Heavy-Mass EFT: Scale Separation & Methods

• Heavy-mass EFT is a framework that separates heavy scales from low-energy dynamics using matching and controlled expansions.
• It integrates out heavy fields or reformulates them via velocity labels, resulting in local operators suppressed by inverse powers of the heavy mass.
• The approach underpins precise predictions across neutrino physics, QCD, gravitational scattering, and inflationary models through its versatile formulations.

Heavy-mass effective field theory denotes a family of EFT constructions built around a hierarchy between a heavy scale and the kinematic or dynamical scales of interest. In the simplest Wilsonian setting, heavy degrees of freedom are integrated out and replaced by local operators suppressed by inverse powers of the heavy mass. In HQET-like formulations, by contrast, the heavy field is retained but rewritten in terms of a fixed velocity label and residual momentum. More recent work extends the same scale-separation logic to worldline/background EFTs for extreme mass ratios, nonrelativistic Born–Oppenheimer descriptions of doubly heavy hadrons, SCET frameworks with finite masses or mass modes, and inflationary EFTs with Hubble-scale extra fields (Cohen, 2019).

1. Scale separation and common structure

A unifying premise is the presence of separated scales, typically written as $m \ll M$, with a power-counting parameter $\lambda \sim m/M$. The EFT is obtained by matching a UV theory onto a lower-energy theory at a scale $\mu_M \sim M$, then evolving Wilson coefficients to lower scales by RG equations. In this formulation, heavy particles cannot generate low-energy nonlocal singularities, so their effects appear analytically in external momenta and light masses and are encoded in local operators or threshold coefficients (Cohen, 2019).

The detailed organization depends on the problem. Some theories use a canonical-dimension expansion in $1/M$, others a chiral expansion, others an inverse-heavy-mass expansion around a fixed velocity $v^\mu$, and others a mass-ratio expansion such as $\lambda = m_L/m_H$. What remains common is matching, decoupling, and a hierarchy-controlled truncation.

Realization	Organizing parameter	Representative sources
Wilsonian decoupling EFT	$E/M$, $m/M$, $1/M$	(Borisov et al., 2023, Ribeiro et al., 2019, Cohen, 2019)
Velocity-based heavy-particle EFT	$k/m$ with $\lambda \sim m/M$0	(Brandhuber et al., 2021, 1908.10308)
Extreme-mass-ratio EFT	$\lambda \sim m/M$1	(Cheung et al., 2023, Cheung et al., 2024)
Strong-coupling two-heavy EFT	$\lambda \sim m/M$2 with BO potentials	(Soto et al., 2020, Soto et al., 2020)
Massive SCET / mass-mode EFT	scale ordering among $\lambda \sim m/M$3, $\lambda \sim m/M$4, $\lambda \sim m/M$5, $\lambda \sim m/M$6	(Kang et al., 2016, Gritschacher et al., 2013)

This diversity is not accidental. The literature uses “heavy-mass EFT” both for decoupling theories in the Appelquist–Carazzone sense and for heavy-particle theories in which the heavy degree of freedom remains explicit but kinematically reduced.

2. Integrating out heavy fields

The cleanest realization is the decoupling EFT obtained by removing heavy propagating modes and expanding in inverse powers of the heavy scale. In neutrino physics, the review of seesaw models presents this as a textbook application: heavy sterile neutrinos with masses far above the electroweak scale are integrated out, generating the unique dimension-five Weinberg operator. After electroweak symmetry breaking this yields a Majorana mass matrix $\lambda \sim m/M$7, or parametrically $\lambda \sim m/M$8. In type-I seesaw matching, the low-energy coefficient is fixed by the UV parameters through

$\lambda \sim m/M$9

with active–sterile mixing of order $\mu_M \sim M$0 (Borisov et al., 2023).

The same logic appears in electroweak resonance EFT. In the non-linear EWET/HEFT/EWChL framework, heavy vector, axial-vector, and scalar resonances are integrated out at tree level, giving low-energy constants $\mu_M \sim M$1 suppressed by inverse resonance masses. Short-distance conditions such as the two-Goldstone vector form factor and the Weinberg sum rules then translate these LECs into predictions or bounds on resonance masses. The paper reports that present constraints push the heavy resonance scale into the TeV regime, with the strongest bound $\mu_M \sim M$2 at $\mu_M \sim M$3 CL from $\mu_M \sim M$4 (Rosell et al., 2021).

