Local Renormalized Functional Theory

• Local Renormalized Functional Theory is a set of methods that organizes functional renormalization by extracting local or quasi-local sectors with renormalized coefficients to encode short-distance fluctuation effects.
• This approach unifies different frameworks—such as Wilsonian effective actions, functional RG flows, and density-functional as well as DMFT methods—by controlling ultraviolet divergences via local operations.
• It enables precise, scalable computations and predictions in systems like near-critical fluids and quantum field theories by renormalizing local coefficients and absorbing nonlocal remnants into controlled remainders.

Searching arXiv for papers directly relevant to “Local Renormalized Functional Theory” and closely related formulations. Local Renormalized Functional Theory designates a cluster of renormalization frameworks in which the primary object is a local or quasi-local functional whose coefficients, kernels, or counterterms already encode short-distance fluctuation effects. The phrase is used explicitly in the theory of near-critical fluids confined between parallel plates, where a local free-energy density is equipped with locally renormalized coefficients so as to reproduce nonclassical critical behavior (Okamoto et al., 2011). Closely related constructions appear in several other settings: Wilsonian effective functionals split into local relevant or marginal terms plus irrelevant remainders (Mastropietro, 2023), effective actions expanded in local operators and evolved by exact RG equations (Gurau et al., 2014), flowing Wess–Zumino actions linking functional and local renormalization groups (Codello et al., 2015), finite functional Callan–Symanzik flows with local or quasi-local counterterm functionals (Braun et al., 2022), renormalized adiabatic kernels derived from local TDDFT kernels (Olsen et al., 2014), local renormalized quasiparticle descriptions in DMFT (Hewson, 2016), renormalized Hamiltonian nets built from local algebras (Miyao, 2018), and renormalized local index formulas on Lie manifolds (Bohlen et al., 2016). This suggests that the term is best understood not as a single universally standardized formalism but as a family of locality-preserving renormalization strategies.

1. Terminological scope and defining structure

In its broadest technical sense, the subject concerns renormalization performed at the level of a functional rather than solely at the level of individual amplitudes. Mastropietro’s review formulates this most explicitly in Wilsonian language: after integrating ultraviolet modes, one obtains a sequence of effective functionals $V_h$ that decompose into a local relevant or marginal part and an irrelevant remainder, schematically

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$

with the local part carrying the running couplings and the remainder containing the nonlocal irrelevant terms (Mastropietro, 2023). The advanced overview of renormalization expresses the same idea through a theory-space expansion of the effective action,

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$

so that renormalization becomes the scale dependence of local operator coefficients inside a functional object $\Gamma_\mu$ (Gurau et al., 2014).

In this vocabulary, “local” does not necessarily mean strictly ultralocal. The same sources repeatedly distinguish local bare actions, quasi-local effective actions, and genuinely nonlocal remainders. In the perturbative multiscale formulation, high subgraphs are quasi-local because they “look almost local when seen through their external edges”; in the Wilsonian formulation, marginal and relevant terms can be rewritten as local monomials, while the remaining terms are irrelevant and generally nonlocal (Gurau et al., 2014). This suggests a working definition: a local renormalized functional theory is one in which renormalization is organized by extracting, evolving, or modifying the local or quasi-local sector of a functional, with the nonlocal sector either suppressed, controlled, or generated in a secondary role.

A persistent misconception is that such theories must belong to a single doctrinal lineage. The literature instead shows several non-equivalent realizations. In some papers, locality is encoded by local operators and running couplings in a Wilsonian effective action; in others, by local gradient terms with renormalized coefficients; in still others, by local counterterm flows, local algebras, or local kernel constructions. The unifying feature is structural rather than canonical.

2. Renormalized local functional theory for near-critical fluids

The most literal use of the term appears in the theory of a near-critical fluid confined between two parallel plates in equilibrium with a bulk reservoir, developed for Ising-class fluids including one-component fluids near a gas–liquid critical point and binary mixtures near demixing (Okamoto et al., 2011). The local singular free-energy density is written as

$$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$

with total singular free energy

$$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$

Here $\psi$ is the order parameter, $\mu_\infty$ is the reservoir chemical potential difference, and $f_s$ is the surface free-energy density.

