Hopping Parameter Expansion in Lattice QCD

• Hopping Parameter Expansion (HPE) is an expansion technique in lattice QCD that expresses the fermion determinant as a sum over gauge-invariant closed trajectories.
• It decouples temporal winding effects by isolating chemical potential contributions, enabling effective heavy-quark thermodynamic analyses through Polyakov-loop sectors.
• Systematic high-order corrections and algorithmic reformulations of HPE enhance simulation efficiency and precision near heavy-quark critical points.

Searching arXiv for the cited HPE literature to ground the article in current records. {"query":"hopping parameter expansion lattice QCD heavy quark finite density Wilson fermions arXiv", "max_results": 10} {"query":"(Ejiri et al., 2023) hopping parameter expansion", "max_results": 5} Hopping parameter expansion (HPE) is an expansion in powers of a lattice hopping parameter around a limit in which propagation between neighboring sites is suppressed. In lattice QCD with Wilson fermions, it is the Taylor expansion of the fermion determinant, or of $\ln \det M$, about the heavy-quark point $\kappa=0$, where $\kappa$ is small for heavy quarks. In that setting, each order of the expansion is a sum over closed lattice trajectories, so the determinant is reorganized into Wilson-loop and Polyakov-loop sectors (Wakabayashi et al., 2021). Closely related expansions also appear in strong-coupling treatments of bosonic lattice models, transfer-matrix analyses of gauge theories with matter, and linked-cluster formulations of lattice field theory (Kimura, 2012).

1. Formal definition and loop interpretation

For Wilson fermions, the lattice Dirac operator is commonly written as

$M=1-\kappa Q$

or equivalently

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$

with $\kappa$ related naively to the bare quark mass by

$\kappa=\frac{1}{2m+8}$

in one standard convention, and by

$\kappa_f=\frac{1}{2am_f+8}$

when the lattice spacing is written explicitly (Aarts et al., 2014, Kanaya et al., 2022). The HPE is then

$$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$

or, in the Wilson-kernel notation,

$$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$

Because $\kappa=0$0 or $\kappa=0$1 is a nearest-neighbor hopping matrix, $\kappa=0$2 and $\kappa=0$3 are nonzero only for closed paths of length $\kappa=0$4. The order in $\kappa=0$5 is therefore the loop length, and the determinant becomes a weighted sum of gauge-invariant closed trajectories (Aarts et al., 2015, Wakabayashi et al., 2021).

This loop interpretation is the structural reason HPE is useful. In heavy-quark lattice QCD, short contractible loops renormalize gauge-action-like terms, while temporally winding loops couple directly to thermal observables such as the Polyakov loop. The lowest nontrivial nonwinding term is the plaquette contribution at order $\kappa=0$6, whereas the leading thermal term is the straight Polyakov loop at order $\kappa=0$7 (Kitazawa et al., 2023). For example, in one normalization,

$\kappa=0$8

so already at leading order the expansion produces a shift of the plaquette coupling and a linear source for the Polyakov loop (Kanaya et al., 2022).

The same logic appears outside QCD. In the spin-1 Bose–Hubbard model, the strong-coupling expansion is an expansion in the hopping amplitude $\kappa=0$9 around the atomic limit $\kappa$0, with the Mott state as the unperturbed vacuum and defect energies computed order by order in $\kappa$1 (Kimura, 2012). In $\kappa$2-dimensional gauge theory with fundamental matter, the transfer matrix is expanded in powers of

$\kappa$3

so that powers of $\kappa$4 count matter-hopping processes in the heavy-mass regime (1705.01549).

2. Temporal winding, chemical potential, and heavy-quark thermodynamics

At finite density, HPE acquires a particularly transparent structure because the chemical potential enters only through temporal hoppings. In the Wilson-fermion matrix, the temporal forward and backward terms carry factors $\kappa$5, so a closed path with temporal winding number $\kappa$6 acquires $\kappa$7 (Ejiri et al., 2023). The expansion coefficients can therefore be decomposed as

$\kappa$8

or equivalently into $\kappa$9 and $M=1-\kappa Q$0 pieces, with the latter generating the complex phase of the determinant (Ejiri et al., 2023).

