4

Thermal ¯uctuations and their fundamental laws

The subject of investigation in this chapter is the fundamental law of nature that
relates the quantum ¯uctuation radiation of an object of any physical nature with its
dissipative properties in macroscopic scales and is called the ¯uctuation±dissipation
theorem ± FDT. Attention is chie¯y given to the physical aspect of the problem.
Two approaches, which are important for remote sensing and instrumental applica-
tions, are analysed in this chapter. They are the quasi-stationary FDT approxima-
tion, called the Nyquist formula, and the geometric-optical approximation, the
Kirchho€ law. In addition, methodological issues of the application of FDT
results under real remote sensing conditions are considered.

4.1 THERMAL RADIATION AND THERMAL FLUCTUATIONS: A

HISTORICAL REVIEW

One of the fundamental factors that explains the principal signi®cance of thermal
radiation (sometimes called thermal electric ¯uctuations) in remote sensing and
astrophysical applications, is the fairly transparent physical relationship between
the recorded radiation and the internal thermal structure of a physical object and
its physical-chemical and physical-geometric features. In fact, all fundamental results
of both remote sensing (of the Earth and planets) and astrophysics obtained up to
the present, are largely based on using the results of remote observation of thermal
radiation (thermal ¯uctuations), which is generated and reveals itself (certainly, in
the observational respect) in di€erent parts of the electromagnetic spectrum,
depending on its temperature and physical properties. Of course, in addition to
thermal radiation, many other electromagnetic emissions either fall from space to
the Earth or are formed directly under Earth conditions. These emissions also have a
¯uctuation character, but are not pertinent to thermal radiation physics. The sepa-
ration or extraction of various types of emissions from the experimental data
128 Thermal ¯uctuations and their fundamental laws [Ch. 4

sometimes represents a complicated scienti®c problem in itself. In this book, as

mentioned above, we shall consider thermal radiation issues only.
All physical objects having physical temperatures other than absolute zero are
continuously emitting a ¯uctuating electromagnetic ®eld arising from internal
energy, which stipulates the possibility of spontaneous transitions between
vibration±rotation levels of molecules in gases, oscillations of molecules in liquid
and solid bodies and oscillations of a lattice in solid bodies, with subsequent de-
excitation of electromagnetic quanta. The radiation has a typically quantum
character and cannot be described within the framework of the classical Maxwellian
theory of electromagnetism. The radiation energy covers a very broad range of
wavelengths and has (as is usually stated in radiophysics and optics) a continuous
spectrum of rather complicated form, the position of maximum of which depends on
the thermal temperature of matter. As to terminological approaches, the literature
o€ers a spectrum of names for this radiation: emission, emitted radiation, Planck's
emission, black-body radiation, thermal emission, radiothermal emission, radio-
emission, grey-body radiation, outgoing radiation.
The study of various parts of the electromagnetic spectrum of thermal radiation,
and thermal ¯uctuations in general, has proceeded rather non-uniformly in the
historical respect (Rytov, 1953, 1966; Levin and Rytov, 1967; Schopf, 1978). In
studying thermal electrical ¯uctuations and thermal radiation there are two ranges
of questions, the relation between which was elucidated long ago, but only in the
early 1950s was it formulated mathematically as a uni®ed theory called the
¯uctuation±dissipation theorem.
One of the areas we speak about arose considerably earlier in time; it concerns
the issues of the thermal radiation of heated bodies considered in the optical wave-
length band. Researchers have been interested in the relation between the emitting
body and the environment from the beginning of the nineteenth century (P. Prevost,
B. Stewart, A. J. Angstrom). But only G. R. Kirchho€ had sucient insight into the
subject to elucidate the primary ideas concerning `ray-radiation' (emission) and
absorption. Kirchho€ 's work has rested upon a discovery made some months
before that event by Kirchho€ and Bunsen: they found that Fraunhofer's absorption
lines in the solar spectrum coincided with the lines of emission of known vapours and
gases. Kirchho€ himself evaluated his discovery as the proof of the fact that matter
outside the Earth consists of known chemical elements. Doubtless, this was one of
the ®rst outstanding discoveries in astrophysics (see Schopf (1978) for more details).
One of the fundamental results obtained by Kirchho€, on the basis of applica-
tion of thermodynamic laws to equilibrium thermal radiation, was the proof of the
fact that the spectral density of this radiation is a universal function of frequency and
temperature. The complete determination of the form of the universal function
constituted the problem of the next stage of development in thermal radiation
theory. The ®nal solution of this problem, based on the quantum hypothesis and
resulting in the expression for spectral density of the equilibrium (absolutely black-
body) radiation, that is valid for any frequency, was given by M. Planck. The
detailed and fascinating presentation of the (sometimes dramatic) history of this
discovery was given in a book by H.-G. Schopf (1978).
Sec. 4.1] 4.1 Thermal radiation and thermal ¯uctuations: a historical review 129

