Velocity Gauge: Theory & Applications

• Velocity gauge is a formulation where electromagnetic coupling is expressed via the vector potential rather than through dipole interaction, maintaining momentum-space integrity.
• It offers computational advantages in periodic and topological systems by avoiding singularities associated with the position operator while ensuring gauge equivalence.
• Practical applications require precise treatment of field-dependent currents and basis truncation to accurately model nonlinear responses and transport phenomena.

Velocity gauge denotes a class of formulations in which electromagnetic coupling is expressed through the vector potential rather than the dipole coupling $-q\,\mathbf r\cdot \mathbf E(t)$ of the length gauge. In quantum light–matter theory this typically means minimal coupling, $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$, or, in Bloch systems, a momentum-space substitution such as $\mathbf k\to \mathbf k+\mathbf A(t)$. In classical electrodynamics, the same term can also refer to the “v-gauge,” the one-parameter family defined by $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$, with Lorenz and Coulomb gauges recovered as special cases. Across atomic, molecular, solid-state, and electromagnetic applications, the central theme is representation rather than altered physics: exact formulations are gauge equivalent, while practical discrepancies arise from basis truncation, incomplete operator transformations, or omission of field-dependent current terms (Passos et al., 2017, Faisal, 2008, Giri et al., 2022).

1. Definition, scope, and formal structure

In quantum mechanics, the velocity gauge is the formulation of light–matter coupling in which the external electromagnetic field is introduced through the vector potential $\mathbf A(t)$, rather than through the scalar potential $-q\,\mathbf{E}(t)\cdot\mathbf{r}$. For crystals and other periodic systems this is often written in minimal-coupling form, and in Floquet or band-theory settings it is frequently expressed through a Peierls-type substitution $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$. This representation preserves translation symmetry or crystal momentum in ways that are often computationally advantageous in periodic solids (Passos et al., 2017, Dabiri et al., 16 Mar 2025, Xu et al., 2023).

The contrast with the length gauge is structural. In the length gauge, the perturbation is written directly as $\mathbf r\cdot\mathbf E(t)$, so position-operator matrix elements, Berry connections, or $\mathbf k$-derivatives appear explicitly. In the velocity gauge, the perturbation is instead attached to momentum, current, or band-velocity operators, and the current operator itself can acquire explicit field dependence. This distinction becomes especially important in finite-band models, nonlinear response, and truncated Hilbert spaces, where naive use of infinite-basis formulas is not generally valid (Passos et al., 2017, Ahmad et al., 29 Jun 2026).

In classical electrodynamics, the velocity gauge or v-gauge is a continuous family of gauges indexed by a parameter $v$, with condition

$H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$0

The limiting choices $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$1 and $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$2 recover the Lorenz and Coulomb gauges, respectively. This usage is conceptually distinct from the quantum-optical “velocity gauge,” but both center on vector-potential representations and gauge freedom (Giri et al., 2022).

2. Gauge equivalence and hybrid transformations

A foundational result is that the familiar velocity-gauge and length-gauge perturbation series are not different physical theories but different series representations of the same transition amplitude. A “master” perturbation expansion derived in the minimal-coupling velocity-gauge framework generates a one-parameter family of branches labeled by $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$3, with $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$4 yielding the traditional velocity-gauge series and $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$5 yielding the traditional length-gauge series. The two are therefore “equal but in a common domain of convergence (if it exists)” (Faisal, 2008).

The atomic dynamic Stark effect provides a sharper illustration. Under the “hybrid” gauge transformation studied in “Relativistic and Radiative Corrections to the Dynamic Stark Shift: Gauge Invariance and Transition Currents in the Velocity Gauge” (Jentschura et al., 2018), the scalar and vector potentials are transformed, but the wave function remains unaltered. For a laser polarized along $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$6, the leading dynamic Stark shift is written in the length gauge as

$H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$7

with $H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$8, whereas in the velocity gauge

$H(t)=H_0(\mathbf r,\mathbf p-q\mathbf A(t))$9

The $\mathbf k\to \mathbf k+\mathbf A(t)$0 contribution is the seagull term. The standard identity

$\mathbf k\to \mathbf k+\mathbf A(t)$1

shows that the leading-order dynamic Stark shift is gauge invariant even in the hybrid formulation (Jentschura et al., 2018).