This decomposition is central in heavy-quark finite-temperature QCD. The leading winding contribution is the singly wound Polyakov-loop term, while higher winding sectors are parametrically smaller in the regime studied around the heavy-quark critical point (Wakabayashi et al., 2021). In thermodynamic applications, the determinant can be reorganized by net winding number $M=1-\kappa Q$1, yielding a grand potential of the form

$M=1-\kappa Q$2

with $M=1-\kappa Q$3 and $M=1-\kappa Q$4 built from connected Yang–Mills expectation values of Polyakov-loop operators (Tohme et al., 13 Aug 2025).

A notable consequence is the leading-order prediction for quark-number susceptibility ratios in heavy-quark QCD. In the deconfined phase, the leading sector is $M=1-\kappa Q$5, and the paper reports

$M=1-\kappa Q$6

whereas in the confined phase, center symmetry removes $M=1-\kappa Q$7 and $M=1-\kappa Q$8, so the leading sector is $M=1-\kappa Q$9, giving

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$0

(Tohme et al., 13 Aug 2025). The same HPE analysis yields an analytic formula for the quark excitation energy in the deconfined phase,

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$1

and a baryonic decomposition in the confined phase in terms of winding-three Polyakov-loop cumulants (Tohme et al., 13 Aug 2025).

3. Effective theories for the heavy-quark critical point

In finite-temperature heavy-quark QCD, HPE is not merely a formal determinant expansion; it is the basis of an effective theory in plaquette and Polyakov-loop variables. Keeping the leading HPE terms gives

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$2

with

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$3

in the formulation used for two-flavor heavy-quark QCD (Kitazawa et al., 2023). This converts the fermion determinant into an effective external field for the deconfinement order parameter, and it enables efficient pseudo-heat-bath and over-relaxation updates instead of full dynamical-fermion simulation (Kitazawa et al., 2023).

That strategy has been used to study the endpoint where the first-order deconfining transition turns into a crossover. In the $M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$4 and $M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$5 heavy-quark studies, leading-order terms are included in the simulation measure, next-to-leading-order terms are incorporated by reweighting, and finite-size scaling of the Binder cumulant of $M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$6 is used to extract the critical point (Kanaya et al., 2022, Kitazawa et al., 2023). For $M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$7, one cited NLO determination is

$M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$8

(Ejiri et al., 2023).

A key HPE result at finite density is that the critical line can be expressed through an effective Polyakov-loop coupling $M_{xy}(\kappa)=\delta_{xy}-\kappa B_{xy},$9. For $\kappa$0 flavors,

$\kappa$1

so the same critical $\kappa$2 corresponds to smaller $\kappa$3 as $\kappa$4 increases (Ejiri et al., 2023). The reported physical conclusion is that the first-order phase-transition region in the heavy-quark region becomes narrower exponentially with increasing chemical potential, and that the critical hopping parameter decreases roughly exponentially with $\kappa$5 (Ejiri et al., 2023). In this regime, the sign problem remains mild because the critical $\kappa$6 moves deeper into the heavy-quark region, and the phase fluctuation is controlled by a small effective coupling (Ejiri et al., 2023).

4. Higher orders, convergence, and algorithmic reformulations

A recurrent theme in the HPE literature is the distinction between formal convergence and practical truncation error. A detailed worst-case analysis with all gauge links set to unity computes HPE coefficients to more than $\kappa$7th order and finds that the expansion converges up to

$\kappa$8

the free Wilson-fermion chiral-limit value (Wakabayashi et al., 2021). This result addresses a common objection: the issue is not that HPE has an exceptionally small convergence radius, but that the order required for accurate truncation grows with $\kappa$9 and with the critical $\kappa=\frac{1}{2m+8}$0 (Wakabayashi et al., 2021).

