8.1 Dielectric function

Definition and significance

The dielectric function describes how a material responds to an applied electric field as a function of frequency. It's the bridge between what electrons do at the atomic scale and what you can actually measure with light, making it central to nearly everything in the optical properties of solids.

Concept of dielectric function

The dielectric function $\varepsilon(\omega)$ is a complex, frequency-dependent quantity that encodes how a material polarizes in response to an applied electric field. It relates the displacement field $\mathbf{D}$ to the electric field $\mathbf{E}$ through:

$\mathbf{D} = \varepsilon(\omega)\,\varepsilon_0\,\mathbf{E}$

Because $\varepsilon(\omega)$ depends on frequency, it captures the full electronic structure of the material: where it absorbs, where it's transparent, and where it reflects. Every optical measurement you'll encounter in this unit ultimately traces back to this single function.

Role in material properties

• Optical properties: The refractive index $n$ and absorption coefficient $\alpha$ are both derived directly from $\varepsilon(\omega)$.
• Electronic screening: In semiconductors and metals, $\varepsilon(\omega)$ determines how effectively conduction electrons screen the Coulomb interaction between charges.
• Plasmonic resonances: The condition $\varepsilon_1(\omega) = 0$ (with small $\varepsilon_2$) defines where collective electron oscillations (plasmons) occur in metallic nanostructures.
• Energy storage: The real part of the dielectric function governs how much electrostatic energy a capacitor can store.

Mathematical formulation

Complex dielectric function

The dielectric function splits into real and imaginary parts:

$\varepsilon(\omega) = \varepsilon_1(\omega) + i\,\varepsilon_2(\omega)$

• $\varepsilon_1(\omega)$ (real part): describes energy storage and dispersion. It tells you how much the material slows and bends light.
• $\varepsilon_2(\omega)$ (imaginary part): describes energy dissipation and absorption. Peaks in $\varepsilon_2$ correspond to frequencies where the material strongly absorbs photons.

Together, the magnitude $|\varepsilon(\omega)|$ and the loss angle $\delta = \tan^{-1}(\varepsilon_2/\varepsilon_1)$ give a complete picture of the material's response.

Frequency dependence

The dielectric function varies dramatically across the electromagnetic spectrum:

• At low frequencies (well below any resonance), $\varepsilon(\omega)$ approaches the static dielectric constant $\varepsilon_s$.
• At very high frequencies (far above all resonances), $\varepsilon(\omega) \to 1$ because electrons can't keep up with the field, and the material looks like vacuum.
• Near resonance frequencies (electronic transitions, phonon modes), $\varepsilon(\omega)$ changes rapidly. These are the dispersion regions where absorption is strong and the refractive index varies quickly with frequency.

Kramers-Kronig relations

The real and imaginary parts of $\varepsilon(\omega)$ are not independent. Causality (the fact that a material can't respond before the field arrives) enforces a strict mathematical link between them:

$\varepsilon_1(\omega) = 1 + \frac{2}{\pi}\,\mathcal{P}\!\int_0^{\infty} \frac{\omega'\,\varepsilon_2(\omega')}{\omega'^2 - \omega^2}\,d\omega'$

$\varepsilon_2(\omega) = -\frac{2\omega}{\pi}\,\mathcal{P}\!\int_0^{\infty} \frac{\varepsilon_1(\omega') - 1}{\omega'^2 - \omega^2}\,d\omega'$

Here $\mathcal{P}$ denotes the Cauchy principal value. The practical payoff is significant: if you measure only the reflectivity $R(\omega)$ over a broad frequency range, you can use Kramers-Kronig analysis to reconstruct both $\varepsilon_1$ and $\varepsilon_2$ without needing a separate phase measurement.

Physical interpretation

Microscopic vs. macroscopic response

At the microscopic level, individual atoms and molecules develop small dipole moments in response to a local electric field. The macroscopic dielectric function emerges from averaging over enormous numbers of these microscopic dipoles.

The catch is that the local field at an atom isn't the same as the applied macroscopic field, because surrounding dipoles modify it. Local field corrections (such as the Clausius-Mossotti relation) bridge this gap. For heterogeneous materials like composites or thin-film stacks, effective medium theories (Maxwell-Garnett, Bruggeman) approximate the macroscopic $\varepsilon$ from the dielectric functions of the constituents.

Polarization mechanisms

Different polarization mechanisms dominate at different frequency scales, each with a characteristic response time:

As you increase the frequency of the applied field, the slower mechanisms "freeze out" one by one, and $\varepsilon(\omega)$ decreases in a stepwise fashion. This is why the static dielectric constant is always larger than the optical (high-frequency) one.