Another, no less important, result which is very signi®cant for the practice of
remote sensing and astrophysical investigations was the theoretical proof, in the
geometric-optical approximation, of the law called the Kirchho€ law. This law
states that the ratio between `emissive' ability (or radiation intensity) and
absorbing ability is identical for all bodies (irrespective of their shape, chemical
composition, aggregate state, surface properties, etc.) for a given temperature and
for a given frequency. Subsequent investigations have shown that the universal
constant in the Kirchho€ law is closely related to the spectral intensity of equilibrium
radiation inside the enclosure of a thermostat. At present, several forms of the
Kirchho€ law presentation are used in the theory and practice of remote sensing
and astrophysical investigations, the physical sense of which is, certainly, identical.
Some of these forms will be described in Chapter 6.
Another area where researchers have again (though much later) encountered
thermal ¯uctuations is the so-called `noise' in electrical circuits and, ®rst of all, in
amplifying devices, whose noise properties have already been mentioned many times
in this book (see Chapter 3). The close relation between electrical noises and thermal
radiation lies in the fact that this radiation represents a wave electromagnetic ®eld
generated by thermal electrical ¯uctuations in physical bodies of various natures.
The physical explanation of the fact that a uni®ed and rather general theoretical
approach to such closely related physical phenomena has been absent for a long time
lies in the great distinction between the frequencies of the electromagnetic oscilla-
tions of interest in each of the aforementioned areas. The questions related to
thermal radiation arose and have been studied as optical problems using the
methods of geometric optics (see section 1.6). On the other hand, electrical noise
was found experimentally in the band of low radio frequencies, which made it
possible to consider them within the framework of the theory of quasi-stationary
currents only (see section 1.6).
However, in the 1940s the intensive development of radar engineering gave rise
to considerable growth in the sensitivity of radio and radar receiving equipment
(Skolnik, 1980; Brown, 1999). This made it possible to reliably record, in the
decimetre and centimetre bands, thermal electromagnetic radiation coming from
natural physical objects situated both on the Earth's surface, and in space. It was
this technological basis on which the new science ± radio-astronomy ± arose and
continues to progress actively now (Esepkina et al., 1973; Ruf, 1999; Barreiro, 2000;
Kardashiov, 2000). A little later aerospace radio thermal location (microwave radio-
metry) and scatterometry of the Earth surface arrived and continue to be eciently
developed (Basharinov et al., 1974; Sharkov and Etkin, 1976; Bass et al., 1977;
Raney, 1983; Kalmykov, 1996; Carver et al., 1985; Shutko, 1986; Massonnet,
1996). Thus, the areas of thermal radiation and electrical noise `have touched each
other' closely in the microwave band.
Though the existence of electrical ¯uctuations of thermal origin in radio-
engineering circuits and receivers has been obvious since the ®rst steps in the
development of Brownian motion theory in statistical physics at the beginning of
the twentieth century, their experimental detection became possible as a result of
the improvement of radio engineering devices and, ®rst of all, amplifying systems at
130 Thermal ¯uctuations and their fundamental laws [Ch. 4

the end of the 1920s. In 1927 J. B. Johnson found that at an output of the ampli®er,
to the input of which the active resistance is connected, additional noise ± the chaotic
voltage ± was observed. As was found out later, this noise is of purely Gaussian type,
and its intensity (the mean square ± the variance) grows linearly with resistance R at
an input and with increasing physical temperature. Almost simultaneously with these
experiments H. Nyquist, using the existing physical concept of a random electro-
motive force (emf), localized in the active circuit, has showed that the spectral
intensity (Wiener's spectrum), G‡ f †, of the ¯uctuation emf, localized in the
arbitrary passive two-terminal circuit with impedance Z_ j2f †, is
G‡ f † ˆ 4kT Re Z_ j2f †; 4:1†
where k is the Boltzmann constant (see Appendix A), and T is the absolute tem-
perature. In such a form, this formula, called the Nyquist formula (or the Nyquist
theorem), gives the spectral intensity in the unit interval of positive frequencies and is
valid in the non-quantum region of frequencies and temperatures, i.e. for hf  kT
(here h is Planck's constant, see Appendix A). The rigorous quantum-mechanical
generalization of this formula, whose necessity was still pointed out by Nyquist
himself, was performed much later, however, as a result of the quantum-mechanical
derivation of the ¯uctuation±dissipation theorem (see, for instance, Levin and
Rytov, 1967). The complete form of the spectral density, which is valid both for
low temperatures, and for suciently high frequencies, hf > kT, is as follows:
 
hf
‡
G f † ˆ 2hf coth Re Z_ j2f † 4:2†
2kT
where coth x ˆ exp 2x† ‡ 1† exp 2x† 1† is the hyperbolic cotangent. From
quantum mechanics we know the expression for the mean energy of the so-called
quantum oscillator:
 
hf hf hf hf
 !; T† ˆ  f ; T† ˆ ‡ ˆ coth ; 4:3†
2 exp hf =kT† 1 2 2kT
in this case
h
hf ˆ 
h! ˆ !:
2
In such a case the Nyquist formula can have the more compact quantum form:
G‡ f † ˆ 4 f ; T† Re Z_ j2f †: 4:4†
The further development of the theory of thermal ¯uctuations resulted in the
appearance of a set of derivations of this formula and in far-reaching generalizations,
from which it issues as a very simple special case. First of all, we should point out
here the transition from concentrated ¯uctuation forces to detached random ®elds
(both electrical and magnetic, in the general case) and the construction of spatial
correlation functions for spectral amplitudes of detached ®elds in the frequency
bands not limited by the quasi-static condition (see section 1.6) (Rytov, 1953).
At that time, in the early ®fties, H. B. Callen with co-workers proved the
Sec. 4.2] 4.2 The ¯uctuation±dissipation theorem: a qualitative approach 131