A scalar heavy-light model in curved spacetime provides a particularly explicit derivation. For a light scalar $\mu_M \sim M$5 coupled cubically to a heavy scalar $\mu_M \sim M$6, integrating out $\mu_M \sim M$7 gives the exact nonlocal operator

$\mu_M \sim M$8

which is then expanded as

$\mu_M \sim M$9

At leading order this produces a local quartic light-field interaction with matching

$1/M$0

while higher-derivative and curvature corrections are suppressed by $1/M$1 (Ribeiro et al., 2019).

Inflationary EFT displays the boundary between integrating out a heavy field and retaining it. In quasi-single-field inflation, an additional scalar $1/M$2 with $1/M$3 must remain dynamical during horizon crossing, but when $1/M$4 and the time dependence is adiabatic, $1/M$5 can be integrated out. The resulting single-field EFT contains local operators such as $1/M$6 and modifies the Goldstone sound speed $1/M$7 (Noumi et al., 2012).

3. Velocity-dependent heavy-particle EFTs

A different branch keeps the heavy field explicit but reorganizes it around a large mass and fixed timelike velocity. The defining kinematics is

$1/M$8

with $1/M$9 and soft residual momentum $v^\mu$0. In the heavy-mass limit, propagators linearize to $v^\mu$1, and interactions reduce to eikonal couplings. This is the structure used in heavy-mass effective theory for tree amplitudes with two heavy particles and arbitrary gauge bosons or gravitons (Brandhuber et al., 2021).

Within that framework, a new gauge-invariant double-copy construction yields BCJ numerators written directly in terms of field strengths $v^\mu$2 and $v^\mu$3. The numerators are local in the massless channels, automatically satisfy Jacobi identities and crossing symmetry, and are constructed explicitly up to six particles. The heavy-particle poles $v^\mu$4 are physical EFT poles rather than spurious nonlocalities (Brandhuber et al., 2021).

The all-multiplicity completion of this program identifies a kinematic algebra for the pre-numerators as a quasi-shuffle Hopf algebra. For two heavy particles and $v^\mu$5 gluons, the number of terms in the pre-numerator is the Fubini number $v^\mu$6, reflecting the ordered-partition structure of the algebra. The corresponding double-copy formula generates the gravitational amplitudes relevant for black-hole scattering and gravitational-wave emission (Brandhuber et al., 2021).

Heavy Black Hole Effective Theory applies the same HQET logic to long-range gravity. With

$v^\mu$7

HBET organizes amplitudes in $v^\mu$8, or equivalently in $v^\mu$9. The heavy propagators become

$\lambda = m_L/m_H$0

and the EFT isolates the nonanalytic sectors responsible for classical post-Minkowskian dynamics. The paper uses this setup to compute the classical and leading quantum 2PM scattering amplitude of two heavy spin-$\lambda = m_L/m_H$1 particles (1908.10308).

A common limitation is explicit in this literature: velocity-based heavy-particle EFT is naturally adapted to scattering or soft exchange, but not automatically to bound-state loop kinematics. HBET, for example, develops the usual HQET pinch singularity when the heavy velocities coincide (1908.10308).

4. Background-field and extreme-mass-ratio formulations

A further development treats the infinitely heavy limit exactly as a classical background and expands in the mass ratio. The relevant parameter is

$\lambda = m_L/m_H$2

identified with the self-force expansion. At 0SF, the heavy particle moves inertially and generates a background Coulomb field or Schwarzschild metric, while the light particle follows a probe trajectory in that background. At 1SF and beyond, one restores the heavy body’s recoil by integrating out its deviation from the background worldline (Cheung et al., 2023).

In gravity, the resulting effective action is

$\lambda = m_L/m_H$3

where the recoil operator is

$\lambda = m_L/m_H$4

In electromagnetism the analogous operator is built from the heavy-particle electric field. These nonlocal worldline terms are the precise finite-mass corrections missing from a pure probe-in-background description (Cheung et al., 2023).

The formalism treats the Schwarzschild or Coulomb background not merely as a classical solution but as a resummation of infinitely many flat-space diagrams. Probe geodesics or Lorentz-force trajectories play the same role for the light body. PM or PL expansion of the background then reconstructs multiloop classical scattering integrands. The 2024 extension verifies known two-loop classical scattering results and adds new calculations for dyons and for particles coupled to extra scalar or vector fields that interact directly only with the lighter particle (Cheung et al., 2024).

An important structural point is that recoil is not optional. In both the 2023 and 2024 formulations, background-field diagrams and recoil-insertion diagrams are separately incomplete; only their sum reproduces the correct conservative dynamics and gauge-invariant answer (Cheung et al., 2023).

