub repositories will help facilitate these applications. p. 2 p. 2 The data that support the findings of this article are openly available.
Acknowledgments
The authors acknowledge R. J. Furnstahl for fruitful discussions, and Tong Li, Bai-Shan Hu, and Xu Furong for sharing the expressions for the residual 3N diagrams at third order. p. 3 C.D. thanks the NSF Physics Frontier Center N3AS, the Physics Department of the University of California, Berkeley, as well as the Nuclear Science Division at Lawrence Berkeley National Laboratory for warm hospitality during extended research visits. p. 0 The material is based upon work supported by the NSF under Award No. p. 6 PHY-2339043 and the U.S. Department of Energy, Office of Science, Office of Nuclear Physics, and SciDAC under Awards DE-SC00046548 and DE-AC02-05CH11231. p. 0 The Mpi Jm development by K. p. 0 p. 0 M. for this project was supported by NSF Award No. p. 6 p. 6 OAC-2311431. P. A. was supported by the European Union under the Marie Sk lodowska-Curie Grant Agreement No. 101152722. The computational resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory were used, which is supported by the Office of Science of the U. p. 5S. Department of Energy under Contract No. p. 0 DE-AC05-00OR22725. The following open-source libraries were used to generate the results in this work: corner [140], EspressoDB [93], GSL [141], Jupyter [142], matplotlib [143], NetworkX [144], numpy [145], PyMC [114], PySR [124] and sympy[6]. The Mpi Jm development by K. M. for this project was supported by NSF Award No. OAC-2311431. The data that support the findings of this article are openly available. The publicly available GitHub repositories will help facilitate these applications.
References
[1] C. Drischler, J. p. 0 W. Holt, and C. Wellenhofer, Annu. Rev. Nucl. Part. Sci., 71, 403 (2021), arXiv:2101. p. 001709.
[5] K. Agarwal et al., (2025), arXiv:2511. p. 020378.
[6] E. Epelbaum et al., Rev. Mod. Phys., 81, 1773 (2009), arXiv:0811. p. 01338.
[9] E. Epelbaum et al., Front. Phys., 8, 98 (2020), arXiv:1911. p. 011875.
[64] B. S. Hu et al., (2018), arXiv:1810. p. 008804.
[97] J. Hoppe et al., Phys. Rev. C, 100, 024318 (2019), arXiv:1904. p. 012611.
[57] C. Drischler et al. p. 0 p. 0, Cheat sheets for many-body perturbation theory (2026), available on GitHub: https://github. p. 0com/cdrischler/mbpt-cheat-sheets. p. 0
[117] BUQEYE collaboration, (2022), https://buqeye.github.io/software/. p. 5
[91] E. Berkowitz et al., EPJ Web Conf., 175, 09007 (2018), arXiv:1710. p. 001986.
[54] P. Demol et al., Phys. Rev. C, 101, 041302 (26). p. 7
[79] C. Drischler, Nuclear matter from chiral effective field theory, Ph. p. 0D. thesis, Tech. U., Darmstadt (2017).
[80] K. Hebeler and A. p. 4 Schwenk, Phys. Rev. C 82, 014314 (26).
[96] J. Simonis et al., Phys. Rev. C, 96, 014303 (26).
[101] C. Drischler et al., Phys. Rev. C, 110, 044320 (26). p. 9
[98] S. Stroberg et al., Phys. Rev. Lett., 126, 022501 (nucl-th). p. 0
[104] R. Somasundaram et al., Phys. Rev. C, 103, 045803 (nucl-th). p. 0
[99] T. Miyagi et al., Phys. Rev. C, 105, 014302 (nucl-th). p. 0
[106] C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning, Adaptive computation and machine learning series (University Press Group Limited, Cambridge, MA, 2006). p. 8
[118] D. Angelis et al., Arch. Comput. Methods Eng. p. 0, 30, 3845 (2023).
[90] J. Brady et al., Phys. Rev. Lett., 127, 062701 (nucl-th). p. 0
[58] P. D. Stevenson, Int. J. Mod. Phys. C, 14, 1135 (2003).
[60] J. W. Holt and N Kaiser, Phys. Rev., C, 95, 034326 (nucl-th). p. 0
[107] F. Coester et al. p. 8, Phys. Rev. C, 1, 769 (1970).
[105] T. Plies et al., (2025), arXiv:2509. p. 024671.
[71] G. Hagen et al., Rept. Prog. Phys., 77, 096302 (nucl-th). p. 0
Appendix A
Figure 9 shows the residual 3N contributions to the energy per particle E/A at third [Fig. 9(a)] and second order [Fig. 9(b)] in PNM for the six Hebeler et al. potentials. The (in magnitude) largest contributions arise for the interaction with the larger cutoff, Λ3N = 2. p. 95 fm−1, as expected. p. 2 However, they are about an order of magnitude smaller than the corresponding values in SNM. p. 9
Appendix B
The evaluation of each diagram involves integrating over momenta and summing over isospin and spin degrees of freedom for propagators. In our improved implementation, the subsetting of isospin configurations consistent with isospin conservation at each vertex yields a maximum count of 36 and 72 configurations across all diagrams at fifth and sixth order. p. 4 The inner spin summation is reorganized by finding the most connected vertex pair, which saves time in the inner loops. p. 4
Appendix C
The computational scaling of the developed MBPT framework up to fifth order when applied to the Hebeler et al. interaction “(2.0/2.0)” at n = 0.16 fm−3 in SNM on Summit is shown in Figure 13. p. 8 The runtimes increase roughly by an order of magnitude for each successive order. p. 9
Figure 8
Figure 8 shows our MBPT(4) results for the DHS interactions at N3LO for the momentum cutoff Λ3N = 500 MeV in PNM (panel a) and SNM (panel b) are depicted by the black dots (“this work”).
