Maier S.A. Plasmonics: Fundamentals and Applications. Майер С.А. Плазмоника: Теория и приложения

Подождите немного. Документ загружается.

Acknowledgments

I wish to thank my colleague Tim Birks for all his efforts in proof-reading

an early draft of this book and his helpful criticism, and David Bird for his

encouragement to undertake this project. Thanks also to my student Charles

de Nobriga for working through a more advanced version of this text, and of

course to my wife Mag for all the lovely distractions from writing ...

PART I

FUNDAMENTALS OF PLASMONICS

Introduction

Research in plasmonics is currently taking place at a breathtaking pace, and

we can expect that many more will join the game in the near future. But for

the newcomer, where to start? Before diving into particular sub-ﬁelds, funda-

mental or application-driven, a solid basis for an understanding of the more

specialized literature is clearly desirable. This part of the text aims to help in

building such a core knowledge. The ﬁrst chapter sets the groundwork by de-

scribing the optical properties of metals, starting with Maxwell’s equations and

the derivation of the dielectric function of the free electron gas. The following

three chapters introduce surface plasmon polaritons both at single interfaces

and in multilayer structures, and describe experimental techniques for their ex-

citation and observation. Chapter 5 adds the second important ingredient of

the game, localized surface plasmons in metallic nanostructures. The ﬁrst part

of the book closes by describing electromagnetic modes at low frequencies,

where surface plasmon polaritons based on metals become highly delocalized,

and surface structuring must be employed to create more conﬁned modes.

Chapter 1

ELECTROMAGNETICS OF METALS

While the optical properties of metals are discussed in most textbooks on

condensed matter physics, for convenience this chapter summarizes the most

important facts and phenomena that form the basis for a study of surface plas-

mon polaritons. Starting with a cursory review of Maxwell’s equations, we

describe the electromagnetic response both of idealized and real metals over a

wide frequency range, and introduce the fundamental excitation of the conduc-

tion electron sea in bulk metals: volume plasmons. The chapter closes with a

discussion of the electromagnetic energy density in dispersive media.

1.1 Maxwell’s Equations and Electromagnetic Wave

Propagation

The interaction of metals with electromagnetic ﬁelds can be ﬁrmly under-

stood in a classical framework based on Maxwell’s equations. Even metallic

nanostructures down to sizes on the order of a few nanometres can be described

without a need to resort to quantum mechanics, since the high density of free

carriers results in minute spacings of the electron energy levels compared to

thermal excitations of energy k

T at room temperature. The optics of met-

als described in this book thus falls within the realms of the classical theory.

However, this does not prevent a rich and often unexpected variety of optical

phenomena from occurring, due to the strong dependence of the optical prop-

erties on frequency.

As is well known from everyday experience, for frequencies up to the vis-

ible part of the spectrum metals are highly reﬂective and do not allow elec-

tromagnetic waves to propagate through them. Metals are thus traditionally

employed as cladding layers for the construction of waveguides and resonators

for electromagnetic radiation at microwave and far-infrared frequencies. In this

6 Electromagnetics of Metals

low-frequency regime, the perfect or good conductor approximation of inﬁnite

or ﬁxed ﬁnite conductivity is valid for most purposes, since only a negligible

fraction of the impinging electromagnetic waves penetrates into the metal. At

higher frequencies towards the near-infrared and visible part of the spectrum,

ﬁeld penetration increases signiﬁcantly, leading to increased dissipation, and

prohibiting a simple size scaling of photonic devices that work well at low

frequencies to this regime. Finally, at ultraviolet frequencies, metals acquire

dielectric character and allow the propagation of electromagnetic waves, albeit

with varying degrees of attenuation, depending on the details of the electronic

band structure. Alkali metals such as sodium have an almost free-electron-like

response and thus exhibit an ultraviolet transparency. For noble metals such

as gold or silver on the other hand, transitions between electronic bands lead

to strong absorption in this regime.