¯uctuation±dissipation theorem (FDT), which generalized Nyquist's result: ®rst, to

dissipative systems of an arbitrary physical nature; second, to the quantum region of
frequencies and temperatures; and, third, to thermodynamic ¯uctuations described
by any number of discrete functions of time. Thus, the possibility of regular applica-
tion of FDT to distributed physical systems was opened up. This is just what was
done a little later when it was applied to the Maxwell equations (Landau and
Lifshitz, 1957), and the general FDT formulation was established for the case of
distributed dissipative systems (Levin and Rytov, 1967).
Thus, as a result of the `merging' of the two aforementioned directions, the
theory of thermal ¯uctuations (thermal radiation) in electrodynamics represents
one of the most important applications of the general theory of thermal ¯uctuations
to arbitrary macroscopic systems. First of all, we shall indicate the principal impor-
tance of using this theory in the observational practice of microwave remote sensing,
namely in those cases where the size of physical bodies is of the same order as the
working wavelength (see section 1.6), and the di€raction phenomena in interaction
problems make a noticeable and, sometimes (as in problems of emission and scatter-
ing from the wavy sea surface), overwhelming contribution.
Prior to considering the results of the application of the theory of thermal
¯uctuations to remote sensing, we shall discuss brie¯y, and mainly at the qualitative
level, the physical essence of the ¯uctuation±dissipation theorem.

4.2 THE FLUCTUATION±DISSIPATION THEOREM:

A QUALITATIVE APPROACH

The ¯uctuation±dissipation theorem is one of the fundamental laws of statistical

physics. It establishes for an arbitrary dissipative physical system the relationship
between the spectral density of spontaneous equilibrium ¯uctuations and its non-
equilibrium properties, the energy dissipation in a system in particular. The detailed
quantum-mechanical derivation of this theorem can be found both in original works
by H. B. Callen with co-authors (Callen and Welton, 1951; Callen and Green, 1952),
and in a series of textbooks and monographs on statistical physics (Landau and
Lifshitz, 1957; Rytov, 1953, 1966; Levin and Rytov, 1967).
To elucidate the qualitative physical issues, it is sucient to consider the special
case where the ¯uctuations in a system are determined by a single random quantity.
We designate it by  and assume that at the equilibrium state its mean value is zero.
Suppose also, that the system is situated in a thermostat and, accordingly, obeys the
canonical Gibbs (J. W. Gibbs) distribution. Further, the interaction with a thermo-
stat is supposed to be weak, so the system's energy can be introduced, which is
uniquely determined by the state of the system itself. Using quantum-mechanical
approaches, it can be shown that the spectral density of equilibrium ¯uctuations of
the quantity  is expressed in terms of the levels of energy En of the considered system
and magnitudes of matrix elements n;m . However, the actual calculation of En and
n;m for a real macroscopic system requires consideration of the micromechanism of
¯uctuations, and, generally, for real physical bodies the problem seems to be
132 Thermal ¯uctuations and their fundamental laws [Ch. 4

virtually hopeless. Note that the intensity of spontaneous internal thermal ¯uctua-
tions is described by means of En and n;m . Certainly, the macroscopic dynamics of a
system is not re¯ected at all in any way in the quantum-mechanical structure, since
the macroscopic process in a system, by which we mean the average variation of
system's parameters , may be caused in a dissipative system by the external e€ect of
macroscopic forces only.
The theoretical and practical value of FDT consists, in particular, in the fact
that for thermodynamic equilibrium systems it gets rid of the necessity to ®nd
directly En and n;m for a real physical body by expressing the spectral density of
¯uctuations in terms of a particular macroscopic characteristic of a system ± its
general susceptibility.
Suppose the system under consideration is disturbed by the e€ect of the external
force f t†. Now let this force be suciently small that the macroscopic response can
be found from the linearized equation of motion and, accordingly, the spectral
amplitude  to be linearly related with the spectral amplitude of the disturbing
force f !†:
 !† ˆ _ j!† f !†: 4:5†
Quantity _ j!† ˆ 0 !† j 00 !†, determined by relation (4.5) and called the
general susceptibility, is introduced for weak e€ects allowing for linearization of
the macroscopic equations of motion for the physical system under consideration.
Since for real f t† quantity  t† should also be real, we have _ j!† ˆ _ j!†, i.e.
0
!† is an even function, and 00 !† an odd one.
Often in physical practice the basic equation for a system is used in the form of
macroscopic response for the rate of time variation of the basic parameter of a
system, and then the relation for spectral amplitudes of velocity  t† and external
force f t† is presented as:
_
 !† ˆ Y_ j!† f !†; 4:6†
where coecient Y_ j!† is called the admittance of a system. The general suscept-
ibility _ j!† is associated with the admittance of a system by relation
Y_ j!† ˆ j! _ j!† (see Appendix B, equation (B.7)). The reciprocal quantity to the
admittance is called the impedance of the system, Z_ j!† ˆ 1=Y_ j!†. If we address
the theory of electrical circuits (see sections 1.6 and 2.6), then we ®nd that similar
parameters are introduced for describing the processes in electrical circuits as well.
The quantum-mechanical consideration of energy dissipation in a system even-
tually results in the following important relation between the spectral density of
¯uctuations, G !†, and the general susceptibility, which just expresses the
physical essence of FDT:
 !; T† 00
G !† ˆ !†: 4:7†
!
Here  !; T† is the mean energy of a quantum oscillator (4.3); in this case the
frequencies are considered in the whole frequency band (both positive and negative).
The fundamental relation (4.7) indicates that the spectral intensity of equilib-
rium ¯uctuations is determined by the imaginary part of a system's susceptibility,
Sec. 4.2] 4.2 The ¯uctuation±dissipation theorem: a qualitative approach 133