Susceptibility and permittivity

The electric susceptibility $\chi_e(\omega)$ quantifies how much polarization $\mathbf{P}$ a field produces:

$\mathbf{P} = \varepsilon_0\,\chi_e(\omega)\,\mathbf{E}$

The dielectric function and susceptibility are related by:

$\varepsilon(\omega) = 1 + \chi_e(\omega)$

The displacement field then takes the familiar form:

$\mathbf{D} = \varepsilon_0\,\mathbf{E} + \mathbf{P} = \varepsilon_0\bigl(1 + \chi_e\bigr)\mathbf{E} = \varepsilon_0\,\varepsilon(\omega)\,\mathbf{E}$

In most condensed matter contexts, "relative permittivity" $\varepsilon_r$ and "dielectric function" $\varepsilon(\omega)$ refer to the same quantity.

Dielectric function in solids

Three classical models form the backbone for understanding $\varepsilon(\omega)$ in different classes of solids. Each adds one layer of physics on top of the last.

Drude model (metals)

The Drude model treats conduction electrons as a classical gas that scatters with a characteristic relaxation time $\tau = 1/\gamma$:

$\varepsilon(\omega) = 1 - \frac{\omega_p^2}{\omega^2 + i\gamma\omega}$

where the plasma frequency is $\omega_p = \sqrt{Ne^2 / m\varepsilon_0}$, with $N$ the conduction electron density. Separating real and imaginary parts:

• $\varepsilon_1(\omega) = 1 - \frac{\omega_p^2}{\omega^2 + \gamma^2}$
• $\varepsilon_2(\omega) = \frac{\omega_p^2\,\gamma}{\omega(\omega^2 + \gamma^2)}$

This model successfully explains why metals are highly reflective below $\omega_p$ (where $\varepsilon_1 < 0$) and become transparent above it. For most metals, $\omega_p$ falls in the ultraviolet, which is why they reflect visible light. The Drude model also works well for doped semiconductors at infrared frequencies.

Free electron model (collisionless limit)

Setting $\gamma = 0$ in the Drude model gives the collisionless free electron result:

$\varepsilon(\omega) = 1 - \frac{\omega_p^2}{\omega^2}$

This is purely real, meaning no absorption. Below $\omega_p$, $\varepsilon < 0$ and electromagnetic waves are evanescent (total reflection). Above $\omega_p$, $\varepsilon > 0$ and waves propagate. This idealized limit is useful for understanding the sharp plasma edge in reflectivity spectra.

Lorentz oscillator model (insulators and semiconductors)

For bound electrons, the Lorentz model adds a restoring force with resonance frequency $\omega_0$:

$\varepsilon(\omega) = 1 + \frac{Ne^2}{m\varepsilon_0}\,\frac{1}{\omega_0^2 - \omega^2 - i\gamma\omega}$

Near $\omega_0$, $\varepsilon_2$ shows a strong absorption peak, and $\varepsilon_1$ exhibits anomalous dispersion (it decreases with increasing frequency). Away from resonance, $\varepsilon_1$ increases with frequency (normal dispersion), which is the origin of the familiar prismatic splitting of white light. Real insulators often require a sum of multiple Lorentz oscillators to fit the full spectrum.

Experimental techniques

Optical spectroscopy

• Reflectivity measurements over a broad frequency range, combined with Kramers-Kronig analysis, yield both $\varepsilon_1$ and $\varepsilon_2$.
• Spectroscopic ellipsometry measures the change in polarization state of light upon reflection, directly determining the complex refractive index $\tilde{n} = n + ik$ without needing Kramers-Kronig transforms.
• Terahertz time-domain spectroscopy probes the low-frequency dielectric response, useful for studying phonons and free carrier dynamics.

Electron energy loss spectroscopy (EELS)

EELS measures the energy lost by fast electrons transmitted through a thin sample. The loss function is directly related to the dielectric function:

$\text{Loss function} = -\text{Im}\!\left[\frac{1}{\varepsilon(\omega)}\right]$

Peaks in the loss function correspond to bulk and surface plasmon excitations. EELS provides dielectric information over a very wide energy range (0–100 eV) and, in a transmission electron microscope, can map dielectric properties with nanometer spatial resolution.

Applications in condensed matter

Optical properties of materials

The complex refractive index connects directly to the dielectric function:

$\tilde{n} = n + ik = \sqrt{\varepsilon(\omega)}$

From this you get the reflectivity at normal incidence, $R = \left|\frac{\tilde{n} - 1}{\tilde{n} + 1}\right|^2$, and the absorption coefficient $\alpha = 2\omega k / c$. These relationships are the foundation for designing anti-reflection coatings, optical filters, and photonic crystals.

Plasmonics

Surface plasmon polaritons exist at metal-dielectric interfaces when $\varepsilon_1$ of the metal is negative and its magnitude exceeds that of the dielectric. These modes confine electromagnetic energy to subwavelength scales, enabling surface-enhanced Raman spectroscopy (SERS), biosensing, and the design of metamaterials with exotic optical properties (e.g., negative refraction).

