---
title: Frequency-Dependent Electric Dipole Polarizability
url: https://www.emergentmind.com/topics/frequency-dependent-electric-dipole-electric-dipole-polarizability
type: topic
---

# Frequency-Dependent Electric Dipole Polarizability

Frequency-dependent electric dipole-electric dipole polarizability is the linear-response quantity that relates an applied electric field at angular frequency $\omega$ to the induced electric dipole moment. In stationary systems it is the Fourier-domain form of a causal dipole-response kernel; in tensor notation it is written as $\alpha_{ij}(\omega)$ or, in response-theory form, $\alpha_{AB}(-\omega;\omega)=-\langle\!\langle \mu_A;\mu_B\rangle\!\rangle_\omega$. In isotropic settings it reduces to a scalar, whereas in anisotropic, hyperfine-resolved, cavity-coupled, periodic, or temporally modulated systems it is intrinsically tensorial and may depend on additional geometric, quantum-number, or time arguments. Recent treatments span high-precision atomic clocks, empirically assembled oscillator-strength distributions, cavity QED response theory, periodic coupled-cluster theory, time-varying particles, Wigner crystals, and multilayer nanoshells [2308.09378] [1508.01986] [2606.24513] [2002.12297].

## 1. Definition as a linear-response function

In the stationary time domain, the induced dipole moment is expressed as a causal convolution,
\[
p(t)=\int_0^\infty \alpha(\gamma)\,E(t-\gamma)\,d\gamma,
\]
while the frequency-domain relation takes the form
\[
p(\omega)=\epsilon_0 \alpha_{ee}(\omega)\,E(\omega)
\]
for a stationary, linear, isotropic particle [2002.12297]. In tensor form for molecular or periodic systems, the induced dipole satisfies
\[
m_i(\omega)=\sum_j \alpha_{ij}(\omega)E_j(\omega),
\]
and coupled-cluster linear response uses the convention
\[
\alpha_{ij}(\omega)=-\langle\!\langle \mu_i;\mu_j\rangle\!\rangle_\omega
\]
[2508.01266]. In QED-HF, the corresponding working definition is
\[
\alpha_{AB}(-\omega;\omega)=-\langle\!\langle \mu_A^e;\mu_B^e\rangle\!\rangle_\omega
\]
[2606.24513].

Atomic Stark-shift theory makes the same object directly observable. For a monochromatic field, Chakraborty and Sahoo write the leading light shift of a hyperfine level as
\[
\Delta E_{\rm light}=-\frac{1}{2}\,\alpha_{F,M_F}(\omega)\,\mathcal{E}_L^2(\omega),
\]
with $\alpha_{F,M_F}(\omega)$ reducing to the static polarizability at $\omega=0$ [2308.09378]. The same response quantity therefore underlies microscopic spectroscopy, optical trapping, clock-state Stark shifts, and macroscopic constitutive relations.

A persistent source of confusion is not the physics but the notation. The literature represented here uses atomic units, SI units, and the dimensionless relative polarizability $\alpha_{ee}$; it also alternates between scalar, Cartesian-tensor, hyperfine-tensor, and two-time formulations. These are different representations of the same linear electric-dipole response rather than different observables.

## 2. Spectral, sum-over-states, and correlational representations

A standard stationary representation expresses $\alpha(\omega)$ through discrete and continuum oscillator strengths. For magnesium, the real- and imaginary-frequency forms are
\[
\alpha(\omega)=\sum_n \frac{f_n}{(E_n-E_0)^2-\omega^2}+\int_0^\infty dE\,\frac{df/dE}{E^2-\omega^2},
\]
\[
\alpha(i\omega)=\sum_n \frac{f_n}{(E_n-E_0)^2+\omega^2}+\int_0^\infty dE\,\frac{df/dE}{E^2+\omega^2},
\]
with the Thomas-Reiche-Kuhn sum rule
\[
S(0)\equiv \sum_n f_n+\int f(\varepsilon)\,d\varepsilon=N
\]
and the static limit
\[
S(-2)=\alpha(0)
\]
[1508.01986]. In that construction, enforcing $S(0)=12$ required inclusion of 3s, 2p, 2s, and 1s strength, and the resulting static polarizability was $\alpha(0)=72.2\,a_0^3$ [1508.01986].