These dispersive properties can be described via a complex dielectric func-

tion ε(ω), which provides the basis of all phenomena discussed in this text.

The underlying physics behind this strong frequency dependence of the optical

response is a change in the phase of the induced currents with respect to the

driving ﬁeld for frequencies approaching the reciprocal of the characteristic

electron relaxation time τ of the metal, as will be discussed in section 1.2.

Before presenting an elementary description of the optical properties of met-

als, we recall the basic equations governing the electromagnetic response, the

macroscopic Maxwell equations. The advantage of this phenomenological ap-

proach is that details of the fundamental interactions between charged parti-

cles inside media and electromagnetic ﬁelds need not be taken into account,

since the rapidly varying microscopic ﬁelds are averaged over distances much

larger than the underlying microstructure. Speciﬁcs about the transition from

a microscopic to a macroscopic description of the electromagnetic response of

continuous media can be found in most textbooks on electromagnetics such as

[Jackson, 1999].

We thus take as a starting point Maxwell’s equations of macroscopic elec-

tromagnetism in the following form:

∇·D = ρ

ext

(1.1a)

∇·B = 0 (1.1b)

∇×E =−

∂B

∂t

(1.1c)

∇×H = J

ext

∂D

∂t

. (1.1d)

These equations link the four macroscopic ﬁelds D (the dielectric displace-

ment), E (the electric ﬁeld), H (the magnetic ﬁeld), and B (the magnetic induc-

Maxwell’s Equations and Electromagnetic Wave Propagation 7

tion or magnetic ﬂux density) with the external charge and current densities

ext

and J

ext

. Note that we do not follow the usual procedure of presenting the

macroscopic equations via dividing the total charge and current densities ρ

tot

and J

tot

into free and bound sets, which is an arbitrary division [Illinskii and

Keldysh, 1994] and can (especially in the case of metallic interfaces) confuse

the application of the boundary condition for the dielectric displacement. In-

stead, we distinguish between external

(

ext

, J

ext

)

and internal

(

ρ,J

)

charge

and current densities, so that in total ρ

tot

= ρ

ext

+ ρ and J

tot

= J

ext

+ J.The

external set drives the system, while the internal set responds to the external

stimuli [Marder, 2000].

The four macroscopic ﬁelds are further linked via the polarization P and

magnetization M by

D = ε

E + P (1.2a)

H =

B − M, (1.2b)

where ε

and μ

are the electric permittivity

and magnetic permeability

vacuum, respectively. Since we will in this text only treat nonmagnetic me-

dia, we need not consider a magnetic response represented by M, but can limit

our description to electric polarization effects. P describes the electric dipole

moment per unit volume inside the material, caused by the alignment of micro-

scopic dipoles with the electric ﬁeld. It is related to the internal charge density

via ∇·P =−ρ. Charge conservation

(

∇·J =−∂ρ/∂t

)

further requires that

the internal charge and current densities are linked via

J =

∂P

∂t

. (1.3)

The great advantage of this approach is that the macroscopic electric ﬁeld

includes all polarization effects: In other words, both the external and the in-

duced ﬁelds are absorbed into it. This can be shown via inserting (1.2a) into

(1.1a), leading to

∇·E =

tot

. (1.4)

In the following, we will limit ourselves to linear, isotropic and nonmagnetic

media. One can deﬁne the constitutive relations

D = ε

εE (1.5a)

B = μ

μH. (1.5b)

≈ 8.854 ×10

−12

F/m

≈ 1.257 ×10

−6

H/m

8 Electromagnetics of Metals

ε is called the dielectric constant or relative permittivity and μ = 1therela-

tive permeability of the nonmagnetic medium. The linear relationship (1.5a)

between D and E is often also implicitly deﬁned using the dielectric suscepti-

bility χ (particularly in quantum mechanical treatments of the optical response

[Boyd, 2003]), which describes the linear relationship between P and E via

P = ε

χE. (1.6)

Inserting (1.2a) and (1.6) into (1.5a) yields ε = 1 +χ.

