Optical Properties of Metals

M. PARKER
GIVENS
Universitg of Rochester, Rochester, New York

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
11. Electromagnetic Theory. .. ......... 316
1. Relations between the 0 ............. 316
2. The Flow and Conservation of Energy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318

2. Drude’s Theory.. .....

1. Drude’a Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336

2. An Alternate Method., . . . . . . . . . . . . . . . . . . . . . . . . . 338
3. Schulz’s Method.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . 342

2. Copper, Silver, and Gold

. . . . . . . . 348

5. Aluminum.. ..
........................

1. Introduction
I n everyday experience the optical properties of metals are associated
with high reflectivity and low transmission. I n a more quantitative way
the optical properties of metals are given in terms of two constants: the
index of refraction, n, and the absorption coefficient k. The two constants
may be defined by the following equations. First
c
n 3 -
2,
313
314 M. PARKER GIVENS

where c is the velocity of light in vacuum and u is the phase velocity in

the metal of a plane wave having constant amplitude along a wave front.
This is the definition usually given for the refractive index of a dielectric ‘
substance, along with the added condition that the amplitude of the
wave be constant. The absorption coefficient Ic is defined by the rate of
decrease of amplitude E of the electromagnetic wave as it progresses
through the metal
E = E0e(-2rk/X)~. (1.2)
Here z is the coordinate in the direction of propagation, Eo is the ampli-
tude a t z = 0, and X is the wavelength in vacuum. The constancy of
amplitude over a wave front is still assumed.
Together these two constants form the complex index N of the metal.
This is defined by the equation
N E n - ik. (1.3)
Many writers prefer to replace the absorption coefficient k by the
extinction coefficient K . The two are related by the simple equation
k = nK. Equation (1.2) then becomes
E = Eoe(-2rK/hn)8

where A, is the wavelength in the metal. The complex index becomes

N = n(l - i~).
The extinction coefficient K has been mentioned here for completeness.
I n order to avoid confusion, the absorption coefficient k will be used
exclusively throughout this article.
The term “complex index” for the quantity defined in Eq. (1.3) is an
apt one. If the index in the equation
8 = Eoeiw[t-(n~/~)l

for transparent media, is replaced by the complex index N , the resulting

equation
8 = EOel-2rk/X)zeio [t-(ns/c)l
(1.4)
has the properties required by the defining Eqs. (1.1) and (1.2). Here
8 is the electric field as a function of z and t. This concept is convenient
because it enables some of the formulas of dielectric optics to be converted
easily to the metallic case. For example, the reflectivity of a glass is
I(n - l)/(n + l)12. Inserting the complex index, the reflectivity R of a
 OPTICAL PROPERTIES O F METALS 315

metal is found to be‘

(n - 1)’ + k2
R =
+
(n 1 ) 2 + k2’

Care is required. For example Eq. (1.5)is obtained from the formula for
+
glass only when the latter is written in the form I(n - l)/(n l)lz. The
expression (n - l)”/(n +
l)’, which is equally valid for the case of glass,
will not give the correct expression when converted to the situation for
metals.
The conversion of optical formulas to the metallic case in this way
provides little or no insight into the physics involved and so provides

FIQ.1. Refraction of light at a plane boundary.

no aid in understanding or interpreting the equations. It is preferable,

therefore, to treat the metallic case directly. Transparent media can
then be considered as a special case in which k = 0. This procedure is
straightforward and involves none of the risks described in the previous
paragraph.
I n defining the index of refraction and absorption coefficient of a metal
it was specified that the amplitude was constant along a wave front. This
is usually the case only for normal incidence. For example, if a plane light
wave of uniform intensity in air (or vacuum) strikes a metal surface
obliquely as in Fig. 1, a given part of the refracted wave front will pene-
trate the metal by varying amounts depending upon whether it is among
the earlier or later parts of the wave front to strike the surface. I n the
figure, A B is a homogeneous. wave front moving in air which has just
touched the metal a t A ; CD is the game wave front a short time later
when the disturbance formerly a t B has just penetrated the surface of
the metal. The wave front a t D has not been attenuated by passing
through metal whereas the disturbance a t C has traveled a distance AC
l G. 5008, “Theoretical Physics.” English translation by I. M. Freeman, 2nd ed.,
p. 352. Hafner Pub. Co., New York and Blackie & Sons, Glasgow.
316 M. PARKER GIVENS