The absorptive part can also be derived from fluctuation-dissipation theory. For a Wigner crystal with dipole operator $\mu(t)=-e\,x(t)$, the imaginary part is written as
\[
\alpha''(\omega)=\frac{1}{2\hbar}\left[1-e^{-\beta\hbar\omega}\right]\int_{-\infty}^{+\infty} dt\,e^{i\omega t}\langle \mu(t)\mu(0)\rangle,
\]
or equivalently through the dipole spectral density [2509.07590]. In the high-temperature coarse-grained limit the same work obtains $\alpha''(\omega)\propto \omega$ for $\omega\ll \Omega_W$, whereas for $\omega>\Omega_W$ the absence of resonant modes causes $\alpha''(\omega)$ to drop sharply, implying transparency above the Wigner cutoff [2509.07590].

The real part follows from causality through a Kramers-Kronig integral,
\[
\alpha'(\omega)=\frac{2}{\pi}\,P\!\int_0^\infty d\omega'\,\frac{\omega'\alpha''(\omega')}{\omega'^2-\omega^2},
\]
a relation stated explicitly for the Wigner-crystal treatment and, in the nonstationary case, satisfied by the temporal complex polarizability for each fixed observation time [2509.07590] [2002.12297]. In this sense, frequency-dependent dipole polarizability simultaneously encodes dispersion, absorption, and the oscillator-strength budget of the system.

## 3. Angular-momentum structure, hyperfine resolution, and Stark shifts

For hyperfine-resolved atomic levels, the polarizability decomposes into scalar, axial-vector, and tensor parts. For the $^{133}$Cs clock states,
\[
\alpha_{F,M_F}(\omega)=\alpha_F^S(\omega)+\mathcal{A}\frac{M_F}{2F}\cos\theta_k\,\alpha_F^A(\omega)
+\frac{3M_F^2-F(F+1)}{F(2F-1)}\frac{3\cos^2\theta_p-1}{2}\,\alpha_F^T(\omega),
\]
where $\mathcal A$ is the degree of polarization, $\theta_k$ is the angle between propagation direction and quantization axis, and $\theta_p$ is the polarization angle [2308.09378]. Chakraborty and Sahoo further separate each component into hyperfine-independent and hyperfine-induced pieces,
\[
\alpha_F^{\mathcal K}=\alpha_F^{\mathcal K(2,0)}+\alpha_F^{\mathcal K(2,1)},\qquad \mathcal K=S,A,T,
\]
where $(m,n)$ denotes $m$ E1 interactions and $n$ M1 hyperfine interactions [2308.09378].

A central selection rule is that the atomic tensor polarizability $\alpha_J^T$ vanishes for $J<3/2$. For the $6S_{1/2}$ ground state of cesium, this means that any tensor shift at the hyperfine level arises solely from hyperfine-induced third order [2308.09378]. The vector component also acquires hyperfine-level dependence through hyperfine angular factors and the M1 operator.

The same framework directly yields clock-state Stark shifts. The differential clock polarizability is
\[
\Delta \alpha_{\rm clock}(\omega)=\alpha_{F=4,M_F}(\omega)-\alpha_{F=3,M_F}(\omega),
\]
and the frequency shift is
\[
\Delta \nu(\omega)=-\frac{1}{2h}\,\Delta \alpha_{\rm clock}(\omega)\,\mathcal E^2
\]
[2308.09378]. For the static scalar differential shift, the Stark-shift coefficient is
\[
k_s=-\frac{1}{2}\left[\alpha_{F=4}^{S(2,1)}(0)-\alpha_{F=3}^{S(2,1)}(0)\right]
=-2.274(10)\times 10^{-10}\ {\rm Hz}/({\rm V/m})^2,
\]
in excellent agreement with the precise measurement $-2.271(4)\times 10^{-10}\ {\rm Hz}/({\rm V/m})^2$ [2308.09378]. The same study gives $\alpha_{6S}^S(0)=401.0(6)$ a.u., $\alpha_{6S}^S(936\,{\rm nm})=3022.1(40)$ a.u., and $\alpha_{6S}^S(1064\,{\rm nm})=1170.8(16)$ a.u., making explicit the strong frequency dependence relevant to optical trapping and clock interrogation [2308.09378].