The last important constitutive linear relationship we need to mention is that

between the internal current density J and the electric ﬁeld E, deﬁned via the

conductivity σ by

J = σ E. (1.7)

We will now show that there is an intimate relationship between ε and σ ,

and that electromagnetic phenomena with metals can in fact be described using

either quantity. Historically, at low frequencies (and in fact in many theoretical

considerations) preference is given to the conductivity, while experimentalists

usually express observations at optical frequencies in terms of the dielectric

constant. However, before embarking on this we have to point out that the

statements (1.5a) and (1.7) are only correct for linear media that do not exhibit

temporal or spatial dispersion. Since the optical response of metals clearly

depends on frequency (possibly also on wave vector), we have to take account

of the non-locality in time and space by generalizing the linear relationships to

D(r,t)= ε





ε(r − r



,t −t



)E(r



) (1.8a)

J(r,t)=





σ(r −r



,t −t



)E(r



). (1.8b)

ε and σ therefore describe the impulse response of the respective linear re-

lationship. Note that we have implicitly assumed that all length scales are

signiﬁcantly larger than the lattice spacing of the material, ensuring homo-

geneity, i.e. the impulse response functions do not depend on absolute spatial

and temporal coordinates, but only their differences. For a local response, the

functional form of the impulse response functions is that of a δ-function, and

(1.5a) and (1.7) are recovered.

Equations (1.8) simplify signiﬁcantly by taking the Fourier transform with

respect to



dtdre

(

K·r−ωt

)

, turning the convolutions into multiplications. We

are thus decomposing the ﬁelds into individual plane-wave components of

wave vector K and angular frequency ω. This leads to the constitutive rela-

Maxwell’s Equations and Electromagnetic Wave Propagation 9

tions in the Fourier domain

D(K,ω) = ε

ε(K,ω)E(K,ω) (1.9a)

J(K,ω) = σ(K,ω)E(K,ω). (1.9b)

Using equations (1.2a), (1.3) and (1.9) and recognizing that in the Fourier do-

main ∂/∂t →−iω, we ﬁnally arrive at the fundamental relationship between

the relative permittivity (from now on called the dielectric function)andthe

conductivity

ε(K,ω) = 1 +

iσ(K,ω)

. (1.10)

In the interaction of light with metals, the general form of the dielectric re-

sponse ε(ω, K) can be simpliﬁed to the limit of a spatially local response via

ε(K = 0,ω) = ε(ω). The simpliﬁcation is valid as long as the wavelength λ

in the material is signiﬁcantly longer than all characteristic dimensions such as

the size of the unit cell or the mean free path of the electrons. This is in general

still fulﬁlled at ultraviolet frequencies

Equation (1.10) reﬂects a certain arbitrariness in the separation of charges

into bound and free sets, which is entirely due to convention. At low frequen-

cies, ε is usually used for the description of the response of bound charges to a

driving ﬁeld, leading to an electric polarization, while σ describes the contri-

bution of free charges to the current ﬂow. At optical frequencies however, the

distinction between bound and free charges is blurred. For example, for highly-

doped semiconductors, the response of the bound valence electrons could be

lumped into a static dielectric constant δε, and the response of the conduction

electrons into σ



, leading to a dielectric function ε(ω) = δε +

iσ



(ω)

.Asimple

redeﬁnition δε → 1andσ



→ σ



δε will then result in the general form

(1.10) [Ashcroft and Mermin, 1976].

