Drude frequency response ($\sigma (\omega )={\sigma }_0/(1-i\omega \tau )$)

Use physical electric fields $\mathbf{E}(t)=\mathrm{Re}[\hat{\mathbf{E}}{e}^{-i\omega t}]$ and current densities $\mathbf{J}(t)=\mathrm{Re}[\hat{\mathbf{J}}{e}^{-i\omega t}]$, with peak complex amplitudes. After the initial transient has decayed, substitution into the Drude model equation gives the complex conductivityHere the mobile species is electrons. With the alternative $e^{+i\omega t}$ convention, the denominator is $1+i\omega \tau$ and the imaginary part changes sign; the physical response is identical.^[ref]

For $E_x(t)=E_0\cos \omega t$ with $\omega >0$, the real-time current density isThus a positive argument of complex conductivity in the $e^{-i\omega t}$ convention means that current density lags the electric field. At $\omega \tau =1$, its real part and imaginary part are both ${\sigma }_0/2$, its modulus is ${\sigma }_0/\sqrt{2}$, and the lag is $\pi /4$.

For $\omega \tau \ll 1$, complex conductivity tends to the positive constant ${\sigma }_0$: current density nearly follows the electric field. For $\omega \tau \gg 1$ within this model,The current density amplitude decreases and its response becomes predominantly inertial. Average power absorbed per volume is $\overline{p}=\frac{1}{2}\mathrm{Re}\sigma |\hat{\mathbf{E}}{|}^2\ge 0$. These limits describe the material response to a specified local electric field, not the complete electromagnetic response of a finite specimen with contacts. The Drude response has its pole at $\omega =-i/\tau$, consistent with the decaying causal response of a first-order relaxation equation.

The simple Drude model does not predict the microscopic momentum relaxation time. It neglects electronic band structure and Fermi-Dirac statistics; a useful semiclassical extension replaces $m$ by an appropriate band effective mass, but neither a single scalar mass nor one constant momentum relaxation time describes every material. Several charge carriers, energy-dependent scattering, and transitions between electronic bands can spoil the simple Drude frequency response. A successful low-angular frequency fit does not justify extrapolation to arbitrarily large angular frequencies.

For charge $-e$, field $B\hat{z}$, time dependence $e^{i\omega t}$ and circular components $E_{\pm }=E_x\pm i{E}_y$, the active positive-frequency response is the plus component and its denominator contains $\omega -{\omega }_c$. Reversing the field or exchanging helicity labels swaps the signs. A printed circular conductivity must state those conventions together.

A doped semiconductor at frequencies far below its bound-electron resonances can haveFor a weakly absorbing nonmagnetic half-space adjacent to vacuum, the normal-incidence optical reflectance is minimized near refractive-index matching, ${ϵ}_r\simeq 1$. When ${ϵ}_b>1$ and ${\gamma }_c$ is small this gives ${\omega }_{\min }\simeq {\omega }_{pc}/\sqrt{{ϵ}_b-1}$. The zero of the real part of ${ϵ}_r$, near ${\omega }_{pc}/\sqrt{{ϵ}_b}$, is a distinct frequency. Increasing the carrier density shifts the minimum to shorter wavelength if the carrier effective mass and background response remain fixed. With appreciable loss, the minimum must be found from the full complex Fresnel equations.

For a weakly absorbing doped semiconductor with one dominant carrier species and approximately constant background dielectric function ${ϵ}_b>1$, the Drude model gives a normal-incidence reflectivity minimum near ${ϵ}_r=1$. Thus ${\omega }_{pc}^2=({ϵ}_b-1){\omega }_{\min }^2$, with ${\omega }_{pc}^2=n{e}^2/({ϵ}_0{m}^{\ast })$ and ${\omega }_{\min }=2\pi c/{\lambda }_{\min }$. Eliminating the plasma frequency proves the displayed mass estimate. The real-dielectric zero instead has ${\omega }_0\simeq {\omega }_{pc}/\sqrt{{ϵ}_b}$ and is the nearby screened plasma edge, so its wavelength requires a different denominator. With appreciable loss, fit the full complex Fresnel equations rather than identifying the minimum with an exactly real dielectric value.