At the bulk critical temperature $T=T_c$, the singular bulk terms are chosen as power laws with nonclassical exponents,

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$0

and for $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$1 the paper uses

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$2

The corresponding local correlation length is

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$3

The theory is therefore local in the sense that it is expressed as a spatial integral of $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$4 and $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$5, but renormalized in the sense that the coefficients already incorporate the exact critical exponents rather than mean-field ones.

Away from $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$6, the theory introduces a local variable $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$7 measuring the local “distance” from criticality and defines the coefficients

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$8

with $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$9 fixed implicitly by

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$0

This is the central renormalization mechanism of the model: the functional remains local, but its coefficients are locally renormalized by the critical scaling structure. The authors then derive the Euler–Lagrange equation

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$1

together with the first integral

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$2

which determines the slit profile.

The formulation yields universal scaling in terms of the slit variables

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$3

with $$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$4. In the geometry of two identical strongly adsorbing walls, the theory predicts a line of first-order capillary condensation outside the bulk coexistence curve when the component favored by the walls is slightly poor in the reservoir. The transition terminates at a capillary critical point

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$5

equivalently

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$6

The theory is especially notable for its Casimir-force predictions. The force amplitude is defined by

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$7

and the excess grand-potential amplitude by

$$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$8

At $$\Gamma_\mu[\varphi]=\sum [\mathcal Z_\mu]^{n/2}\frac{1}{n!}\lambda_\mu^{(n,p,\sigma)}\,\mathcal O^{(n,p,\sigma)}(\varphi),$$9, the critical-point value is estimated as $\Gamma_\mu$0. At $\Gamma_\mu$1, $\Gamma_\mu$2 is maximized at $\Gamma_\mu$3, where

$\Gamma_\mu$4

while $\Gamma_\mu$5 is maximized near $\Gamma_\mu$6, with

$\Gamma_\mu$7

More generally, the paper emphasizes that the Casimir amplitudes can be larger than their critical-point values by $\Gamma_\mu$8–$\Gamma_\mu$9 times between the capillary-condensation line and the bulk coexistence curve. The framework remains approximate: it is one-dimensional, neglects lateral fluctuations, yields mean-field behavior at the capillary critical point, and uses a simple continuation inside the coexistence curve. Even so, it is the clearest explicit example of a renormalized local functional theory in the literature surveyed here.

3. Wilsonian locality and renormalized effective functionals

In quantum field theory, the most systematic foundation for the subject is Wilsonian. Mastropietro’s review formulates renormalization in terms of regularized Euclidean functional integrals, shellwise integration, and scale-dependent effective potentials $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$0 (Mastropietro, 2023). Starting from a regularized theory, one repeatedly integrates out ultraviolet shells and rewrites the result as a new effective functional. At each scale, the dangerous part is extracted into finitely many local relevant or marginal terms, while the rest is left in an irrelevant nonlocal remainder. The paper states this directly: it is convenient “to write the marginal or relevant terms as local ones, that is with all the fields computed in the same point; this is always possible at the price of extracting enough irrelevant terms.”

The perturbative side of the same logic is encoded by power counting. For $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$1 theory in $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$2,

$$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$3

and in QED,

$$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$4

A theory is perturbatively renormalizable when $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$5 is independent of perturbative order and $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$6 only for graphs corresponding to monomials already present in the bare action (Mastropietro, 2023). The advanced overview presents the same content in multiscale language: the propagator is sliced into scales, high subgraphs form an inclusion forest, and localization operators $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$7 contract divergent high subgraphs to local counterterms (Gurau et al., 2014). In $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$8, only $$f_{\rm loc}=f(\psi)+\frac{1}{2}C(\psi)|\nabla\psi|^2,$$9, $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$0, and $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$1 need local counterterms, while the effective expansion introduces running local couplings $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$2, $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$3, and $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$4.

The functional viewpoint is decisive. The advanced overview defines the quantum effective action by the Legendre transform of $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$5,

$$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$6

and emphasizes that $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$7 admits a semi-local expansion in effective vertices rather than only strictly local bare terms (Gurau et al., 2014). This furnishes a precise sense in which local renormalized functional theory is broader than bare-action renormalization: the renormalized theory is encoded in an effective functional whose organization by local operators, running couplings, and quasi-local derivative expansions reflects the locality of ultraviolet divergences.