## 4. Computational realizations and many-body methodology

High-accuracy polarizabilities now emerge from several distinct computational pipelines. For cesium clock states, the dominant valence contributions were evaluated by combining relativistic coupled-cluster matrix elements with measurements, while core and mixed terms were treated with random phase approximation. The electronic structure used Dirac-Hartree-Fock with the Dirac-Coulomb Hamiltonian, augmented by Breit and low-order QED corrections, and many-body correlation was included through RCCSD and RCCSDT [2308.09378]. The paper decomposes the response into valence, core, core-core, core-valence, and valence-core sectors and shows that continuum “tail” contributions change the static scalar Stark coefficient by about $8\%$, making them mandatory for precision work [2308.09378].

For magnesium, the dynamic function was assembled empirically from discrete resonance oscillator strengths and continuum photoionization cross sections. The adopted principal line strength $f=1.75$ brought the oscillator-strength distribution into TRK and $\alpha(0)$ consistency; choosing $f=1.80$ or $1.83$ would have increased $S(0)$ to $12.05$ or $12.08$ and $\alpha(0)$ to $73.4$ or $74.6\,a_0^3$, outside recommended ranges unless other inputs were retuned [1508.01986]. This construction is not a formal many-body response calculation, but it is an internally constrained reconstruction of the same response function.

In cavity QED Hartree-Fock, the response amplitudes obey a coupled electron-photon linear system,
\[
{\bf M}(\omega){\bf Z}_C(\omega)=-{\bf V}_C,
\]
with electronic TD-HF blocks, photon blocks at $\omega_{\rm cav}\pm \omega$, and a light-matter coupling vector ${\bf g}$ [2606.24513]. The dynamic polarizability is then extracted as
\[
\alpha_{AB}(-\omega;\omega)=\sum_{ia}\mu^e_{A,ia}\left[X^e_{B,ia}(\omega)+Y^e_{B,ia}(\omega)\right]
\]
[2606.24513]. The implementation was benchmarked against real-time TD-QED-HF and agreed to $\lesssim 0.01\%$ across all tested coupling strengths [2606.24513].

For periodic systems, the first implementation at the CCSD level with periodic boundary conditions was reported for 1D chains in CCResPy, an open-source software based on Python and the NumPy library [2508.01266]. The response uses a symmetric CC linear-response formulation with perturbed amplitudes at $\pm\omega$, collective orbital-$k$ indices, and DIIS acceleration. The periodic dipole operator in the length gauge introduces a technical difficulty: “missing integer” terms in the $U_k$ derivative matrix are usually set to zero, and the paper identifies this as the dominant residual source of discrepancy for longitudinal components in LiH chains [2508.01266].

## 5. Nonstationary, cavity-dressed, and periodic generalizations

When time-translation invariance is broken, the polarizability becomes a two-time kernel,
\[
p(t)=\int_0^\infty \alpha(\gamma,t)\,E(t-\gamma)\,d\gamma,
\]
and the temporal complex polarizability is defined by
\[
\alpha_T(\omega,t)=\int_0^\infty \alpha(\gamma,t)e^{-j\omega\gamma}\,d\gamma
\]
[2002.12297]. Under periodic modulation,
\[
\alpha_T(\omega,t)=\sum_n \alpha_n(\omega)e^{jn\Omega_m t},
\]
so a single drive frequency produces Floquet sidebands at $\omega+n\Omega_m$ [2002.12297]. A key point made explicitly in the time-varying-particle treatment is that the resulting effective permittivity is fundamentally different from the conventional stationary Drude-Lorentz model with time-dependent parameters simply substituted into it [2002.12297].

Cavity dressing modifies the response even for off-resonant perturbing frequencies. In QED-HF the photon enters the electronic response equations through a self-consistent feedback source proportional to
\[
\frac{\omega_{\rm cav}^2}{\omega_{\rm cav}^2-\omega^2}\,
(\boldsymbol\lambda\!\cdot\!\hat{\boldsymbol\mu}^e)_{ia}
\sum_D \lambda_D \alpha_{DC}(-\omega;\omega),
\]
so the cavity effect depends on the projection $\boldsymbol\lambda\!\cdot\!\boldsymbol\alpha_C(-\omega;\omega)$ [2606.24513]. The same analysis identifies special perturbing frequencies at which the percent change in $\alpha_{AA}(-\omega;\omega)$ crosses zero independent of $|\lambda|$, consistent with the condition
\[
\boldsymbol\lambda\!\cdot\!\boldsymbol\alpha_C(-\omega;\omega)=0
\]
[2606.24513]. Numerically, p-nitroaniline shows enhancements up to approximately $11\%$, $38\%$, and $83\%$ in $\alpha_{xx}$, $\alpha_{yy}$, and $\alpha_{zz}$ near $\omega\approx 0.18$ a.u. at $\lambda=0.05$ a.u. [2606.24513].