through the metal and thus been reduced in amplitude. Therefore the
amplitude decreases exponentially from D to C even though the line CD
lies in the wave front. Thus the planes of constant amplitude are parallel
to the surface of the metal and not parallel to the wave front, except in
the special case of normal incidence. We designate by a the angle between
the planes of constant phase (wave fronts) and the planes of constant
amplitude. It is shown readily that electromagnetic waves of this type
form a satisfactory solution of Maxwell’s equation provided the phase
velocity and absorption coefficient are suitable functions of a. Suppose
the index of refraction n(a) and k ( a ) are,defined by Eqs. (1.1) and (1.2),
along with the specification that the planes of constant phase and con-
stant amplitude make the angle a , not necessarily zero, with each other.
Then it can be shown2 by electromagnetic theory that these quantities
are related to n and k, the usual optical constants, by the equations
n(a)k(a) = n(O)k(O) = nk
nz(a)- k 2 ( 4 = nz(0)- k2(0) E n2 - k2. (1.6)
cos2 a
~

If k = 0 (transparent media), n(a) is independent of a and Snell’s law

holds. For a metal, n(a) is dependent upon a and Snell’s law holds only
to the extent that n(a) is constant, even though the medium be isotropic.
The optical constants are dependent upon the frequency or wave-
length of the incident light.

II. Electromagnetic Theory

1. RELATIONS
BETWEEN THE OPTICALAND ELECTRICAL
CONSTANTS
If the subject of this article had been approached from the viewpoint
of electromagnetic theory, it would have been normal to begin by present-
ing Maxwell’s equations for an uncharged conductor:

Curl x = --
e a s + -8
c at c

Curl8 = -- ax
-
c at
div 8 = 0
div 3c = 0
where & and 3C are the electric and magnetic fields, e is the dielectric
constant, p the permeability, u the conductivity, and t the time. An
uncharged, conducting medium is assumed, and cgs units are used. At
2 A. I. Mahan, J . Opt. SOC.Am. 46, 913 (1956).
 OPTICAL PROPERTIES OF METALS 317

optical frequencies p = 1 for all substances and so may be ignored in the

calculations.
The magnetic field X? may be eliminated between the first two equa-
tions by differentiating the first with respect to t and taking the curl of
the second. The result is

This is the differential equation of a wave motion; the following equa-

tion is a solution:
= Eoeiwte*i.\/.o'-i4*aw(o/c). (2.6a)

The application of Eq. (2.2) requires that

X? = H o e i W t e ~ i . \ / ~ ' - ~ 4 * u w ( ~ / c ) (2.6b)
where lHol = lEol I(n - ik)l. It is also required that Ho be perpendicular
to the direction of propagation and to Eo. The factor (n - ik) indicates
that X differs in phase from E by tan-' (kin).
Equations (2.6) represent a plane sinusoidal wave traveling parallel
to the z axis and having the planes of constant phase parallel to the
planes of constant amplitude. Spherical waves and plane waves for which
a # 0 are among the other possible solutions.
The equation above takes on the conventional form
= Eoe'O[MNa/~)I
(2.7)
which is identical with Eq. (1.4) if
N = n - ik = 1/E - i ( h u / w ) .
(2.8)
Equation (2.8) relates the optical constants n and Ic with the electrical.
constants c and u. Since it is a complex equation it may be reduced to two
equations, which are
n2 - k2 = E
and (2.9)
nkw = ~UU.

An experimental verification of these two equations would confirm the

applicability of Maxwell's equations to this problem. Unfortunately the
A complex dielectric constant e, or a complex polarizability aCmay be introduced
as an alternate approach.
ec = 1 + 47rao =e + zw
47r.Y
-*

Its relation to the other constants used is apparent.