that is the quantity describing the dynamic behaviour of a linearized macroscopic

system beyond any relation with the ¯uctuations. Note here the following important
point: the dynamics of a system under strong external e€ects, where the macroscopic
equations of motion can also be nonlinear, has no relation to thermodynamic
¯uctuations. So, the macroscopic non-linearity (the deviation from Ohm's law in
electrical circuits) can be revealed only for essentially non-equilibrium distribution of
current carriers in conductors, or, in other words, beyond the FDT action
framework (Levin and Rytov, 1967).
According to (4.7), the total intensity (the variance) of ¯uctuations in a system is:
1 1
1  !; T† 00
2 ˆ G !† d! ˆ !† d!: 4:8†
1  1 !
Let us consider now another form of FDT, which is fairly often utilized. It is
based on the so-called Langevinian conception of ¯uctuation forces. (It was this
conception that was utilized by Nyquist for deriving his formula (4.1).) These
equivalent random forces are introduced into linearized macroscopic equations of
motion of a system as a `reason' for ¯uctuations, i.e. these forces are introduced
along with true external forces. Thus, by quantity f !† in equation (4.5) can be
meant the spectral amplitude of a random equivalent force, and equation (4.5) can
be understood as the equation relating spectral amplitudes of macroscopic random
processes  t† and f t†. Recalling the relation between the spectra of linearly bound
processes (2.65), we obtain from relations (4.5) and (4.7) the formula for the spectral
intensity Gf of the ¯uctuation force f t†:
G !†  !; T 00 !†
G f !† ˆ ˆ : 4:9†
j _ j!†j 2 ! j _ j!†j2

As an example, we shall use this formula for deriving the Nyquist relation. As we
noted above (section 1.6), in the general case for linear concentrated circuits the
generalized coordinates are the charges qj and the generalized velocities are the
currents q_ j ˆ Ij . The spectral amplitudes of currents Ij and electromotive forces %j
are speci®ed by Kirchho€'s generalized equations (Krug, 1936; Rytov, 1966)
X
Tj ˆ Y_ jk j!† %k
k
X ; 4:10†
%j ˆ Z_ jk j!† Ik ;
k

here Y_ jk j!† and Z_ jk j!† are mutually reciprocal matrices of the admittance and
impedance k of electrical circuits. If we deal with a single active resistance R, then its
admittance is 1=R and, accordingly, the general susceptibility is _ j!† ˆ 1=j!R, and
its imaginary part is 00 !† ˆ 1=!R. Thus, the intensity of ¯uctuations according to
relation (4.9) will be equal to:
1
2
2 ˆ % 2 ˆ R  !; T† d!: 4:11†
 0
134 Thermal ¯uctuations and their fundamental laws [Ch. 4

In the classical approximation, i.e. when 

h!  kT and, here,  !; T† ˆ kT, and
using the Langevinian conception (4.9), we (transferring to positive frequencies)
obtain the well-known Nyquist relation we have already used many times before:

2 ˆ %2 ˆ 4kTR
f : 4:12†

In studying the limiting sensitivity of mechanical systems, such as mechanical

gravitational wave detectors (Yamamoto et al., 2001), the same Nyquist formula is
used in a slightly di€erent form (certainly, without changing the physical essence of
the phenomenon), namely, in the form of the relationship between the thermal noise
spectrum and the mechanical response of a system:

4kT
G !† ˆ Im H_ j!†: 4:13†
!
The transfer function, H_ j!†, is written as

X_ j!†
H_ j!† ˆ ; 4:14†
F_ j!†

where F_ j!† and X_ j!† are the Fourier components of the applied mechanical force
and the displacement at the observation point, respectively. The imaginary part of
the transfer function represents the phase lag between the force and the displace-
ment, which is related to the dissipation of a system.
Below we shall summarize the basic qualitative components of FDT.

(1) Any dissipative system of arbitrary physical nature possesses spontaneous equi-
librium ¯uctuations whose intensity is determined by the macroscopic dissipative
properties of a system. As examples of various FDT applications to concen-
trated and distributed systems, we point out the investigations of thermal ¯uc-
tuations in liquids (Landau and Lifshitz, 1957), in mechanical systems, in
plasma, in electronic gas, in hydrodynamics and, which is closest to our
subject, the studies of thermal ¯uctuations of electromagnetic ®elds (Levin
and Rytov, 1967).
(2) The FDT action spreads to any relationship between frequencies and tempera-
tures, beginning with the classical limit hf  kT, both for low temperatures and
for high frequencies, hf  kT .
(3) In applying FDT to the electrodynamics, its action spreads to any relationship
between the geometrical size of a system and the working wavelengths of the
¯uctuation electromagnetic ®eld of radiation. In the case of geometrical optics,
L  , FDT asymptotically `transfers' into the form of Kirchhoff's law, and in
the quasi-static case, L  , into the Nyquist formula (4.1). Surprising is the
fact that in the intermediate, most complicated (diffraction) case, L  , it was
possible to ®nd in the most general case (Rytov, 1966; Levin and Rytov, 1967) a
rather transparent relationship between the emissive and absorbing properties of
media (Rytov's formulae) (see section 4.3).
Sec. 4.3] 4.3 Thermal ¯uctuations in the electrodynamics 135