Periodic boundary conditions produce a different kind of environment dressing. In the 1D LiH calculations, going from an isolated molecule to a periodic chain inverts the relative magnitude of the polarizability tensor elements: molecular $\alpha_{xx}<\alpha_{yy}$, whereas periodic $\alpha_{xx}>\alpha_{yy}$ [2508.01266]. The paper interprets this as enhanced longitudinal response and reduced transverse response induced by intermolecular coupling in the chain [2508.01266]. This suggests that in low-dimensional solids the frequency-dependent dipole polarizability is not merely a localized molecular observable repeated in space, but a collective response per unit cell.

## 6. Nanostructures, collective modes, applications, and limits of validity

For spherical nanoparticles in the long-wavelength approximation, the frequency-dependent dipole polarizability is already complex because the permittivity is complex and dispersive. For a homogeneous sphere,
\[
\alpha(\omega)=4\pi \epsilon_0 \epsilon_m R^3\,
\frac{\epsilon(\omega)-\epsilon_m}{\epsilon(\omega)+2\epsilon_m},
\]
and the Fröhlich condition is approximately $\mathrm{Re}[\epsilon(\omega)+2\epsilon_m]=0$ [2403.15153]. For multilayer nanoshells, the normalized polarizability obeys a recursion,
\[
a_n(\omega)=\frac{\epsilon_n(\omega)-f_{n-1}(\omega)\epsilon_{n+1}(\omega)}
{\epsilon_n(\omega)+2f_{n-1}(\omega)\epsilon_{n+1}(\omega)},
\]
with $f_{n-1}$ determined by the inner structure’s scaled polarizability [2403.15153]. The corresponding Fröhlich function,
\[
F_n(\omega)=\mathrm{Re}\!\left[\epsilon_n(\omega)+2f_{n-1}(\omega)\epsilon_{n+1}(\omega)\right],
\]
locates dipolar localized surface plasmon resonances [2403.15153]. The recursive formula reproduces usual long-wavelength-approximation results and was compared with Mie-theory and FEM spectra for structures with up to $n=6$ layers, but the same paper notes the expected limits of the approximation: under-prediction of absorption magnitudes, smaller redshifts, breakdown at larger size parameter, and possible nonlocal or quantum effects for very thin metallic layers [2403.15153].

At the atomic and molecular scale, the same response function is the input for dispersion and surface interactions. The empirically constructed magnesium $\alpha(i\omega)$ yields
\[
C_6({\rm Mg\!-\!Mg})=642.4,\qquad
C_9({\rm Mg})=34{,}480,\qquad
C_3({\rm Mg|perfect\ conductor})=1.69
\]
with a quadrature value $1.687$ for $C_3$ [1508.01986]. The magnesium study also shows that incomplete inner-shell modeling is most visible in $\alpha(i\omega)$ for $\omega\gtrsim O(1)$ a.u., affecting $C_3$ and higher-order dispersion more than $\alpha(0)$ [1508.01986].

The Wigner-crystal formulation extends the same concept to a lattice of localized electrons. There the imaginary part is computed from the dipole-moment correlator in the harmonic approximation and, with phenomenological damping, becomes a thermally weighted continuum of Lorentz-like oscillators up to the Wigner cutoff $\Omega_W$ [2509.07590]. A standard ellipsometry set-up is proposed as a test of the predicted low-frequency linear rise of $\alpha''(\omega)$ and the transparent regime for applied frequencies greater than $\Omega_W$ [2509.07590].

Several limitations recur across these otherwise disparate realizations. The cesium calculations show that continuum tails and mixed core-valence terms, though small, are necessary for percent-level accuracy [2308.09378]. The periodic CCSD implementation avoids explicit damping and therefore omits genuinely resonant regions unless a small imaginary shift is introduced [2508.01266]. The QED-HF implementation likewise uses Hermitian response without linewidths or explicit Kramers-Kronig analysis [2606.24513]. The long-wavelength nanoshell recursion neglects retardation and multipoles by construction [2403.15153]. These are not contradictions among definitions of polarizability; they are regime-specific truncations applied to the same frequency-dependent electric dipole-electric dipole response function.

Source: https://www.emergentmind.com/topics/frequency-dependent-electric-dipole-electric-dipole-polarizability