318 M. PARKER GIVENS

electrical constants cannot be measured directly a t optical frequencies

and can be obtained only by applying Eqs. (2.9) to the optical constants.
OF ENERGY
2. THE FLOWAND CONSERVATION
It is instructive a t this point to demonstrate that the energy absorbed
from the light wave is converted into heat and that the classical Joulean
heating accounts for the entire energy conversion.
If Poynting's vector is used to calculate the energy flow, Eqs. (2.6)
or the equivalent derived from Eq: (1.4)lmay be used. Poynting's equation
may be taken in the form
C
s=-€XX (2.10)
4.r
if the real parts of E and K are inserted. We find

(2.11)
and, upon taking the time average over one cycle,

(2.12)

Notice that Eq. (2.12) contains n (which represents the phase velocity)
but is independent of anlaw (or the group velocity). This subject was
treated in a more detailed way by Brillouin.s
The time average of the energy removed from the wave per unit
volume per unit time, W (ergs/cm* sec), may be calculated as follows:

(2.13)

This is to be compared with the rate of production of heat.

The Joule heat produced per unit time and unit volume in a conduct-
ing medium W' is the product of the electric field and the current density.
The current density is the conductivity u times the field. Again, the real
part of Eq. (2.6) or Eq. (1.4) must be used. Thus

the time average of which is

(2.14)

8 L. Brillouin, Congr. intern. EEec. 2, 739 (1932).

 OPTICAL PROPERTIES OF METALS 319

I n view of Eq. (2.9), this is the same as Eq. (2.13). Thus the energy lost
from the beam is just accounted for in terms of the heat produced
electrically.

111. Classical Attempts to Predict the Optical Constants

I n the previous paragraphs, the discussion has revolved around the
optical properties of metals when expressed in terms of the macroscopic
constants n and k or e and u. When Maxwell’s equations are employed,
these constants are considered as experimentally measurable quantities.
Moreover they have been measured for many metals, as will be discussed
later on. It is natural in physics however to hope to derive such macro-
scopic constants from other constants or from first principles. A variety
of attempts have been made in this direction and the more important
will occupy our attention for the next few pages.

1. THEHAGEN-RUBENS
RELATION
The Hagen-Rubens4 relation for long wavelengths was developed
theoretically by Drude6 and confirmed experimentally by Hagen-Rubens.
The relation states that

This is demonstrated most readily by assuming that the conductivity u

has the direct current value uo, and that the currents in the metal are in
phase with the electric field. This second assumption is equivalent to
assuming that e = 1. When these conditions are used, Eqs. (2.9) become

and (3.2)

If n2and k2 are large compared to one, we see that they are approximately
equal. Equation (3.1), then follows immediately. Substituting Eq. (3.1)
into Eq. (1.5) for the reflectivity a t normal incidence, and taking account
of the fact that 1 << ( u o / Y ) , we find

This form of the relation was tested experimentally by Hagen and

Rubens.
E. Hagen and H. Rubens, Ann. Physik 11, 873 (1903).
6 P. Drude, Ann. Physik 14, 677 and 936 (1904).
320 M. PARKER GIVENS

It follows from Eq. (2.12) that the intensity falls off exponentially in
the manner e(-'/"o) where

The assumption that the current is in phase with the field is equivalent
to the assumption that the period 1 / w is long compared with the relaxa-
tion time of the electron in the metal.
2. DRUDE'STHEORY
Drude's treatment was more fundamental and was based on the
assumption that metals contain free electrons which experience viscous
damping and no other forces except the applied electric field. This situa-
tion is described by the following equation of motion:
d2Y
m-
dt2
+ my -
dY = -eEoeiot
dt (3.5)

in which -e and m are the electronic charge and mass, y is the coordinate
describing the position of the electron (the electric vector of the light
wave is selected t o be the y direction), and y is a constant describing the
viscous damping. The damping arises from the collisions of the electron
with the lattice. The steady-state solution of this equation is