In general, ε(ω) = ε

(ω)+iε

(ω) and σ(ω) = σ

(ω)+iσ

(ω) are complex-

valued functions of angular frequency ω, linked via (1.10). At optical frequen-

cies, ε can be experimentally determined for example via reﬂectivity studies

and the determination of the complex refractive index ˜n(ω) = n(ω) + iκ(ω)

of the medium, deﬁned as ˜n =

√

ε. Explicitly, this yields

However, spatial dispersion effects can lead to small corrections for surface plasmons polaritons in metallic

nanostructures signiﬁcantly smaller than the electron mean free path, which can arise for example at the tip

of metallic cones (see chapter 7).

10 Electromagnetics of Metals

= n

−κ

(1.11a)

= 2nκ (1.11b)



+ε

(1.11c)

κ =

. (1.11d)

κ is called the extinction coefﬁcient and determines the optical absorption of

electromagnetic waves propagating through the medium. It is linked to the

absorption coefﬁcient α of Beer’s law (describing the exponential attenuation

of the intensity of a beam propagating through the medium via I(x)= I

−αx

)

by the relation

α(ω) =

2κ(ω)ω

. (1.12)

Therefore, the imaginary part ε

of the dielectric function determines the

amount of absorption inside the medium. For



, the real part n

of the refractive index, quantifying the lowering of the phase velocity of the

propagating waves due to polarization of the material, is mainly determined by

. Examination of (1.10) thus reveals that the real part of σ determines the

amount of absorption, while the imaginary part contributes to ε

and therefore

to the amount of polarization.

We close this section by examining traveling-wave solutions of Maxwell’s

equations in the absence of external stimuli. Combining the curl equations

(1.1c), (1.1d) leads to the wave equation

∇×∇×E =−μ

∂

∂t

(1.13a)

K(K · E) −K

E =−ε(K,ω)

E, (1.13b)

in the time and Fourier domains, respectively. c =

√

is the speed of light

in vacuum. Two cases need to be distinguished, depending on the polarization

direction of the electric ﬁeld vector. For transverse waves, K ·E = 0, yielding

the generic dispersion relation

= ε(K,ω)

. (1.14)

For longitudinal waves, (1.13b) implies that

ε(K,ω) = 0, (1.15)

The Dielectric Function of the Free Electron Gas 11

signifying that longitudinal collective oscillations can only occur at frequencies

corresponding to zeros of ε(ω). We will return to this point in the discussion

of volume plasmons in section 1.3.

1.2 The Dielectric Function of the Free Electron Gas

Over a wide frequency range, the optical properties of metals can be ex-

plained by a plasma model, where a gas of free electrons of number density

n moves against a ﬁxed background of positive ion cores. For alkali metals,

this range extends up to the ultraviolet, while for noble metals interband transi-

tions occur at visible frequencies, limiting the validity of this approach. In the

plasma model, details of the lattice potential and electron-electron interactions

are not taken into account. Instead, one simply assumes that some aspects of

the band structure are incorporated into the effective optical mass m of each

electron. The electrons oscillate in response to the applied electromagnetic

ﬁeld, and their motion is damped via collisions occurring with a characteristic

collision frequency γ = 1/τ . τ is known as the relaxation time of the free

electron gas, which is typically on the order of 10

−14

s at room temperature,

corresponding to γ = 100 THz.

One can write a simple equation of motion for an electron of the plasma sea

subjected to an external electric ﬁeld E:

x + mγ

x =−eE (1.16)

If we assume a harmonic time dependence E(t) = E

−iωt

of the driving ﬁeld,

a particular solution of this equation describing the oscillation of the electron

is x(t) = x

−iωt

. The complex amplitude x

incorporates any phase shifts

between driving ﬁeld and response via

x(t) =

m(ω

+iγ ω)

E(t). (1.17)

The displaced electrons contribute to the macroscopic polarization P =−nex,

explicitly given by

P =−

m(ω

+iγ ω)

E. (1.18)

Inserting this expression for P into equation (1.2a) yields

D = ε

(1 −

+iγ ω

)E, (1.19)

where ω

is the plasma frequency of the free electron gas. Therefore we

arrive at the desired result, the dielectric function of the free electron gas: