ωFQFμ: Fluctuating Charge and Dipole Model

• ωFQFμ is a polarizable embedding framework that assigns fluctuating atomic charges and dipoles to capture anisotropic electronic polarization and charge transfer effects.
• It unifies regularized charge–charge, charge–dipole, and dipole–dipole interactions using parameterization based on electronegativity, hardness, and polarizability.
• The method efficiently computes excitation energies and IR spectra in QM/MM setups and extends to modeling 2D materials with unique in-plane screening.

The ωFQFμ (Fluctuating Charge and Fluctuating Dipole) model is a fully polarizable embedding framework widely used in quantum/classical (QM/MM) hybrid calculations, particularly for molecular spectroscopy and excitation energy computations in condensed-phase systems and two-dimensional materials. In this approach, each atom in the classical (MM) region is assigned both a dynamically adjusting atomic charge and an induced dipole, enabling a self-consistent description of anisotropic electronic polarization and charge transfer effects in the environment. The ωFQFμ formalism unifies regularized charge–charge, charge–dipole, and dipole–dipole electrostatics with parameterization based on atomic electronegativity, chemical hardness, and polarizability, and has been generalized to accommodate arbitrary geometries and material dimensionality.

1. Theoretical Foundations and Energetics

The total energy in the ωFQFμ scheme consists of three components: the pure quantum subsystem (QM), the polarization energy of the classical environment (MM), and their electrostatic coupling. The MM region's energy is given by a quadratic functional of the vector of fluctuating atomic charges $q_i$ and dipoles $\mu_i$:

$$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$

where $T^{qq}$, $T^{\mu\mu}$, and $T^{q\mu}$ are the regularized interaction kernels (charge–charge, dipole–dipole, and charge–dipole), and $\chi$ is the vector of atomic electronegativities. On-site self-interaction regularization ensures the diagonal elements of $T^{qq}$ and $T^{\mu\mu}$ realize the atomic "hardness" $\eta_i$ and inverse polarizability $\mu_i$0, respectively.

The inclusion of QM/MM coupling terms (the interaction between QM density and MM multipoles) creates a coupled nonlinear system, solved iteratively to self-consistency. The QM Fock matrix is augmented with polarizable terms dependent on current $\mu_i$1 and $\mu_i$2 values, and the MM multipoles respond to both the QM electrostatic field and each other (Giovannini et al., 2019, Giovannini et al., 2019).

2. Physical Parameters and Regularization

Three atomic physical parameters are central: electronegativity $\mu_i$3, hardness $\mu_i$4, and polarizability $\mu_i$5.

• $\mu_i$6 determines the preference for attracting electrons (driving the equilibrium charge);
• $\mu_i$7 sets the energetic penalty for charge distortion (diagonal of $\mu_i$8);
• $\mu_i$9 controls the cost of induced dipole (diagonal of $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$0).

These parameters are chosen to reproduce known gas-phase atom properties and, by construction, enable transferability and constrain the regularized Gaussian kernels, which remove pathologies associated with delta-function (point) multipoles (Giovannini et al., 2019).

3. Linear Response, Excitations, and Analytical Gradients

For ground-state QM/MM calculations, the system is solved via coupled self-consistent field and polarization equations, producing equilibrium $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$1 and $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$2. Excitation energies are obtained by embedding-augmented Casida linear response (LR), where the classical region responds to the transition density at each iteration. The LR formalism is augmented by a polarization kernel $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$3 constructed from on-the-fly calculated perturbed $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$4 and $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$5 arising from the transition density, leading to an LR eigenproblem with Casida's structure but embedding-augmented kernels (Giovannini et al., 2019).

A corrected Linear Response (cLR) protocol accounts for state-specific MM relaxation to the full excited-state density, introducing a post-response correction to the excitation energies. Analytical first and second derivatives (for gradients, Hessians, and IR intensities) are available by solving coupled perturbed Hartree–Fock/Kohn–Sham and polarization response equations. All polarization responses enter via the explicit dependence of the total energy and dipole on $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$6 and $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$7 and their parametric derivatives (Giovannini et al., 2019).

4. Implementation and Computational Considerations

The ωFQFμ implementation relies on assembling the block-dense polarization matrix $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$8 from $$E_\mathrm{MM}(q,\mu) = \frac{1}{2} q^T T^{qq} q + \frac{1}{2} \mu^T T^{\mu\mu} \mu + q^T T^{q\mu} \mu + \chi^T q$$9, $T^{qq}$0, $T^{qq}$1, as well as charge conservation constraints. At each ground-state and response step, a linear system $T^{qq}$2 is solved for the current polarization variables and Lagrange multipliers. Ground-state and excited-state calculations differ only in the right-hand side (RHS), making the approach computationally efficient, with typical LU factorization and solution costs scaling as $T^{qq}$3 for full (dense) MM embedding (Giovannini et al., 2019). Reuse of decomposed matrices across states and snapshots is standard.

IR intensities are calculated from analytically derived Hessians and total dipole derivatives, fully incorporating both electronic and MM response contributions in normal-mode analysis. The required derivatives of $T^{qq}$4 and $T^{qq}$5 with respect to nuclear displacements are computed via additional polarization response solves.

5. Extensions: Two-Dimensional Kernels and ωFQFμ for 2D Materials

For two-dimensional materials, the standard 3D Coulomb interaction kernels are replaced with exact 2D charge–charge (Rytova–Keldysh), charge–dipole, and dipole–dipole kernels. The electrostatic energy is thus:

$T^{qq}$6

where $T^{qq}$7 is the Rytova–Keldysh potential, $T^{qq}$8 and $T^{qq}$9 its gradient and Hessian (yielding charge–dipole and dipole–dipole interactions). At large distances, these kernels asymptote to their 3D Coulombic forms, but at short distances reflect the unique in-plane screening and polarizability of the monolayer (Kezerashvili et al., 2022). This modification to the ωFQFμ machinery enables accurate modeling of planar charge redistribution and dielectric response in 2D materials.

6. Applications and Performance

Extensive benchmarking of QM/FQFμ embedding has been carried out for optical solvatochromic shifts (vertical excitation energies) and infrared (IR) spectra of molecular solutes in condensed phases. The inclusion of both fluctuating charges and dipoles (beyond charges alone) routinely yields near-quantitative agreement with experimental solvatochromic shifts and accurately reproduces IR band shapes and intensity patterns, particularly when anisotropic polarization or hydrogen-bonding is important (Giovannini et al., 2019, Giovannini et al., 2019). The method provides a consistent improvement over continuum (QM/PCM) or fluctuating-charge-only (QM/FQ) approaches, especially for systems with strong local polarizability.

7. Context in Broader Polarizable Embedding and Electrostatic Models

The ωFQFμ model embodies a generalization of the charge equalization principle and embedded atom models by incorporating explicit fluctuating dipoles, utilizing well-regularized Gaussian multipole kernels. It enables rigorous, parameterizable, and computationally efficient simulation of electronic polarization phenomena across both 3D and 2D materials, as well as their interfaces. In this respect, ωFQFμ is a representative of the emerging generation of atomistic, physically-parameterized embedding models that can seamlessly interpolate between molecular and extended (crystalline, 2D) regimes while retaining analytical tractability for both ground-state and excited-state properties (Giovannini et al., 2019, Giovannini et al., 2019, Kezerashvili et al., 2022).