The conductivity is the ratio to & of the component of the electric current
density in phase with the electric field and therefore is the real part of
noe d y
EoeiVt dt

where no is the number of free electrons per unit volume. This gives

Similarly, the polarizability a is obtained by dividing the component of

the current which is 90' out of phase with the field, or the imaginary part
of the ratio, by the angular frequency w . Thus

The dc values of these quantities are obtained by putting w = 0:

(ro = -.
noe2
(3.9)
my
 OPTICAL PROPERTIES OF METALS 321
From this, the value of the constant y may be estimated to be lo+’* sec-l.
The reciprocal of y is the relaxation time of the electron gas, which is
frequently denoted by the symbol r . I n like manner

(3.10)

Equations (3.7) and (3.8) show that for frequencies sufficiently low,
w2 << y2, u, and a may be replaced by their dc values. Since u = ay,
a w is small compared with u in this Bame spectral region. This means
that the current is in phase with the field and the assumptions of Eq.(3.2)
are justified.
Drude visualized the electrons as having a mean free path comparable
to the interatomic distance and assumed that the effective number of
free electrons would be equal to the number of valence electrons. The
advent of the Pauli exclusion principle, Fermi-Dirac statistics, the band
theory of solids,6 and the concept of “effective mass” has modified the
picture in some respects, but has not altered the predicted relations
between the optical constants and the dc conductivity.
As we shall see later, the optical constants of many metals show a
dependence on frequency in the infrared (A > 10 p ) which approximates
the form of Eq. (3.1). I n most cases, however, the dc conductivity must
be divided by a number between 3 and 10.This agreement should not be
expected until X > 100 p for good conductors.
3. SHORTWAVELENGTHS: THE FORMULA OF ZENER
At the short wavelength end of the spectrum the electron is assumed
to oscillate back and forth along a path much shorter than the mean
free path under the influence of the alternating field. The number of
collisions per unit time is then small compared to the frequency and a
negligible amount of energy is absorbed by the electrons in the metal.
It follows that the conductivity is zero. Since 2 m k w = u [Eq.(2.9)], we
conclude that one of the optical constants must be zero. Two cases
should be distinguished. I n the first, a > 0, so that k = 0 and n = 4.
I n the second, z < 0, n = 0, and k = G.
The free electron formulas of Drude [Eqs. (3.7) and (3.8)] may be
applied? to this case by assuming y 2 << w2.We find
a = - - noe2 (3.11)
mw2
(3.12)
3. C. Kittel, “Introduction to Solid State Physics,” Chapters 10 and 11. Wiley,
New York, 1956.
’ c. Zener, Nature 132, 968:(1933).
322 M. PARKER GIVENS

Since the dielectric constant e equals 1 + h a , its value will be positive

when
noe2
47r-
mu2
<1 (3.13)

and negative when the inequality is reversed. The first of these cases
applies to the high-frequency or short wavelength side of a critical
wavelength A,, defined by the relation

(3.14)

It corresponds to a transparent medium having negligible absorption.

The opposite case, in which A > A,, describes total reflection at the metal

2.8.
cz'"
'
'
4'
P
2.4-E 2 c
0 "
* Y '4
,#'
2 . 0 4 2 kr'
/'
t 1.6- r'
X5 'f'
$1.2 #'

DUNCAN a DUNCAN
+ WOOD
Q .4 IVES a BRIGGS
--- CALCULATED
0
1200 2800 4400 6Ooo
WAVE-LENGTH IN ANGSTROMS
FIG.2. Theoretical and experimental values of the optical constants of sodium in
the region near the critical wavelength (after Ives and Briggs).