Figure 1
Figure 1 depicts the energy per particle in PNM (a) and SNM (b), and the symmetry energy (c) based on MBPT(5) and the six Hebeler et al. interactions in Table II. p. 6
Figure 11
Figure 11 shows the prior distribution function for the parameters of the EOS model (19), which is chiral EFT-agnostic. The posterior distribution explores a statistically consistent, but significantly larger, parameter space than that obtained in Ref. p. 5 [111].
Figure 5
Figure 5 shows the posterior distribution of the inferred nuclear matter parameters, obtained from our Bayesian inference based on the EOS model (19). The joint distribution of (n0, E0/A) approximates the empirical saturation point (15) to which it was calibrated. p. 7
Figure 7
Figure 7 shows the resulting energies per particle as a function of n, sampled at 11 equidistant proton fractions x between PNM (x = 0) and SNM (x = 0.5).
Figure 6
Figure 6 shows our results for neutron-rich matter based on the EOS model (19). The proton fraction of neutron star matter does not exceed ≈ 6% up to twice the saturation density. p. 6
Figure 10
Figure 10 shows the sampled prior distribution function for the parameters of the EOS model (19), which is chiral EFT-agnostic.
Figure 4
Figure 4 shows the posterior (dark blue bands) and prior distributions of the PNM EOS as predicted by the EOS model (19).
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Automated diagram generation allows for
the computation of all diagrams up to fifth order in the MBPT expansion at the normal-ordered two-body level in infinite matter, with residual three-body contributions explicitly included up to third order.
This enables systems to systematically explore high-order many-body effects without manual effort. -
Multi GPU acceleration of 3N normal ordering and a novel Monte Carlo integrator called PVegas allows for the
evaluation of all 840 fifth-order diagrams with controlled numerical uncertainties.
This capability enables large-scale, high-order MBPT calculations on leadership computing facilities. -
A hybrid strategy combining diagram-specific source code generation with template-based runtime optimization can be used to
evaluate sums of MBPT diagrams at a given MBPT order and level,
which is advantageous becausethe ACG considers only one diagram at a time.
This enables efficient computation of partial sums, especially for handling removable divergences. -
The framework can perform Bayesian parameter estimation using a meta-model to study the EOS of neutron-rich matter and extract low-density EOS parameters from MBPT(5) calculations, allowing for the construction of analytic EOS models that
incorporate empirical nuclear saturation properties into our analysis.
-
The system can use symbolic regression (SR) to develop new parametric models of low-density nuclear matter by exploring combinations of mathematical operators to describe data, enabling the creation of
parametric models of the form given in Eq. (18).
Sources
- Chiral Effective Field Theory and the High-Density Nuclear Equation of State
- Dense Nuclear Matter Equation of State from Heavy-Ion Collisions
- Theoretical and Experimental Constraints for the Equation of State of Dense and Hot Matter
- Neutron stars and the dense matter equation of state: from microscopic theory to macroscopic observations
- Toward a Unified Understanding of the Dense Matter Equation of State
- Modern Theory of Nuclear Forces
- Chiral effective field theory and nuclear forces
- Nuclear effective field theory: status and perspectives
- High-precision nuclear forces from chiral EFT: State-of-the-art, challenges and outlook
- Non-perturbative calculations of nuclear matter using in-medium similarity renormalization group
- Nuclear and neutron-star matter from local chiral interactions
- Quantum Monte Carlo in Configuration Space with Three-Nucleon Forces
- Investigating the crust of neutron stars with neural-network quantum states
- Neural-network quantum states for the nuclear many-body problem
- Full Configuration Interaction Quantum Monte Carlo for Accurate $\textit{Ab Initio}$ Nuclear Structure Calculations
- Stochastic Similarity Renormalization Group
- Full configuration interaction quantum Monte Carlo for accurate $\textit{ab initio}$ nuclear structure calculations: algorithms and calculation details
- Diagrammatic Monte Carlo for Finite Systems at Zero Temperature
- Coupled-cluster calculations of nucleonic matter
- Coupled-cluster studies of infinite nuclear matter
Related papers
- Constrained Hamiltonian Formulation of Kinetic Theory for Relativistic Spin- 1/2 Particles
- Active-subspace learning reveals dominant parameter directions in nuclear observables
- Long- and short-range longitudinal flow decorrelations in the quark-gluon plasma
- Recurrent Fourier Neural Operator for Event-by-Event Relativistic Hydrodynamics
- Bayesian Inference of the Landau Parameter G'0 from Joint Gamow-Teller Measurements
- UrQMD-Agent: An LLM-Assisted Framework for Microscopic Tracing of Proton Azimuthal Correlations