The subtlety appears once one adds a perturbation $\mathbf k\to \mathbf k+\mathbf A(t)$2. In the length gauge, the first-order correction consists of six standard contributions. In the velocity gauge, the analogous six terms are not sufficient: two additional terms, $\mathbf k\to \mathbf k+\mathbf A(t)$3 and $\mathbf k\to \mathbf k+\mathbf A(t)$4, arise because the perturbation modifies the transition current. The necessary correction is

$\mathbf k\to \mathbf k+\mathbf A(t)$5

With these terms included, the corrected gauge-invariance relation is

$\mathbf k\to \mathbf k+\mathbf A(t)$6

This is the operator-level mechanism by which gauge invariance is restored in the velocity gauge when perturbative radiative or relativistic corrections are treated in the hybrid scheme (Jentschura et al., 2018).

3. Field-dependent currents, basis truncation, and gauge consistency

A persistent source of confusion is the appearance of gauge-dependent results in truncated calculations. Several works identify the same underlying mechanism: the velocity gauge remains formally consistent, but only if its full operator structure is retained. Failures are therefore typically failures of truncation or bookkeeping rather than failures of gauge equivalence itself (Passos et al., 2017, Yakovlev et al., 2017).

For crystalline solids in a truncated Bloch basis, the velocity gauge converges slowly with respect to the number of bands and can violate the adiabatic low-frequency limit. “Adiabatic corrections for velocity-gauge simulations of electron dynamics in periodic potentials” (Yakovlev et al., 2017) shows that the missing contribution from omitted high-energy bands is, to a very good approximation, an adiabatic function of the vector potential. This permits analytic corrections to the current density, built only from band energies, momentum matrix elements, and $\mathbf k\to \mathbf k+\mathbf A(t)$7, which restore the missing adiabatic response and compensate the violation of the Thomas–Reiche–Kuhn sum rule in finite-band calculations (Yakovlev et al., 2017).

A complementary finite-band construction is given in “Nonlinear optical responses of crystalline systems: Results from a velocity gauge analysis” (Passos et al., 2017). There the velocity-gauge Hamiltonian is obtained by a unitary transformation from the length-gauge form and expanded through nested commutators,

$\mathbf k\to \mathbf k+\mathbf A(t)$8

The resulting conductivity formulas are valid to any order and for any finite band model, provided the full Brillouin zone is used. The work emphasizes that one cannot take formulas derived for an infinite-band minimal-coupling Hamiltonian and naively truncate the final answer (Passos et al., 2017).

In dc and nonlinear transport, the same theme reappears as field-dependent current vertices. In the velocity-gauge density-matrix treatment of intrinsic second-order dc conductivity, the current operator follows from

$\mathbf k\to \mathbf k+\mathbf A(t)$9

and expands as

$\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$0

The second-order current therefore contains three pieces,

$\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$1

with $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$2 and $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$3 essential for gauge invariance; keeping only $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$4 produces spurious results (Ahmad et al., 29 Jun 2026). In the velocity-gauge Keldysh formalism, the analogous role is played by the Peierls contact vertices

$\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$5

whose inclusion removes the apparent low-frequency poles and makes the gauge-invariant decomposition explicit (Shibata, 21 Jun 2026).

4. Periodic solids, topology, and nonlinear transport

The velocity gauge has become particularly prominent in periodic solids because it avoids explicit position-operator ambiguities and, in many formulations, keeps the problem diagonal in crystal momentum. In topological materials this is technically important: the length-gauge semiconductor Bloch equations sample Berry connections and dipole matrix elements that can become singular or discontinuous at special points in the Brillouin zone, whereas in the velocity gauge the $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$6-channels are decoupled and the singular structure is not entered through a finite-difference derivative operator (Kim et al., 2021).

This advantage is illustrated in “Theory for all-optical responses in topological materials: the velocity gauge picture” (Kim et al., 2021). Using the Haldane model as a prototype Chern insulator, the work finds good qualitative agreement of the velocity gauge compared to the length gauge and the Time-Dependent Density Functional theory in the case of topologically trivial materials such as MoS$\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$7. For velocity gauge and length gauge, the two-band Haldane model reproduces key HHG spectra features: the selection rules for linear and circular light drivers, the linear cut-off law scaling, and the anomalous circular dichroism (Kim et al., 2021).