A recurrent misunderstanding is that locality in these constructions excludes nonlocality altogether. The Wilsonian sources instead describe a hierarchy: local bare interactions, local relevant or marginal parts of the renormalized effective functional, quasi-local derivative expansions at finite scale, and nonlocal irrelevant tails. The theory is “local” because renormalization is controlled by local structures, not because the full effective functional is everywhere strictly local.

4. Functional renormalization-group realizations

The functional renormalization group translates the Wilsonian picture into exact flow equations for scale-dependent effective actions. The advanced overview writes the effective average action $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$8 as the Legendre transform of a $$F=\int d{\bf r}\,[f_{\rm loc}-\mu_\infty\psi]+\int dS\,f_s(\psi).$$9-dependent generating functional with regulator term $\psi$0, and derives the Wetterich equation

$\psi$1

with $\psi$2 (Gurau et al., 2014). In practice, FRG calculations use truncations of $\psi$3 that are local or quasi-local, typically local-potential or derivative expansions. Within this framework, local renormalized functional theory becomes an explicit computational scheme rather than only a structural principle.

A purely fermionic realization is given by the local potential approximation for fermionic field theories. There the effective action is truncated to

$\psi$4

with $\psi$5 an arbitrary function of local fermionic invariants, and the difficult fermionic Hessian in the Wetterich equation is factorized so that LPA flows can be derived directly in terms of fermionic composites, without auxiliary bosonic fields (Jakovac et al., 2013). The construction is worked out explicitly for the $\psi$6-flavor Gross–Neveu model and the one-flavor chiral Nambu–Jona-Lasinio model. This is a genuinely local functional RG treatment, because the running object is a local effective potential $\psi$7 of composite operators.

The relation between FRG and the local renormalization group is made explicit in two dimensions by the scale-dependent Wess–Zumino action

$\psi$8

In this setting, the off-critical $\psi$9-function is

$\mu_\infty$0

and the local RG Weyl consistency condition becomes

$\mu_\infty$1

with an explicit FRG representation for the Zamolodchikov–Osborn metric $\mu_\infty$2 (Codello et al., 2015). The significance is conceptual as well as computational: local anomaly data and exact RG flows can be encoded in the same running functional.

Renormalized spectral flows generalize the FRG construction to regulators that do not themselves make the flow ultraviolet finite. The paper derives a finite functional flow

$\mu_\infty$3

where $\mu_\infty$4 is a flowing BPHZ-type counterterm action fixed by renormalization conditions imposed along the flow (Braun et al., 2022). In the Callan–Symanzik limit this becomes a finite functional CS equation, allowing Lorentz-invariant spectral flows in real and imaginary time. The counterterm functional is local for approximations with local vertices that reduce to the classical ones at large momentum, and more generally quasi-local with uniquely fixed momentum dependence. This is a local renormalized functional theory in the exact-flow sense: renormalization is implemented through a functional subtraction that co-moves with the scale $\mu_\infty$5.

A different FRG strategy links local functional truncations to dimensional regularization and $\mu_\infty$6. By choosing a pseudo-regulator

$\mu_\infty$7

one can reproduce the perturbative $\mu_\infty$8 beta functions at one loop and, with sufficiently rich truncations and a careful order of limits, also at two loops (Baldazzi et al., 2020). The resulting flows are presented as nonperturbative extensions of $\mu_\infty$9, recover all the multicritical models in two dimensions, and preserve nonlinearly realized symmetries. Here the local renormalized functional is the derivative-expanded FRG action, but its scheme data are forced to coincide with dimensional regularization.

