Classical microscopic models of optical response of condensed matter v. 2 4.4.2021 1 FK020_Electrodynamics of solids Josef Humlíček, humlicek@physics.muni.cz Two classical models: 1. An ensemble of Lorentzian oscillators, 2. a gas of free charge carriers (Drude model). 2 Lorentzian oscillator 3 Charges bound elastically to their equilibrium positions are dislocated by the electrical force of the electromagnetic wave (the magnetic component is negligible). In a harmonic wave travelling along z, the wavelength is supposed to be large compared to the spatial structure of the investigated charges; the relevant quantity is then the displacement r of the charge e in the plane perpendicular to z. The force is proportional to the intensity of the electric field of the wave,     , .zi k z t oE z t E e   Newtonian equation of motion for the displacement of damped harmonic oscillator of the mass m at an arbitrary position z: 2 2 2 d d . d d i t o o r m r m m r eE e t t         Neglecting the contribution of possible vibrations at the eigenfrequency o, which is small at times long compared to  , the solution is a harmonic movement with the frequency  , 2 2 1 .i t o o e r E e m i           The complex values of intensity lead to complex values of the displacement (they have magnitude and phase). (3.1) (3.2) (3.3) 4 The displacement of the charge produces the dipole moment proportional to the electric field; a is called electrical polarizability. It disappears for zero charge (no dipole moment) and infinite mass (no displacement). The frequency dependence exhibits a resonance close to the eigenfrequency of the oscillator. With N oscillators in a unit volume, the volume density of the dipole moment (polarization) is 2 2 2 1 , , o e p er E m i a a           it is proportional to the intensity of the wave; the dimensionless coefficient (note the presence of the vacuum permittivity, SI units) c is called susceptibility (compare with the Russian term for it – the ability to accept the influence of the external field). The ensemble of oscillators is characterized by ;oP Np N E Ea c   2 2 2 1 . o o Ne m i c        The SI unit of polarization is C/m 2 , the same as that of area density of charge, or of the electric displacement D. (3.4) (3.5) (3.6) 5 Dielectric function is dimensionless proportionality factor between the electric displacement and field intensity: (1+ ) 1 .o o oD E E P E   c  c       Here, S is dimensionless “oscillator strength”, determining the “magnitude of the spectral structure”; it has a further simple meaning, as it is equal to the contribution of the oscillators to static permittivity (dielectric function at zero frequency). The eigenfrequency of oscillators determines the “spectral position” and the damping time  determines the “spectral width”. The usual notation for the real and imaginary part of the dielectric function uses subscripts 1 and 2: 22 2 2 2 2 2 2 1 1 1 , where .o o o o o o SNe Ne S m mi i                     It represents the susceptibility added to to unity (the vacuum contribution to the electric displacement). The ensemble of charged oscillators with the same eigenfrequency and volume density N leads to 1 2 .i    (3.7) (3.8) (3.9) 6 In the complex plane of frequencies, the dielectric function of Lorentzian oscillators has two poles (the roots of the denominator): 2 1 2 1 , ( )( ) oS          This form leads to a common approximate result fro the Lorentzian spectral profile in case of the weak damping. Namely, for 2 2 1 22 2 1 1 , . 2 4 2 4 o o i i                 2 2 2 2 2 1 21 1 . 1 12 2( ) ( )2 4 42 o o o o o oo o S S S i i                               1 , the approximation can be expressed aso   (3.10) (3.11) 7 This approximation can be used only for the frequencies close to the eigenfrequency, due to the following approximation of the poles in the complex plane, usable for small damping. Within this approximation, the imaginary part of  is bell-shaped, symmetric lineshape about the eigenfrequency, with the maximum of 1,2 2and 2 , 2 o o i            2,max .oS   Full width at half-maximum (FWHM) is 1/ .   Real part is antisymmetric about the eigenfrequency. (3.12) (3.13) (3.14) 8 Spectral lineshapes on the real axis of frequencies complies with the general requirement ( ) *( ) ,     where the asterisk denotes complex conjugation. Dielectric function of Lorentz oscillators (S=4, tau*omega_0=5). 9 An example of the difference between the exact and approximate form of the Lorentzian lineshapes for the real (“dispersive”) and imaginary (“absorptive”) parts of the dielectric function: Real (left) and imaginary (right) part of the dielectric function (S=4, tau*omega_0=5), exact (ex), approximate (aprox) forms and their difference (dif). 