The same study shows that higher-order Polyakov-type loop terms are strongly correlated with the ordinary Polyakov loop on a configuration-by-configuration basis,

$\kappa=\frac{1}{2m+8}$1

and reports that this linear relation holds up to the $\kappa=\frac{1}{2m+8}$2th order for the measured ensembles (Wakabayashi et al., 2021). Finite-density work extends this observation to the complex phase,

$\kappa=\frac{1}{2m+8}$3

which permits a “high-order from low-order” strategy in which higher-order effects are absorbed into an effective Polyakov-loop coupling rather than evaluated operator by operator (Ejiri et al., 2023). For $\kappa=\frac{1}{2m+8}$4, the critical line is reported to stabilize for $\kappa=\frac{1}{2m+8}$5, while explicit calculations were carried up to $\kappa=\frac{1}{2m+8}$6 (Ejiri et al., 2023).

These empirical correlations underpin the assessment of truncation errors in critical-point studies. The heavy-quark phase-structure analysis states that LO is fairly accurate for $\kappa=\frac{1}{2m+8}$7, NLO is fairly accurate for $\kappa=\frac{1}{2m+8}$8, and still higher orders are needed for larger $\kappa=\frac{1}{2m+8}$9 (Kanaya et al., 2022). On finer lattices, the $\kappa_f=\frac{1}{2am_f+8}$0 Binder-cumulant study reports

$\kappa_f=\frac{1}{2am_f+8}$1

after incorporating yet higher-order HPE contributions, showing that higher-order corrections remain quantitatively relevant even when LO+NLO captures the basic physics (Kitazawa et al., 2023).

A separate line of work reformulates HPE to all orders while preserving the full Yang–Mills action exactly. One formulation is the direct $\kappa_f=\frac{1}{2am_f+8}$2-expansion,

$\kappa_f=\frac{1}{2am_f+8}$3

and a second is the $\kappa_f=\frac{1}{2am_f+8}$4-expansion,

$\kappa_f=\frac{1}{2am_f+8}$5

which treats temporal hopping and its $\kappa_f=\frac{1}{2am_f+8}$6-dependence analytically and exactly while expanding only spatial hopping (Aarts et al., 2014, Aarts et al., 2015). Simulations with complex Langevin show that at $\kappa_f=\frac{1}{2am_f+8}$7 both expansions approach full-QCD results at sufficiently high order, while at $\kappa_f=\frac{1}{2am_f+8}$8 the direct $\kappa_f=\frac{1}{2am_f+8}$9-expansion breaks down and the $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$0-expansion still converges, already agreeing well with full QCD by about order $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$1 (Aarts et al., 2014).

On the computational side, HPE has also been used as a UV filter in Hybrid Monte Carlo. For two degenerate Wilson fermions, the leading nonvanishing contribution $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$2 is a plaquette term and gives

$$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$3

while numerical tests report a speed-up of roughly a factor $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$4 for $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$5-filtering and $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$6 for $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$7-filtering (Hasenbusch, 2018). More recently, explicit high-order evaluation of $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$8, $$\log \det M=\mathrm{Tr}\log(1-\kappa Q) =-\sum_{n=1}^{\infty}\frac{\kappa^n}{n}\,\mathrm{Tr}\,Q^n,$$9, and $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$0 terms in $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$1 has been made practical with trie-based algorithms, with computational costs of approximately $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$2, $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$3, and $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$4 times that of a single staple evaluation, respectively (Kitazawa et al., 30 Jun 2026).

5. Uses beyond heavy-quark lattice QCD

Although HPE is most closely associated with heavy-quark lattice QCD, the same expansion principle appears in other gauge and many-body systems. In the slave-fermion $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$5-$$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$6 model for underdoped cuprates, fermionic holons are integrated out by a hopping expansion in the holon hopping amplitude. The effective small parameter is stated to be $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$7, and the leading induced term is

$$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$8

This generates an effective bosonic lattice gauge theory with emergent link fields $$\ln \det M(\kappa_f)=-\sum_{n=1}^{\infty}\frac{\kappa_f^n}{n}\,\mathrm{Tr}[B^n].$$9, $\kappa=0$00, and $\kappa=0$01, used to analyze antiferromagnetic, metal–insulator, and superconducting transitions (Shimizu et al., 2010).