A different topological use appears in “Coherent Spectroscopic Probes of Topology: A Velocity-Gauge Perspective” (Bittner et al., 16 May 2025), where the full minimal-coupling Hamiltonian

$\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$8

is retained, including the $\nabla\cdot \mathbf A = -(1/v^2)\,\partial_t\phi$9 term, and nonlinear spectra are formulated through retarded current-current response functions. In the SSH model, the third-order response exhibits characteristic phase inversions and spectral asymmetries that are absent in the trivial phase. The work attributes these features to geometric changes in the Bloch eigenstates and to the role of virtual pathways in shaping the nonlinear signal (Bittner et al., 16 May 2025).

For dc nonlinear transport, gauge-consistent velocity-gauge theories sharpen the distinction between genuine physical terms and artifacts of incomplete implementation. “Length–Velocity Gauge Equivalence of Quantum Geometric Nonlinear Conductivity” (Ahmad et al., 29 Jun 2026) shows that the length gauge and velocity gauge give the same adiabatic dc response when the same retarded continuation is used for all external frequencies and when the velocity gauge current includes all field-dependent vertices. Apparent Fermi sea terms cancel in the full expression, leaving a Fermi surface quantum geometric contribution determined by the band-normalized quantum metric; consequently, a fully gapped insulator has no residual dc nonlinear Hall current in the adiabatic clean limit (Ahmad et al., 29 Jun 2026).

The complementary Keldysh derivation in “Second-order dc conductivity in the velocity-gauge Keldysh formalism” (Shibata, 21 Jun 2026) resolves the clean-limit response into four contributions with distinct lifetime scalings: $\mathbf A(t)$0 In that formulation, the intraband term is a Fermi-surface dipole of the ordinary band quantum metric, while the interband term is written as a Fermi-sea-type response involving a band-normalized quantum metric. The paper also constructs a real two-band model in which the Berry curvature and hence the BCD response vanish identically while the intraband quantum-metric dipole remains finite (Shibata, 21 Jun 2026).

5. Floquet, high-field, and real-time computational formulations

In time-periodic systems, the velocity gauge naturally accommodates drive-induced sidebands and momentum-space structure. “Velocity gauge formulation of nonlinear optical response in Floquet-driven systems” (Dabiri et al., 16 Mar 2025) starts from

$\mathbf A(t)$1

with probe coupling introduced through

$\mathbf A(t)$2

The resulting response is frequency mixed with the drive: for a probe of frequency $\mathbf A(t)$3, the first-order conductivity contains harmonics shifted by integer multiples of $\mathbf A(t)$4, and the second-order response appears at $\mathbf A(t)$5. In the Rabi-model benchmark, the velocity-gauge and length-gauge conductivities show near-perfect agreement at finite frequencies, while the remaining discrepancy is confined to the zero-frequency limit in the velocity gauge, where a spurious divergence can appear (Dabiri et al., 16 Mar 2025).

A first-principles extension is given in “First-principle based Floquet engineering of solids in the velocity gauge” (Tiwari et al., 26 Jan 2025). There the key step is to respect Hilbert-space truncation explicitly through the nested-commutator expansion

$\mathbf A(t)$6

implemented in a maximally-localized Wannier basis. For laser-dressed trans-polyacetylene, the 6-band and 17-band truncated velocity-gauge models reproduce the large-basis benchmark far more efficiently: the paper reports about 4 minutes CPU time per $\mathbf A(t)$7-point for the 17-band truncated velocity gauge, compared with about 1080 minutes for the 140-band conventional velocity gauge (Tiwari et al., 26 Jan 2025).