5. Local renormalization outside relativistic QFT

A conceptually close but formally different use appears in time-dependent density functional theory. The ACFDT-TDDFT work on ground-state energies begins from a local or semi-local adiabatic kernel and identifies the pathology that causes local kernels such as ALDA to fail in ACFDT applications: the high-$f_s$0 tail generates a divergent correlation hole at $f_s$1 (Olsen et al., 2014). For the homogeneous electron gas, the kernel is renormalized by a momentum-space cutoff,

$f_s$2

and generalized to any adiabatic semi-local kernel $f_s$3 through a finite-ranged real-space kernel with cutoff

$f_s$4

The appendix interprets the construction as a weighted-density generalization of a local functional, with $f_s$5 replaced by a locally averaged density $f_s$6. The local starting point is therefore preserved, but the renormalized kernel becomes effectively nonlocal and finite-ranged. Numerically, rAPBE yields a mean absolute percentage error of $f_s$7 for atomization energies of $f_s$8 small molecules and $f_s$9 for cohesive energies of $T=T_c$0 solids, versus $T=T_c$1 and $T=T_c$2 for RPA@PBE, while retaining essentially RPA-like computational scaling $T=T_c$3 and similar performance for barriers, adsorption, and graphene-on-metal systems. The authors do not use the phrase “Local Renormalized Functional Theory,” but the paper explicitly develops a renormalized local or semi-local adiabatic kernel theory within ACFDT-TDDFT.

A distinct local renormalized picture arises in DMFT for the Hubbard model. In the paramagnetic metallic phase, the self-energy is local,

$T=T_c$4

and the low-energy theory is built from renormalized quasiparticle parameters

$T=T_c$5

together with the renormalized local interaction

$T=T_c$6

and counterterm-based renormalized perturbation theory (Hewson, 2016). Even in the low-density limit there is significant renormalization of $T=T_c$7, consistent with Kanamori-type repeated scattering, and on approaching the Mott transition one finds a finite ratio for $T=T_c$8, while $T=T_c$9–$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$00. This suggests a local renormalized effective theory of correlated lattice quasiparticles: the central dynamical data are local one- and two-particle quantities, and response functions are built from repeated quasiparticle scattering with local renormalized vertices.

These examples show that the subject is not confined to high-energy QFT. In density-functional theory, the renormalized object is a kernel derived from a local functional; in DMFT, it is a low-energy local quasiparticle theory. The common pattern is the same: a local starting point is retained, but renormalization reorganizes it into a more accurate effective description.

6. Operator-algebraic and geometric extensions

The operator-algebraic version replaces local Lagrangian densities by nets of local algebras and local Hamiltonians. A renormalized Hamiltonian net consists of triples

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$01

satisfying

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$02

for every bounded region $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$03 (Miyao, 2018). The framework is built on von Neumann algebras $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$04, tensor factorizations over complementary regions, and natural cones from Tomita–Takesaki theory. Under hypotheses (A.1)–(A.5), positivity improvingness of the global renormalized semigroup is equivalent to positivity improvingness of every local cutoff semigroup. For the renormalized Nelson Hamiltonian at fixed total momentum and for the full renormalized Nelson Hamiltonian with confining potential, the resulting semigroups satisfy

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$05

and the method works directly in the Fock representation and covers the massless case. Here renormalization is local in an operator-algebraic sense: the global renormalized Hamiltonian is reconstructed from a coherent net of local pieces.

A geometric version arises in the index theory of Dirac operators on Lie manifolds. Because the heat operator is generally not trace class, the theory introduces a renormalized supertrace defined by finite-part regularization at infinity, together with Getzler-type rescaling implemented on the adiabatic deformation of an integrating Lie groupoid (Bohlen et al., 2016). For a geometric Dirac operator $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$06, the renormalized local index formula is

$$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$07

The local density is the familiar Atiyah–Singer density $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$08; renormalization changes the global pairing through the renormalized supertrace and adds the defect term $$V_h(\phi^{<h})=\mathcal L V_h(\phi^{<h})+\mathcal R V_h(\phi^{<h}),$$09. The rescaling is executed on a rescaled bundle over the adiabatic groupoid, so locality is recovered through deformation and filtration, while renormalization compensates for noncompactness or singularity at infinity.

Taken together, these operator-algebraic and geometric constructions broaden the meaning of local renormalized functional theory. The “functional” need not always be a scalar effective action. It may be a semigroup pairing on a local algebra net or a renormalized supertrace on a convolution algebra. The underlying principle remains stable: the renormalized theory is encoded by local data, while ultraviolet or asymptotic singularities are absorbed into a controlled renormalized functional structure.