5. Heavy-mass EFT in QCD: hadrons, jets, and media

In heavy-hadron spectroscopy, heavy-mass EFT appears in a nonrelativistic and often strong-coupling form. For systems with two heavy quarks and arbitrary light degrees of freedom, the EFT is built from NRQCD without assuming that the heavy-heavy separation is small relative to $\lambda = m_L/m_H$5. The hierarchy is

$\lambda = m_L/m_H$6

with the adiabatic condition $\lambda = m_L/m_H$7. The retained fields are Born–Oppenheimer channel wavefunctions $\lambda = m_L/m_H$8, and the Hamiltonian is expanded as

$\lambda = m_L/m_H$9

At leading order the theory is the Born–Oppenheimer approximation; at next-to-leading order, spin- and velocity-dependent terms break heavy-quark spin symmetry and split the multiplets. The framework is stated to apply uniformly to hybrids, tetraquarks, double heavy baryons, and pentaquarks (Soto et al., 2020).

The companion application to doubly heavy baryons makes the same point more concretely. It uses lattice-QCD static energies as Born–Oppenheimer potentials and derives spectra and spin-averaged mass relations for $E/M$0 systems beyond the compact diquark approximation. The analysis emphasizes that realistic $E/M$1 and $E/M$2 systems need not satisfy $E/M$3; the EFT remains valid provided the heavy pair is nonrelativistic and adiabatically slow (Soto et al., 2020).

In jet physics and medium-induced branching, the appropriate heavy-mass EFT is not HQET but massive SCET. The $E/M$4 construction combines finite heavy-quark masses with Glauber gluons. Collinear momenta scale as $E/M$5, Glauber exchange as $E/M$6, and the heavy-quark mass is taken to scale as $E/M$7. This produces vacuum and in-medium splitting functions for

$E/M$8

and a framework for incorporating medium-induced full splitting functions consistently with NLO calculations of inclusive hadron production (Kang et al., 2016).

A related but distinct SCET construction treats secondary heavy-quark production in jets through mass modes. For thrust in the dijet limit, the hierarchy among $E/M$9, $m/M$0, $m/M$1, and the heavy mass $m/M$2 requires four different EFT descriptions. Massive collinear and soft mass modes are separated coherently from the usual massless collinear and ultrasoft modes, and threshold matching continuously interpolates between the decoupling limit $m/M$3 and the light-mass limit $m/M$4. The paper explicitly relates this framework to a variable fermion number scheme (Gritschacher et al., 2013).

6. Validity, nonlocality, and formal infrastructure

Heavy-mass EFT truncations often generate higher derivatives or nonlocal structures, so their regime of validity is not exhausted by the naive condition $m/M$5. A rigorous analysis of a classical heavy-light scalar model shows that truncating the EFT equations of motion at finite order can produce spurious higher-derivative branches. The physically relevant EFT solutions are instead those whose fields and derivatives remain uniformly bounded as $m/M$6. For a large class of “prepared” initial data, solutions of the UV theory are approximated by EFT solutions on fixed time intervals; for generic bounded-energy data with fast heavy-mode oscillations, the standard EFT fails pointwise but a modified EFT reproduces the solution in an averaged sense (Reall et al., 2021).

This clarifies a recurrent misconception. Several papers explicitly distinguish Appelquist–Carazzone decoupling EFT from heavy-particle EFT in the HQET sense. The curved-space scalar study states that it is not “heavy-particle EFT” in the HQET sense but rather a heavy-mass EFT in which a field of mass $m/M$7 is integrated out when $m/M$8 (Ribeiro et al., 2019). The heavy-ion SCET paper likewise emphasizes that massive SCET is not HQET because the relevant heavy quark is energetic and collinear, with $m/M$9, not nearly static (Kang et al., 2016).

The mathematical infrastructure has also become specialized. Two-loop heavy-particle diagrams with at least one heavy line and an additional ordinary mass scale can be reduced to master integrals and solved analytically using IBP reduction, differential equations, and dimensional recurrence relations. For heavy-heavy and heavy-light vertex sectors, the systems can be transformed to canonical $1/M$0-form and solved in terms of multiple polylogarithms. For the off-shell heavy self-energy sector, elliptic subgraphs obstruct a full $1/M$1-form reduction, so the system is instead brought to $1/M$2 form and solved by a Frobenius expansion around small off-shell energy (Assi et al., 2021).

Taken together, these constructions suggest that heavy-mass effective field theory is best understood not as a single canonical formalism but as a family of scale-separation frameworks. What unifies them is the treatment of heavy dynamics through matching, threshold organization, and controlled expansions in $1/M$3, $1/M$4, or $1/M$5; what differentiates them is whether the heavy sector is integrated out, retained as a velocity-labeled mode, or encoded as a classical background with recoil corrections.