Real-time propagation methods in solids also exploit the periodicity-preserving character of the velocity gauge. “Velocity-gauge real-time TDDFT within a numerical atomic orbital basis set” (Pemmaraju et al., 2017) formulates the time-dependent Kohn–Sham equations in a Bloch representation, with independent $\mathbf A(t)$8-point propagation and a Crank–Nicholson scheme. The implementation in SIESTA is benchmarked against ARTED and Elk for weak-field response, intense-field dynamics, and core-level spectroscopy, and the work notes the well-known fictitious zero-frequency mode in VG-RT-TDDFT linear response (Pemmaraju et al., 2017). “Velocity-gauge real-time time-dependent density functional tight-binding for large-scale condensed matter systems” (Xu et al., 2023) extends this strategy to DFTB+, targeting large periodic condensed-matter systems containing thousands of atoms, with benchmark applications including a 512-atom amorphous silicon supercell (Xu et al., 2023).

The velocity gauge has also been tested in spatially inhomogeneous strong-field settings. “Numerical studies of light-matter interaction driven by plasmonic fields: the velocity gauge” (Chacón et al., 2015) derives the velocity-gauge Hamiltonian for plasmonic fields, transforms it to the length gauge, and shows analytically and numerically that both gauges give equivalent descriptions of the emitted HHG spectra for the single-active-electron helium model, even when the field is spatially inhomogeneous. The same study concludes, however, that the length gauge is often numerically more convenient for converged plasmonic HHG calculations (Chacón et al., 2015).

6. Electrodynamic v-gauge and specialized applications

In classical electrodynamics, the velocity gauge is a generalized gauge condition on the scalar and vector potentials rather than a minimal-coupling prescription for matter. The scalar and vector potentials satisfy

$\mathbf A(t)$9

$-q\,\mathbf{E}(t)\cdot\mathbf{r}$0

so the scalar potential propagates at speed $-q\,\mathbf{E}(t)\cdot\mathbf{r}$1, while the vector potential contains both $-q\,\mathbf{E}(t)\cdot\mathbf{r}$2- and $-q\,\mathbf{E}(t)\cdot\mathbf{r}$3-propagating parts. The associated fields remain unchanged under the usual gauge transformation

$-q\,\mathbf{E}(t)\cdot\mathbf{r}$4

because $-q\,\mathbf{E}(t)\cdot\mathbf{r}$5 and $-q\,\mathbf{E}(t)\cdot\mathbf{r}$6 (Giri et al., 2022).

“Velocity, temporal and generalized Kirchhoff gauges” (Yang et al., 27 Aug 2025) develops the corresponding dyadic Green’s functions. In the generalized velocity gauge,

$-q\,\mathbf{E}(t)\cdot\mathbf{r}$7

the time-domain dyadic Green’s function reduces to the Lorenz-gauge result when $-q\,\mathbf{E}(t)\cdot\mathbf{r}$8 and to the Coulomb-gauge result when $-q\,\mathbf{E}(t)\cdot\mathbf{r}$9. The same work introduces the generalized Kirchhoff gauge, whose frequency-domain Green’s function reduces to the Coulomb gauge at $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$0 and to the temporal gauge at $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$1 for $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$2 (Yang et al., 27 Aug 2025).

Specialized quantum applications further emphasize the practical role of momentum and current operators in velocity-gauge formulations. “Calculation of the nonrelativistic Bethe logarithm in the velocity gauge” (Korobov, 2012) recasts the low-energy radiative correction in terms of the current operator

$\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$3

and reports benchmark values $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$4 and $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$5 for helium and $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$6, respectively (Korobov, 2012). In quantum chemistry, “Velocity Gauge for Oscillator Strength in $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$7SCF theory” (Shen et al., 16 Mar 2026) uses the translational invariance of the momentum operator to avoid the origin dependence that affects length-gauge transition dipoles between nonorthogonal $\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$8SCF determinants. The velocity-gauge oscillator strength,

$\mathbf{k}\rightarrow \mathbf{k}+\mathbf{A}(t)$9

offers origin-independent predictions without additional correction schemes to the Kohn–Sham wavefunctions (Shen et al., 16 Mar 2026).

Taken together, these usages show that “velocity gauge” is not a single domain-specific convention but a broad family of vector-potential-based representations. In exact treatments it is gauge equivalent to alternative formulations; in practical work its distinctive advantages emerge when translational symmetry, momentum-space locality, current-based observables, or origin independence matter most. A plausible implication is that the velocity gauge is especially valuable precisely in those regimes where direct use of the position operator is singular, ambiguous, or numerically fragile.