4.3 THERMAL FLUCTUATIONS IN THE ELECTRODYNAMICS

As we have already indicated, the most important application of the general theory
of thermal ¯uctuations in arbitrary macroscopic systems is the theory of thermal
¯uctuations (radiation) in Maxwell's electrodynamics, as applied, ®rst of all, to
microwave sensing problems. There are two important aspects here.
The ®rst aspect is associated with the fact, that the general conditions of the
macroscopic electrodynamics applicability will be spread to the similar approach
(Stratton, 1941; Landau and Lifshitz, 1957; Levin and Rytov, 1967). First, it is
necessary that the inhomogeneities of macro®elds (the working wavelength of the
electromagnetic radiation) be much larger than the microinhomogeneities caused by
the molecular structure of emitting bodies. This requirement is ful®lled for a broad
range of electromagnetic radiation, including the band of optical frequencies.
Besides, the phenomenological concept of matter as a dielectric continuum (dielectric
formalism) (section 1.6) in Maxwellian theory implies the exclusion from the
statistical electronics, i.e. from accounting for such parameters as the elementary
charge, the number of elementary charges per unit volume, thermal velocities of
microcharges, their free path length, etc. Nevertheless, since the electrodynamics
part of a problem is solved in this case by means of the general Maxwell
equations, the results obtained cover all di€raction phenomena occurring under
the given physical and geometrical conditions, including, naturally, the extreme
cases as well, for instance, the quasi-stationary approximation and the geometrical
optics.
The second aspect concerns the following circumstance. As we have noted
above, thermal radiation has a typically quantum character and cannot be straight-
forwardly described within the framework of the classical Maxwellian theory of
electromagnetism. Within the phenomenological theory framework the ¯uctuation
electromagnetic ®eld is represented as the ®eld generated by random `detached'
sources of Langevinian type, distributed in the volume of a medium under investiga-
tion (Landau and Lifshitz, 1957; Levin and Rytov, 1967). In spite of using a rather
arti®cial approach ± the introduction of detached ¯uctuation ®elds ± such an
approach allows us to formulate any problem on equilibrium thermal ¯uctuations
of electromagnetic quantities as a usual boundary value problem of electrodynamics
and, thereby, to use in thermal radiation problems the full power of di€raction
electrodynamics. And most striking is the fact that the strict (di€raction) theory of
¯uctuation ®elds in electrodynamics can be reduced, in the most general form, to a
simple and delicate form of relationship between ¯uctuation (radiative) and
dissipative characteristics of physical media (Levin and Rytov, 1967). Following
the aforementioned authors, we shall ®rst consider the contents of the electrody-
namic FDT as applied to the electromagnetic ®eld and then the di€raction general-
ization of the Kirchho€ law.
So, as we have already noted, within the phenomenological theory framework
the ¯uctuation electromagnetic ®eld can be considered as the ®eld generated by
random detached currents spread in a medium. To calculate the energy character-
istics of the ¯uctuation ®eld, including spatial characteristics of the ¯uctuation
136 Thermal ¯uctuations and their fundamental laws [Ch. 4

illuminated radiation, it is necessary to know the spatial correlation of the spectral

amplitudes of these random currents, which, strictly speaking, constitutes the matter
of the electrodynamic FDT. Both in distributed and discrete systems, the FDT
allows us to associate the correlation functions of detached ®elds (currents) with
the dissipative properties of a system, which are laid down in macroscopic (linear-
ized) equations of the system. If these equations are Maxwellian (see section 1.6),
then they just determine the spatial correlation of detached electrical and magnetic
®elds, and in this case the dissipative properties of a medium will be described by
macroscopic constitutive equations of the medium (see section 1.6).
In applying the general theory of thermal ¯uctuations to the electromagnetic
®eld the conventional form of ®eld equations is utilized (see section 1.6 and equation
(1.1)). In this case the conductivity current and free charges are not separated from
the polarization current and polarization (i.e. in relations (1.1) quantities j and  are
supposed to be zero). As far as the ¯uctuation `forces' are concerned, they can be
expressed in di€erent ways, either as detached inductions, or as detached strengths,
or as detached currents. If we make use of the last approach, then the macroscopic
equations of the electromagnetic ®eld, to which FDT should be applied, are repre-
sented by the Maxwell equations of the form (in the Gaussian system of units):
@D 4
rot H ˆ ‡ j
@t c e
; 4:15†
@B 4
rot E ˆ j
@t c m
where je and jm are the detached ¯uctuation currents (electrical and magnetic), which
`cause' thermal ¯uctuations of all electrodynamic quantities. Many authors have
utilized various physical approaches in applying FDT to the electromagnetic
theory: the use of the discrete and continuous FDT forms, and the use of the
detailed equilibrium principle, as well as a number of indirect physical considera-
tions. The use of various physical approaches results in the following expression for
spatial spectral amplitudes of random currents for the isotropic medium:
j! !; T† 
jej r† E jek r 0 † ˆ ‰"_ r† "_ r 0 †Š jk  r r 0 †; 4:16†
82
where subscripts j and k denote spatial components j ˆ k ˆ 1; 2; 3†. Recalling the
expressions for the dielectric constant of a medium, this expression can be simpli®ed
and reduced to the form:
! !; T† 00
j je r†j2 ˆ " r†: 4:17†
42
Expressions (4.16) and (4.17) just constitute the matter of the electrodynamic
form of FDT, generalized to continuous dissipative systems, in application to the
electromagnetic ®eld. The physical essence of these expressions corresponds to the
basic matter of FDT, namely: the intensity of electromagnetic ¯uctuations (electro-
magnetic radiation) in a medium is immediately directly associated with macroscopic
Sec. 4.3] 4.3 Thermal ¯uctuations in the electrodynamics 137