Screening effects

In a solid, the bare Coulomb potential $V(q)$ between two charges is reduced to $V(q)/\varepsilon(q, \omega)$. This screening profoundly affects:

• Band structure: quasiparticle energies differ from bare Kohn-Sham eigenvalues due to dynamical screening.
• Exciton binding energies: weaker screening (smaller $\varepsilon$) in low-dimensional systems leads to tightly bound excitons.
• Carrier transport: screened impurity potentials scatter carriers less effectively than bare ones.

Dielectric function vs. conductivity

Relationship and differences

The optical conductivity $\sigma(\omega)$ and dielectric function carry equivalent information but emphasize different physics. They're related by:

$\sigma(\omega) = -i\omega\varepsilon_0\bigl[\varepsilon(\omega) - 1\bigr]$

The conductivity viewpoint is natural when thinking about currents (metals, transport), while the dielectric function viewpoint is natural for polarization (insulators, optics). The real part of the conductivity, $\text{Re}[\sigma(\omega)]$, is proportional to $\varepsilon_2(\omega)$ and directly gives the rate of energy absorption from the field.

Frequency regimes

Anisotropic materials

Tensor representation

In anisotropic crystals (anything less symmetric than cubic), the dielectric function becomes a $3 \times 3$ tensor $\varepsilon_{ij}(\omega)$. The displacement field component along direction $i$ depends on the electric field components along all three directions:

$D_i = \varepsilon_0 \sum_j \varepsilon_{ij}(\omega)\,E_j$

By choosing the principal axes of the crystal, you can diagonalize this tensor so that each axis has its own scalar dielectric function. For nonlinear optical effects, the response generalizes to higher-rank tensors ($\chi^{(2)}_{ijk}$, etc.).

Birefringence and dichroism

• Birefringence occurs when the refractive index differs along different principal axes. An incoming beam splits into two polarization components (ordinary and extraordinary rays) that travel at different speeds. Calcite is the classic example.
• Dichroism occurs when the absorption differs along different axes, so one polarization is attenuated more than the other. Tourmaline crystals exhibit this naturally.

Both effects are exploited in wave plates, polarizers, and other polarization-sensitive optical components.

Quantum mechanical approach

Lindhard dielectric function

The Lindhard function is the quantum mechanical generalization of the free electron dielectric function. It accounts for the Pauli exclusion principle and Fermi-Dirac statistics, giving a wavevector- and frequency-dependent result $\varepsilon(q, \omega)$. Key features include:

• Correct description of Friedel oscillations in the screened potential around an impurity.
• The Kohn anomaly: a singularity in the dielectric response at $q = 2k_F$ that affects phonon dispersion.
• It reduces to the Drude result in the long-wavelength ($q \to 0$) limit.

The Lindhard function serves as the starting point for more sophisticated many-body treatments.

Many-body perturbation theory

Going beyond the independent-particle Lindhard picture requires many-body perturbation theory:

• The GW approximation replaces the bare Coulomb interaction with a dynamically screened one, yielding accurate quasiparticle band structures.
• The Bethe-Salpeter equation (BSE) includes electron-hole interactions, which is essential for correctly describing excitonic peaks in the optical spectrum of semiconductors and insulators.
• Green's function methods provide a systematic framework for incorporating exchange and correlation effects that mean-field theories miss.

Environmental effects

Temperature dependence

Temperature modifies the dielectric function through several channels:

• Thermal excitation changes carrier concentrations in semiconductors, shifting the plasma frequency.
• Increased phonon populations broaden and shift phonon-related absorption features.
• Near phase transitions (e.g., ferroelectric $\to$ paraelectric), the static dielectric constant can diverge or change discontinuously.

Pressure effects

Applied pressure changes interatomic distances and electronic wavefunctions, which in turn modifies band gaps and dielectric response. Pressure can also drive structural phase transitions (e.g., insulator-to-metal transitions) that produce dramatic, discontinuous changes in $\varepsilon(\omega)$. This makes pressure a useful tuning knob for engineering optical and electronic properties.

Computational methods

Density functional theory (DFT)

DFT calculates the ground-state electronic structure from first principles using the Kohn-Sham formalism. From the Kohn-Sham eigenstates, you can compute $\varepsilon_2(\omega)$ via Fermi's golden rule (summing over interband transitions) and then obtain $\varepsilon_1(\omega)$ through Kramers-Kronig. DFT typically underestimates band gaps, so the resulting optical spectra are red-shifted compared to experiment.

Beyond DFT: GW and BSE

To fix the band gap problem, the GW approximation corrects quasiparticle energies using dynamical screening. For optical spectra where excitonic effects matter, the Bethe-Salpeter equation is solved on top of GW-corrected band structures. This DFT $\to$ GW $\to$ BSE pipeline is the current standard for predictive, first-principles calculations of dielectric functions in real materials, though it remains computationally expensive.