Dynamical Bosonization: Methods & Insights

• Dynamical bosonization is a technique that reformulates time-dependent quantum systems by mapping fermionic, anyonic, and supersymmetric dynamics onto bosonic variables.
• It leverages methods such as exact operator dualities in 1+1 dimensions, tensor-network mappings in 2D lattices, and semiclassical approximations for collective modes.
• This approach unifies gauge invariance, topological features, and response function analyses, offering practical insights for renormalization and strongly correlated systems.

Dynamical bosonization is used in recent literature for constructions in which bosonic variables encode time-dependent fermionic, anyonic, or supersymmetric dynamics rather than only static spectra or formal operator correspondences. In the works grouped under this label, the dynamical content can mean exact Heisenberg-equation rewriting in $1+1$ dimensions, exact lattice dualities acting on quantum states, rigorous approximation of many-body evolution by collective bosonic modes, scale-dependent field redefinitions along an RG flow, phase-space or Fermi-surface bosonizations for response functions, or asymptotic expansion phenomena in which an interacting system acquires a bosonic momentum distribution (Ha, 2015, Shukla et al., 2019, Benedikter et al., 2021, Daviet et al., 2021, Mantilla et al., 2020, Patu, 5 Sep 2025).

1. Meanings and scope of the term

In exact $1+1$-dimensional operator bosonization, dynamical bosonization refers to rewriting the fermion equations of motion, current algebra, stress tensor, and interacting dynamics in terms of a real scalar boson. A concise formulation is the current-boson identity

$$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$

together with the chiral operator formulas $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$; in this sense bosonization is an exact operator/canonical reformulation of two-dimensional fermion dynamics, including the massless and massive Thirring models and their relation to free bosons and sine-Gordon theory (Ha, 2015).

A distinct classical meaning appears in supersymmetric integrable systems. For the $\mathcal N=1$ supersymmetric KdV equation, bosonization is implemented by expanding fermionic fields in a finite set of external Grassmann parameters and replacing them by bosonic coefficient functions. The result is a triangular hierarchy consisting of one KdV equation plus linear equations for additional bosonic fields; this is exact within the chosen finite Grassmann-parameter ansatz, but it is not a quantum operator duality (Gao et al., 2011).

Another usage is explicitly finite-frequency and full-band. “Microscopic bosonization of band structures” rewrites the exact kinetic energy of an arbitrary monotonic one-dimensional band in the energy representation, where it becomes linear in energy and can be bosonized exactly. The resulting framework was used to compute real-time correlators and finite-frequency x-ray response from the Fermi level down to the band bottom, rather than only in an infrared Luttinger-liquid regime (1705.01280).

Functional-renormalization-group work uses the phrase in a different technical sense. “Flowing bosonization” keeps the density field $\varphi$ fixed but dynamically redefines the phase field $\vartheta$ along the RG flow so that $\varphi$ and the running phase variable remain conjugate at every scale. Here the issue is not fermion-to-boson transmutation but preservation of the canonical phase-density structure of bosonization itself (Daviet et al., 2021).

2. Exact operator, state, and lattice formulations

A central higher-dimensional development is the tensor-network realization of exact $2$D lattice bosonization. “A tensor network approach to 2D bosonization” constructs a tensor network operator $D$ implementing an exact duality between spinless complex fermions on triangular faces and spin-$1+1$0 bosonic degrees of freedom on edges, with the intertwining relation

$1+1$1

Because the map acts directly on quantum states, it gives an explicit algorithm for bosonizing fermionic tensor-network states, especially fPEPS, into bosonic PEPS. A key ingredient is the choice of spin structure, encoded in bonds of the bosonized network, which allows the construction to work on arbitrary triangulations of orientable $1+1$2D manifolds (Shukla et al., 2019).

A different exact lattice scheme is developed for Majorana systems. “Bosonization of Majorana modes and edge states” maps the even algebra of local Majorana bilinears to commuting on-site generalized-spin variables subject to loop constraints. Exactness requires the parity-matching condition

$1+1$3

When this condition fails, the bosonized theory necessarily contains additional fermionic excitations; on lattices with boundary, such mismatches produce explicit boundary Majorana modes. The construction preserves locality of Hamiltonian dynamics and is therefore a higher-dimensional alternative to Jordan–Wigner strings, but it is exact only for the even operator algebra and one fixed fermion-parity sector (Bochniak et al., 2021).

These exact formulations are narrower than a universal fermion-boson equivalence. In the tensor-network construction the duality is defined on fermion-parity-even operators and odd states are annihilated, while in the Majorana construction the physical Hilbert space is a constrained bosonic sector and full odd operators are not locally represented (Shukla et al., 2019, Bochniak et al., 2021).