dissipation properties in a system, which are re¯ected by constitutive equations of a

medium (see section 1.6). Of course, all spatial features of the medium under study,
re¯ected in constitutive equations (the inhomogeneity, anisotropy), will directly
stipulate the correlation properties of intensity of detached sources and, certainly,
the properties of the ®eld of thermal radiation of a studied physical object, which just
will be recorded by an external observer. The developed approaches are valid for the
homogeneous medium only under the condition of retaining the prerequisites for the
phenomenological description of the medium in Maxwellian electromagnetism. In
other words, in this case the medium can be partitioned into physically elementary
volumes, which are small when compared with macroscopic inhomogeneities of a
medium, but contain a great number of microparticles. Under these conditions the
statistical approach to the description of the state of a medium in such elementary
volumes is conserved. If the conditions of distribution of microparticles in these
volumes are close to equilibrium (see section 4.4) with some local temperature,
then formulas (4.16) and (4.17) can also be spread to a non-uniformly heated (none-
quilibrium) medium, provided that the temperature in the coecient  !; T† is a
function of a point.
However, from the viewpoint of observational remote sensing practice, the
expressions obtained are not sucient, since they determine the state inside the
medium, whereas remote sensing instruments record the electromagnetic radiation
escaped from the medium into free space, where the remote-sensing instruments
are situated. Since the detached currents are distributed over the whole volume of
an emitting body, the obvious method of calculating the external (`illuminated')
electromagnetic ®eld consists in using the regular methods of electrodynamics for
the region determined by the emitting body's form. However, because we deal
with the spatial distribution of detached currents in the medium volume, for the
overwhelming number of real media (or physical bodies) the formulation of a
complete electrodynamic problem can be very complicated. As we have noted
above (section 4.1), the intensities of the ®eld emitted by the medium (which are
of interest for us from the remote sensing point of view) can be found by the
more simple and physically transparent method developed by Levin and Rytov
(1967).
The essence of the method is as follows. The two ®elds are compared: the
radiation ®eld, which is recorded by an instrument in the external (with respect to
the medium studied) space, and the supplementary ®eld of the planar wave coming in
the direction in which the radiation intensity is of interest for ourselves. Those
di€raction ®elds should be taken as supplementary, which are formed at irradiating
the investigated medium by waves issuing from elementary dipoles situated at the
outer space points we are interested in. The application of the electrodynamic
theorem of reciprocity in combination with FDT results in the universal relations
between the spatial correlation functions of spectral amplitudes of the radiation ®eld
of the given medium, on the one hand, and the thermal losses (in this ®eld) of the
di€raction supplementary ®eld generated by the dipole situated at the observation
point. The considered approach is no longer bound by any limitations between the
characteristic body size, L, and the working wavelength,  (as in the case of the
138 Thermal ¯uctuations and their fundamental laws [Ch. 4

Kirchho€ law and the Nyquist formula), and allows us to ®nd any spatial character-
istics of the radiation ®eld at any distance from a body.
If by A r† and B r† are meant any two of six components of strengths E, H of the
thermal radiation ®eld, considered at two di€erent distances r1 and r2 from a body,
then the mean value of bilinear combinations of components of strengths of the
radiation ®eld is related with thermal losses of the total di€raction ®eld from
point sources QAB r1 ; r2 † as follows:
2
A r1 † E B r2 † ˆ  !; T† QAB r1 ; r2 †: 4:18†

And if the question is the intensity of radiation of the whole body volume, con-
sidered at one point of the outer space and having speci®c polarization, then (4.18)
can be transformed to the integral form:
2
jEp r†j2 ˆ ‰!; T r1 †Š dQEE r; r1 †: 4:19†
 V