10 Relative agreement of the exact and approximate form improves for narrower Lorentzian profiles (longer times ) : Real part of the Lorentzian dielectric function (left: S=4, tau*omega_0=15, right: tau*omega_0=50), exact (ex), approximate (aprox), difference (dif). 11 Refractive index is the square root of the dielectric function. Using the common conventions: 2 2 2 2 2 1 1 2 1 1 2, ( )/ 2 , ( )/ 2 , 2 N n ik n k n                 where n and k are real. The real part (n) is always positive, the positive square root is assumed; the sign of k is the same as that of the imaginary part of the dielectric function – it describes the damping of the wave, travelling in the opposite direction for negative frequencies. Complex refractive index of Lorentzian oscillators (S=4, tau*omega_0=5). 12 Complex refractive index determines the reflectivity at planar interfaces. Fresnel amplitude reflectance r for the normal incidence at the interface with vacuum (the unit refractive index) is ( 1 ) /( 1 ) ,i r Re n ik n ik       where R is the squared modulus of r, i.e., the power reflectivity. Normal-incidence reflectivity and the phase angle of r (S=4, tau*omega_0=5). 13 The bands of large reflectivity above the eigenfrequency are called “reststrahlen”. They become more pronounced esp. for small damping of the oscillators. Normal-incidence reflectvance (S=4, tau*omega_0=5, 15 a 50). Complex conductivity 14 The velocity of the charge displacement in the Lorentz model is The current density produced by this movement equals to the charge transferred through unit area per unit time in the direction of the velocity. Assuming the concentration N of the oscillators, the current density is proportional to the field intensity, where the proportionality factor s is called conductivity: 2 2 d 1 . d i t o o r e i E e t m i           The current density is usually not in phase with the driving force, which results in the complex-valued conductivity. Its SI unit is 2 2 2 d 1 . d i t i t o o o r Ne j Ne i E e E e t m i    s            2 A/m 1 . V/m m   (3.15) (3.16) 15 A convenient expression for the conductivity is It uses the dimensionless oscillator strength S. At the eigenfrequency, the conductivity is real, 2 1 2 2 2 2 2 2 2 2 1 ( 1) , where . o o o o o o o Ne i i i m i S Ne i S mi s s s                              2 ( ) .o o oSs     2 2 1 . ( ) 1 ( / / )o o o o o i ii  s  s                This value is convenient in expressing the dimensionless quantity (3.17) (3.18) (3.19) 16 The spectral dependence of complex conductivity on real axis is even and odd in its real and imaginary parts, respectively. The real part decays for large frequencies as 1/ 2 , the imaginary part as 1/ . Conductivity of Lorentz oscillators (S=4, tau*omega_0=5). Drude model for free charge carriers 17 The elastic bonding of charges to their equilibrium positions might disappear (e.g., for electrons in metals, or electrons and holes in semiconductors). This corresponds to the vanishing eigenfrequency of the Lorentzian oscillator. The displacement and velocity of the free charge carriers are Using the carrier density N and their mass m, we arrive at the dielectric function 1 d 1 , . ( / ) d ( / ) i t i t o o e e r E e r i E e m i t m i               2 22 2 2 2 2 2 1 1 1 , where . 11 1 p p p o o Ne Ne i m m i                              The oscillator strength of the Lorentz formula is replaced by the square of plasma frequency; the name invokes the possibility of collective vibrations in the plasma of free carriers at this frequency. NB: setting the eigenfrequency of the Lorentzian oscillator to zero provides an approximation of its behavior for large frequencies, with the motion independent of the elastic bonds. (3.20) (3.21) 18 The two poles of the Drude dielectric function are: As the modulus of complex frequency increases, the dielectric function approaches unity fast enough to warrant the unimportance of the pole at the origin in deriving the Kramers-Kronig relation The real part is on real axis is even, the imaginary part odd. With increasing frequency, the real part is growing monotonically from 1 20 a . i       1 2 2 20 ( ) 12 ( ) d .P              2 1(0) 1 ( )p    towards unity (the response of vacuum, as the charges are not able to follow the fast changes of the electromagnetic field); the imaginary part is singular at zero, and approaches zero as the inverse third power of frequency. (3.22) (3.23) (3.24) 19 Note, in particular, the behavior of the real part (esp. the zero crossing close to the plasma frequency): Dielectric function of the Drude model (omega_p=1, tau=10). Above 0.75, both parts are multiplied by 100. 