3. Collective dynamics and asymptotic bosonic behavior

A rigorous many-body version of dynamical bosonization was established for a homogeneous spinless Fermi gas in three dimensions in the coupled semiclassical and mean-field scaling regime. The fermionic dynamics are formulated in Fock space, and suitably normalized collective particle-hole pair operators localized in patches near the Fermi surface satisfy approximate bosonic commutation relations on states with few excitations. For a specific class of initial data built from finitely many collective particle-hole modes, the many-body Schrödinger evolution is approximated in Fock-space norm by a quasifree bosonic evolution of those collective excitations (Benedikter et al., 2021).

The mechanism is intrinsically dynamical. The bosonic modes are indexed by momentum transfer $1+1$4 and patch label $1+1$5, and the patch occupation numbers are large enough that Pauli blocking is suppressed and approximate CCR emerge. This is not an exact reformulation of the full fermionic theory; it is a quantitative state-dependent approximation for low-energy collective excitations above the Fermi sea (Benedikter et al., 2021).

A very different asymptotic phenomenon appears for strongly interacting one-dimensional $1+1$6-wave anyons. In the strong-attraction limit the ground-state wavefunction is a free-boson product state multiplied by an anyonic phase factor, which yields a complex one-particle reduced density matrix and an asymmetric momentum distribution. In the homogeneous case the zero-temperature momentum distribution is a shifted Lorentzian with tails $1+1$7, while after release from a harmonic trap the long-time momentum distribution converges to that of free bosons in the same trap: $1+1$8 The derivation uses the exact scaling form of the time-dependent reduced density matrix and a stationary-phase argument showing that only the diagonal, bosonic density profile contributes asymptotically (Patu, 5 Sep 2025).

This asymptotic “dynamical bosonization” does not mean that statistics are lost at the level of the full many-body state. The anyonic phase still controls equilibrium off-diagonal correlations and finite-time observables; the bosonic limit applies to the long-time momentum distribution after harmonic expansion (Patu, 5 Sep 2025).

4. Response functions, phase-space hydrodynamics, and critical modes

One important modern direction bosonizes dynamical response rather than static operator algebras. For the $1+1$9 neutral continuum of $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$0D Dirac fermions, the low-energy Hilbert space is projected to singly occupied momentum sites, represented as spins, and then bosonized by a Holstein–Primakoff construction. The resulting boson bilinear Hamiltonian is exactly equivalent to the infinite ladder resummation of the Kadanoff–Baym particle-hole propagator built from self-consistent Hartree–Fock Green’s functions. This makes bosonization a non-perturbative response theory for optical conductivity; at weak coupling it reproduces the known perturbative coefficient $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$1, and numerically the conductivity stays close to the noninteracting value even up to $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$2 (Mantilla et al., 2020).

Higher-dimensional Fermi-surface bosonization in a weak magnetic field leads to a covariant Schwinger algebra for patch density fluctuations,

$$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$3

so the bosons become time-dependent deformations of the Fermi surface transported around the Fermi surface by cyclotron motion. The bosonized theory reproduces fermion propagators, particle-hole correlators, de Haas–van Alphen oscillations, and oscillatory corrections to the Landau zero-sound mode. Landau-level poles appear in the bosonic propagator through the global periodicity around the Fermi surface rather than by explicit Landau-level projection (Barci et al., 2018).

A related but more geometric formalism is multi-orbital coadjoint-orbit bosonization. There the bosonic variables are single-particle distribution functions on phase space, the action is written with Wigner functions and Moyal products, and projection to a single band generates Berry-connection and Berry-curvature corrections automatically. The weak-field projected dynamics become the collisionless Boltzmann equation with Berry curvature, while the compact zero-mode sector produces a topological term

$$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$4

that shifts de Haas–van Alphen oscillations by the Berry phase $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$5. The same framework argues that in interacting systems the relevant phase shift is controlled by the static anomalous Hall conductance, $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$6 (Ye et al., 2024).

Bosonization has also been used to formulate critical dynamics directly in the language of Fermi-surface modes. In the high-dimensional bosonization treatment of the $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$7D Pomeranchuk transition, the softening of an exact bosonized eigenmode $$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$8 yields the critical action

$$j^\mu(x)=\frac{1}{\sqrt{\pi}}\epsilon^{\mu\nu}\partial_\nu\phi(x),$$9

At criticality the propagator scales as $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$0, giving $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$1, upper critical dimension $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$2, and a marginally irrelevant quartic gradient interaction. This sharply contrasts with Hertz–Millis descriptions in which the critical mode is overdamped with $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$3 (Hu et al., 3 Nov 2025).