where the integral is taken over the whole volume, V, of the emitting body (a
medium) with regard to the ®eld of temperature T r1 † non-uniformly distributed
inside a body. Here Ep r† is the projection of the ¯uctuation electrical vector at the
observation point P of the outer space with current radius-vector r on the direction
of the dipole moment, and r1 is the current radius-vector, determining the point of
position of the detached current inside the emitting body. The solution contains both
a wave (far) ®eld, carrying the energy away from a body, and a quasi-stationary
(near) thermal ®eld, which is concentrated in a layer adjacent to the body surface,
whose thickness is of the order of the working wavelength, and rapidly decreases
with the distance from a body. The quasi-stationary ®elds do not participate in the
energy transfer (see Chapter 5), but make their contribution to the volume density of
energy of the ¯uctuation ®eld, which sharply grows near the emitting body surface.
The di€raction e€ects, recorded in the far ®eld, as well as the detection and calcula-
tion of the quasi-stationary ®eld, represent principally new advantages of the ¯uctua-
tion electrodynamics as compared to the classical theory of thermal radiation (the
Kirchho€ laws).
Formulae (4.18) and (4.19) are, in essence, basic equations to the whole theory of
thermal electromagnetic ®elds. These formulae (sometimes called Rytov's formulae),
which relate the second moments of spectral amplitudes of the ¯uctuation ®eld with
thermal losses of the di€raction ®eld of point sources, can be considered as the
generalization of the classical expression of the Kirchho€ law to the di€raction
region (Rytov, 1966; Levin and Rytov, 1967).
Certainly, to ®nd the di€raction ®eld losses it is necessary, again, to solve the
appropriate electrodynamic problems by regular methods. However, these problems
are much simpler, than those considered above, where it was necessary to solve the
problem on spatially distributed detached currents, i.e. the currents distributed in a
complicated manner in the emitting body's volume. In some cases it is possible to use
for this purpose either the existing solutions of di€raction problems or the approx-
imate solutions, where some features of a speci®c problem are used (such as the local
Sec. 4.3] 4.3 Thermal ¯uctuations in the electrodynamics 139

application of geometric optics, the presence of a skin e€ect, the local scales of
surface roughness, etc.).
As an example, we shall consider the results of the solution of the afore-
mentioned problem for the situation, which is often encountered in remote
sensing practice. We mean the case where it is necessary to measure the intensity
of radiation of the absorbing half-space with a smooth boundary with a remote
sensing device standing outside of this space. Such a model situation is a basic
one in analysing any experimental data, obtained in sounding the Earth's surface,
and for this reason we shall repeatedly return to these results throughout this book.
Let the half-space z < 0 be ®lled with an isotropic conducting medium with a
complex dielectric constant, and in the region z > 0 the medium is isotropic too,
but it is transparent with the real index of refraction, n (see section 1.6). For the wave
®eld, i.e. the radiation ®eld, which can be recorded by the external (with respect to
the emitting medium) instrument, the solution of the ¯uctuation electrodynamic
problem should result immediately in the Kirchho€ law. In addition, the solution
will also contain the quasi-stationary ®eld components, which, however, very rapidly
decrease with the distance from the surface and make no contribution to the energy
¯ux. The experimental recording of such a ®eld is a rather complicated and
ambiguous problem. In accordance with the developed methodology (Levin and
Rytov, 1967), the electrical and magnetic dipoles with corresponding dipole
moments are placed at any point of the transparent half-space (over the planar
boundary), and then it is necessary to ®nd the di€raction ®eld losses in the
emitting half-space. This problem, called Zommerfeld's problem, is a classic one in
the problem of radio wave propagation over the Earth's surface (Stratton, 1941;
Alpert et al., 1953).
The complete solution of this problem leads to the following result: the power
characteristics of the ¯uctuation thermal ®eld in the far region (the radiation zone),
the Poynting vector in particular, do not depend on the distance from the medium
and can be expressed, for a ®xed direction and ®xed body angle, as the radiation
intensity (see Chapter 5 for more details) as follows:

I! ˆ I!0 n2 1 jRj2 †; 4:20†

where by I!0 is meant the equilibrium intensity formed inside the emitting body,
R is the Fresnel coecient of re¯ection of a planar electromagnetic wave from the
smooth boundary of the emitting medium (with account taken of polarization
and the angle of observation), and n is the index of refraction of a transparent
external medium. The expression presented is, in essence, the Kirchho€ law for
radiation of the absorbing half-space. The physical sense of this relation is fairly
transparent: the formed equilibrium intensity of an in®nite half-space undergoes
re¯ection at the planar interface boundary. In this case the value of the energy,
re¯ected inside the medium, will be equal to I!0 jRj2 . Thus, the energy will be
illuminated into the outer space and recorded by a remote instrument in accordance
with relation (4.20).
140 Thermal ¯uctuations and their fundamental laws [Ch. 4

4.4 LOCAL THERMAL EQUILIBRIUM

As we have already noted, FDT is valid for systems with thermodynamic equilib-
rium. In statistical physics by thermal equilibrium is meant the physical state, into
which any closed macroscopic system comes after a fairly long time interval has
elapsed. At thermodynamic equilibrium the detailed balance is established; that is,
any elementary process in a system is balanced by the corresponding reverse process.
The detailed balance takes place for the processes, which change the kinetic energy
and the direction of motion of both the macroscopic particles of a system and the
state of elementary particles, atoms, molecules and ions, and the state of their
excitation for the processes of ionization and recombination, dissociation and
formation of molecules, etc. At the thermal equilibrium state the parameters of a
system do not change in time; however, they can undergo thermal ¯uctuations about
their mean values. The thermal radiation arises under the detailed equilibrium con-
ditions in a substance from all non-radiating processes, i.e. from various types of
collisions of particles in gases and plasma, and from exchanging energies of electron
and oscillatory motions in liquids and solid bodies. From the detailed balance of
processes follows a spectrum of important physical consequences which are
expressed as theorems and laws. They include, ®rst of all, FDT, the Planck law of
radiation, the Kirchho€ law of radiation, the Stefan±Boltzmann law of radiation,
Boltzmann's distribution of particles over energies, Maxwellian distribution of
particles over velocities, the law of energy equidistribution a system's degree of
freedom, the ergodic hypothesis. In this case the temperatures, appearing in
formulae describing these laws and distributions, are identical in all parts of the
equilibrium system and for all sorts of particles, i.e. the temperature of the whole
system is meant here.
In the real physical reality, however, for the majority of physical bodies the
conditions of conservation of thermodynamic equilibrium are absent, generally
speaking. This indicates that any physical body emits from its surface some
speci®c portion of the electromagnetic energy which arises inside a body due to
physical-chemical reactions, to internal heat sources, to mass transfer inside a
body and to other causes. The outgoing energy ¯ux, which exists in such cases,
and, accordingly, the gradient (drop) of temperatures between internal and
external parts of a system are directly incompatible with the notion of full thermal
equilibrium.
The important supposition (hypothesis) on local thermal equilibrium (LTE) is
used for similar physical objects. According to this hypothesis, the temperature is
di€erent in di€erent elements of a studied medium; there exists the outgoing energy
radiation ¯ux (the radiation ®eld is anisotropic), but in this case equilibrium is
conserved in very small (elementary) volumes of a medium. But these volumes still
contain such a great number of particles (macroscopic particles, molecules, atoms,
ions, etc.), that their state can be characterized by the local temperature and other
thermodynamic parameters. In their turn, these parameters in macroscopic scales
are not constants, but depend on coordinates and time. But in each elementary
volume a detailed balance is established, which is determined by the local value of
Sec. 4.4] 4.4 Local thermal equilibrium 141