In $\kappa=0$02-dimensional $\kappa=0$03 lattice gauge theory with matter, HPE is performed on the transfer matrix in powers of

$\kappa=0$04

The ground state is found to contain local mesons at $\kappa=0$05, nearest-neighbor mesons crossing a cut at $\kappa=0$06, and longer strings at higher orders (1705.01549). The entanglement analysis based on that HPE shows that Shannon-sector entropy and color entanglement appear at $\kappa=0$07, while the Bell-pair contribution first appears at $\kappa=0$08 in the wavefunction, giving an entropy contribution

$\kappa=0$09

(1705.01549).

In the spin-1 Bose–Hubbard model with antiferromagnetic interaction, the strong-coupling expansion in the hopping amplitude $\kappa=0$10 is carried through third order to determine Mott-state and defect energies. The paper concludes that the Mott insulator phase is considerably more stable against the superfluid phase when filling with an even number of bosons than when filling with an odd number of bosons, reflecting the role of on-site singlet formation (Kimura, 2012).

At a more formal level, the Functional Renormalization Group has been used to reinterpret the hopping expansion as a linked-cluster expansion for the Legendre effective action,

$\kappa=0$11

with the critical hopping parameter identified with the finite radius of convergence of the susceptibilities and with the unstable manifold of a Gaussian or non-Gaussian fixed point of the FRG flow (Banerjee, 2018). This suggests a direct bridge between HPE, linked-cluster graph rules, and nonperturbative renormalization-group resummations.

6. Scope, limitations, and recurring misconceptions

A persistent misconception is that HPE is only a low-order heavy-quark approximation with no systematic extension. The all-orders $\kappa=0$12- and $\kappa=0$13-expansions, the explicit high-order convergence studies, and the algorithmic evaluation of $\kappa=0$14, $\kappa=0$15, and $\kappa=0$16 terms show that the method can be systematically improved to high order (Aarts et al., 2014, Wakabayashi et al., 2021, Kitazawa et al., 30 Jun 2026). A second misconception is that the existence of many higher-order loops automatically destroys predictivity; the heavy-quark critical-point literature instead reports strong configuration-by-configuration correlations between higher-order Polyakov-type terms and the ordinary Polyakov loop, which permit an effective one-coupling description over the regime studied (Ejiri et al., 2023).

The principal limitation is unchanged across applications: HPE is controlled only when the hopping parameter is sufficiently small. In lattice QCD, this means the heavy-quark region; once $\kappa=0$17 is no longer small, neglected higher-order and non-Polyakov-loop structures become important (Ejiri et al., 2023). The required truncation order also grows with $\kappa=0$18, so low-order results that are reliable at $\kappa=0$19 do not automatically remain reliable on finer lattices (Kanaya et al., 2022). In finite density, higher winding sectors $\kappa=0$20 can eventually matter because of the $\kappa=0$21 enhancement, even though they are numerically small in the parameter regions emphasized in the heavy-quark critical-point studies (Wakabayashi et al., 2021).

Another recurring issue concerns the sign problem. The finite-density heavy-quark studies do not claim that HPE removes the sign problem universally; rather, they report that along the heavy-quark critical line the sign problem does not become serious even when the density increases, because $\kappa=0$22 decreases exponentially and the phase fluctuation remains small (Ejiri et al., 2023). This is a regime statement, not a general theorem.

Taken together, the literature presents HPE as a family of controlled expansions around a static or atomic limit, organized by closed trajectories in the hopping matrix. In heavy-quark QCD it reduces the fermion determinant to an effective plaquette–Polyakov-loop theory, makes the $\kappa=0$23-dependence explicit through temporal winding, and supports critical-point, thermodynamic, and algorithmic analyses (Ejiri et al., 2023, Tohme et al., 13 Aug 2025). In broader lattice field theory and condensed-matter contexts, it serves the same structural purpose: integrating out mobile degrees of freedom in powers of their hopping and replacing them with effective local or gauge-invariant operators (Shimizu et al., 2010, 1705.01549, Kimura, 2012).