20 Negative inverse of the dielectric function, ressembles (in a narrow neighborhood of the plasma frequency) the resonance of the Lorentz model. Its imaginary part is proportional to the absorbed energy of longitudinal waves, observable by the EELS technique (Electron Energy Loss Spectroscopy). Note the simple relation: 2 2 2 1 , p i i             1 ( ) 1 .p pi        (3.25) (3.26) 21 The resonance at the plasma frequency has a nearly Lorentzian profile: Negative inverse of the Drude dielectric function (tau*omega_p=10). 22 Conductivity is the convenient response function of free carriers; in the Drude model, As expected, the dc conductivity is proportional to the damping time constant, and inversely proportional to the mass. The frequency dependence depends solely on the damping time. Absorbed energy is proportional to the real part (cf. with the Joule heat U 2 /R developed with the dc voltage U on the resistance R). The simple frequency dependence of the absorptive (real) part allows for a direct check of the (general) sum rule 2 1 2 2 2 2 2 ( 1) 1 , where . ( ) 1 ( ) 1 p o o o o o p i i i i Ne i m   s s s       s s                        2 10 2 ( )d .p o s       When the carrier concentration stays constant, possible changes in the frequency dependence of the absorption processes are limited by the constant area below the real part of conductivity. (3.27) (3.28) 23 Complex conductivity of the Drude model (tau*omega_p=10). Spectral dependence of the complex conductivity has an even real, and odd imaginary part. The real part decays as 1/ 2 , the imaginary part as 1/ , for high frequencies. 24 Both parts of the refractive index diverge with the frequency approaching zero, due to the divergence of the imaginary part of the dielectric function: 2 2 2 2 2 1 1 2 1 1 2, ( )/ 2 , ( )/ 2 . 2 N n ik n k n                 Complex refractive index of the Drude model (tau*omega_p=10). (3.29) 25 Similar to the Lorentz model is the band of high reflectivity at the normal incidence; here it starts at the zero frequency: Normal-incidence reflectance and the phase of amplitude reflectance of the Drude model (tau*omega_p=10). 26 The large values of the reflectivity below the plasma frequency are the consequence of different importance of the closeness of the numerator and denominator in 2 2 2 2 ( 1) , namely 1, or , or 1 . ( 1) n k R n k n n n k         For weak damping, the normal-incidence reflectance is only weakly dependent on . Normal-incidence reflectance of the Drude model (tau*omega_p=3, 30, and 300). (3.30) 27 Travelling plane wave looses its amplitude in the direction of propagation (z), whenever the imaginary part of the refractive index is nonzero. The decrease of the amplitude is exponential, characterized by the “penetration depth of light”. It results from the square of wavenumber, 2 2 2 ( ) ,zk c    where c is the light velocity in vacuum. The amplitude is diminished by 1/e of the initial value (about 37%) on the distance   1 , Im z z c k k     where k is the imaginary part of the refractive index. Spectral dependence of the penetration depth results from the multiplication of 1/k by the vacuum wavelength reduced by the factor of 2 . When travelling this distance, the intensity is reduced by the factor 1/e 2 (about 14%) of its initial value. (3.31) (3.32) 28 In the range of large reflectivity, the penetration dept is small. With the plasma frequency in the midinfrared range, e.g., -1 10 m , 1000 cm , 124 meV ,vac vac      Inverse of the imaginary part of refractive index in the Drude model (tau*omega_p=3, 10, 30). the penetration depth at the half of the plasma frequency is about 3 microns. 29 The usual term for the penetration length at radio frequencies is “skin depth”, as the field penetrates into small depths below the surface of conductors. In our model, the relevant quantity is 1 for 1/ . 2 / p z i k c          The imaginary part of refractive index increases with decreasing frequency as its square root; consequently, the penetration depth increases as the square root inversed. For small frequencies, we can express the penetration depth using the dc conductivity as   1 2 , Im z z o ok  s     where is the permeability of vacuum. For the dc conductivity of 10 5 1/.cm and the “optical” frequency of 1 THz, the penetration depth is about 160 nm, at the “radio” frequency of 100 MHz about 16 microns. In the usual formula for the skin depth, 2 ,z     we easily identify the resistivity (inverse of conductivity) and the magnetic permeability of the conductor, which may be different from the vacuum value. o (3.33) (3.34) (3.35)