temperature, and in this local scale all physical corollaries of the detailed balance
(such as FDT, Boltzmann's and Maxwell's distribution laws, the Kirchho€ law of
radiation, etc.) are valid. At the local thermal equilibrium of a medium's elements the
state of the medium is one of nonequilibrium, in general. So, under these conditions
the thermal radiation is characterized by the value of the temperature at a given
point (locally), but the thermal radiation is not in thermal equilibrium with the
substance at the scale of the whole body (or medium) under study. In such a case
the emission of radiation into the external space and the redistribution of the
temperature regime inside a body (or medium) are possible. To maintain the station-
ary state, in which the gradient thermal ®eld is conserved, the thermal energy losses
must be replenished at the expense of extraneous (and, probably, internal) sources.
The reason for the application of the LTE hypothesis to physical objects both on
the Earth, and in space lies in the circumstance that the radiation absorbed by an
elementary volume of the medium is greatly reprocessed into di€erent forms of
energy before it leaves this volume (i.e. is illuminated). As is known from the
thermodynamics, such a reprocessing at the scale of the elementary volume
proceeds in the direction of establishing thermodynamic equilibrium. So, the
whole absorbed portion of radiation energy falling on the opaque solid body is
rapidly redistributed over internal energy states in accordance with the local equi-
librium distribution inside the solid body. In gases the absorbed radiation energy is
redistributed via various kinds of collisions between gas particles: atoms, molecules,
electrons and ions. In the majority of cases such a redistribution proceeds fairly
rapidly, and the energy levels of gas will be populated in accordance with the equi-
librium distribution corresponding to local conditions (Sobolev, 1997).
The local thermal equilibrium is a good approximation to reality for many
physical objects and their separate sections. Examples of such objects are: the
Earth's atmosphere, surfaces of the Earth, various astrophysical objects. The LTE
hypothesis greatly facilitates the calculation of radiation characteristics of such kinds
of media (using the so-called LTE models). Certainly, there exists a spectrum of
physical conditions in which the LTE assumption is invalid. Examples of such con-
ditions are: (1) highly rare®ed gases, in which the rate and eciency of collisions of
particles resulting in redistribution of absorbed energy are low; (2) very rapid non-
stationary processes with high gradients of parameters, in the course of which the
population density of energy levels has no time to come into correspondence with
new conditions; (3) extreme radiation ¯uxes, in which the absorption of energy and
the population density of the upper energy levels are so great, that, owing to collision
processes, the equilibrium population density of lower levels will not be achieved.
Giving the LTE hypothesis up (in the so-called NLTE models) it becomes necessary
to investigate the relations between collision and radiation processes and their
in¯uence on the energy distribution between various levels, which represents a
fairly complicated problem. Such investigations are carried out in studying shock
waves (large gradients), nuclear explosions (non-stationary processes, large
gradients, extreme ¯uxes), gas dynamics of ¯ights at high altitudes and in outer
space (very low densities). The greatest deviation from LTE conditions is observed
in laser and maser sources, in which the substance with a metastable energy level is
142 Thermal ¯uctuations and their fundamental laws [Ch. 4

excited by the external source. This is because the excited state is metastable and is
chosen such that the population density reaches values that essentially di€er from
equilibrium values (the inverse population density), and is then illuminated into the
external space in a coherent manner.
Such problems are of special interest and are not considered in this book. We
shall suppose here that the local thermal equilibrium exists in the media we shall
investigate later.
Applying the aforementioned radiation laws under local thermal equilibrium
conditions to emission and absorption of thermal radiation in physical bodies, we
can study radiation transfer processes both inside and outside the physical body,
within the framework of the so-called phenomenological theory of radiative
transfer (Chapter 9). The signi®cance of this theory for remote sensing problems
and astrophysical applications can scarcely be exaggerated. In fact, all fundamental
results in remote sensing (and, largely, in the astrophysics) obtained so far are based
to an overwhelming extent on the use of the methodology and interpretation of
conclusions of the theory of radiation transfer under local thermal equilibrium
conditions.