5. Flowing and partial bosonization in renormalization-group and correlated-electron methods

In the nonperturbative FRG treatment of a Luttinger liquid in a periodic potential, the derivative-expansion ansatz for the effective action generates a $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$4 term while gauge invariance fixes the coefficients attached to the original $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$5-field. If one keeps the microscopic $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$6 fixed, the $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$7-sector reproduces the sine-Gordon flow but the infrared action does not take the standard Luttinger-liquid form. The remedy is a scale-dependent field transformation

$\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$8

chosen so that the transformed phase field remains canonically conjugate to $\psi_{R/L}\sim e^{\pm i\sqrt{4\pi}\phi_{R/L}}$9 at every scale. Only with this running reparametrization does one recover the standard low-energy form controlled by the renormalized Luttinger parameter $\mathcal N=1$0 and sound velocity $\mathcal N=1$1 (Daviet et al., 2021).

Correlated-electron diagrammatics uses “partial bosonization” in yet another sense. In the DMFT-based treatment of the extended Hubbard model, only selected collective particle-hole channels are bosonized, yielding dynamical screened interactions $\mathcal N=1$2 and dynamical local fermion-boson vertices $\mathcal N=1$3. The central technical issue is Fierz ambiguity; the proposed consistent choice

$\mathcal N=1$4

suppresses the omitted boson-irreducible pieces of the local four-point vertex and allows charge and spin channels to be treated simultaneously. Here “dynamical” refers to frequency-dependent screening and propagating collective modes, not to RG-scale rebosonization (Stepanov et al., 2019).

These two programs are often conflated but are conceptually distinct. Flowing bosonization in FRG changes the field basis along the scale flow of an already bosonic low-energy theory, whereas DMFT-based partial bosonization introduces dynamical bosonic propagators for selected collective channels while retaining an effective fermion-boson theory rather than a pure bosonic description (Daviet et al., 2021, Stepanov et al., 2019).

6. Gauge fields, topology, and conceptual limits

Gauge backgrounds and topological sectors impose additional consistency conditions on bosonization. In $\mathcal N=1$5 dimensions, bosonization with a background $\mathcal N=1$6 gauge field requires more than the local current map $\mathcal N=1$7. The consistent torus action includes a $\mathcal N=1$8 coupling together with winding-holonomy terms so that gauge invariance, global chiral anomaly matching, and fermion-boson operator correspondence are all satisfied. In this framework, fermionic flux insertion is equivalent to bosonic background charge insertion, and the vanishing of the Dirac partition function in nonzero flux sectors is reproduced by bosonic neutrality conditions (Yao et al., 2019).

A dynamical $\mathcal N=1$9D bosonization duality can also be obtained from a supersymmetric parent theory. By starting from $\varphi$0D mirror symmetry, deriving an $\varphi$1 chiral mirror pair, and then breaking supersymmetry with a controlled D-term deformation, one is led to the bosonization relation between a free Dirac fermion and scalar QED$\varphi$2 with Chern–Simons and BF couplings. In that derivation the bosonic critical field is fixed indirectly by matching phase structure and topological response across the dual pair (Kachru et al., 2016).

For topological-insulator surfaces, the bosonic description involves bulk BF theory together with a gapless surface loop model. Fermionic surface operators appear as Wilson-line composites of bosonic hydrodynamic gauge fields, and gauging the characteristic anomalous one-form $\varphi$3 symmetry produces fermion parity $\varphi$4, the spin sectors, and time-reversal symmetry obeying $\varphi$5. The construction is explicitly dynamical because the loop model carries conformal, nonlocal surface dynamics rather than only topological response (Cappelli et al., 2024).

The literature therefore draws sharp boundaries around what each construction achieves. Exactness may hold only for parity-even algebras or constrained sectors, as in the tensor-network and Majorana lattice dualities; rigorous bosonic evolution may be limited to special low-energy initial states; and response-theoretic bosonizations may be exact only relative to a conserving approximation such as self-consistent Hartree–Fock plus Kadanoff–Baym ladders (Shukla et al., 2019, Bochniak et al., 2021, Benedikter et al., 2021, Mantilla et al., 2020). A plausible implication is that “dynamical bosonization” is best understood not as a single universal technique but as a family of bosonic reformulations whose common feature is direct control over dynamics—real time, frequency dependence, RG flow, or topological transport—within clearly specified operator, state, or effective-field-theory domains.