surface [see Eq. (1.5)]. The electric field within the metal decreases
exponentially with the distance from the surface, just as for light totally
reflected at a glass-air boundary. I n the present case, however, the
total reflection takes place for all angles of incidence, including normal
incidence.
The fact that the conductivity, given by Eq. (3.12), is not strictly8
zero, but is only small, in the first case implies a very slight absorption.
In the second case, the reflectivity is actually less than 100%. The
* G. Grass, 2.Physik 139, 385 (1954).
 OPTICAL PROPERTIES O F METALS 323

remaining light enters the metal and is absorbed strongly. Thus there is a
highly damped wave in the metal. The two cases are easily distinguished.
(See Fig. 2.)
The alkali metals conform reasonably well to the simple free electron
picture. R. W. Woods found that these metals become transparent in
I. COMPARISON
TABLE VALUESOF
OF THE OBSERVEDAND CALCULATED THE
CRITICALWAVELENGTH FOR TRANSMISSION

Metal Observed Calculated

CS 4400 3600
Rb 3600 3200
K 3150 2900
Na 2100 2100
Li 2050 1500

the ultraviolet and measured the wavelength a t which the transparency

begins. Table I gives a comparison of the observed critical wavelength
with that calculated by Eq. (3.14).
4. INTERMEDIATE WAVELENGTHS
If the approximations of Eqs. (3.11) and (3.12) are valid, i.e., y2 << 02,
and if the wavelength is still long compared to A,, i.e., o <<a,, then
Eqs. (3.9) and (3.12) combine to give
g = - noe2y = (n~)~'.
(3.15)
m 2 ma Qo
Equation (3.11) may be rewritten in the form

€ = I - & - . noe2 (3.16)

mo2
These quantities may be determined from experimental values of the
optical constants. Forsterling and Freedericksz'O made such measurements
for several metals in the region from 1 p to 15 p . The present discussion
is appropriate to the half of their experimental range associated with
shorter wavelengths. It was found that the observed values had the
dependence on frequency predicted by Eqs. (3.15) and (3.16). The experi-
mental values of agreed with those predicted by Eq. (3.16) within
R. W. Wood, Phys. Rev. 44, 353 (1933).
10K. Forsterliig and V. Freedericksz, Ann. Physik 40, 200 (1913).
324 M. PARKER GIVENS

about 10%. The values of u were in disagreement by a much larger factor

unless the values of uo employed were smaller than the values measured
directly. The agreement between Eq. (3.16) and the experimental value
of e may be improved slightly by replacing m by an effective mass m*
having a relative value slightly larger than unity. The extent of these
changes is indicated in Table 11.
As was pointed out by Wilson“ the optical properties of metals are
determined by a thin surface layer. The penetration distance 61, required

TABLE11. COMPARISON
OF THE INFRARED MEASUREMENTS FORSTERLINQ
ANDFREEDERICKSZ WITH THE DC VALUES
Schulz has new values of m*/m for Cu, Ag, and Au; see part VII, Section 2.

Ag Au cu Pt Ir

Value of co for best fit 1.4 2.5 1.0 0.12 0.13

Actual value of u0 5.7 4.2 5.3 0.85 1.7
m*/m 1.07 1.13 2.56

00 is in units of 10” sec-l.

if the light intensity is to fall to l/eth of its initial value, is

aI =
x
-. (3.17)
47rk
For sodium metal at = 5896 (Dlight) k is 2.6 and 61 is only 180 A. It is
probable that the conductivity uo is considerably below the bulk value
in a surface layer this thin because of contamination and surface defects.

5. LJa = y2
In the region in which u2 is comparable with y 2 it is not permissible
to introduce approximations; Eqs. (3.7) and (3.8) must be used in their
complete form.
It is interesting to notice that, if the conductivity u = nkv, is plotted
as a function of the dielectric constant E = n2 - k2,Eqs. (3.7) and (3.8)
predict that the result will be a straight line having a negative slope
--y/47r. This prediction involves no approximations and therefore should
be valid over the entire spectral range. Several experimenters have pre-
sented their data in this way in order to show the extent t o which the
observations support the theory. Figure 3 shows one of the best of these
plots as an example. The fact that the experimental points follow a
11 A. H. Wilson, “The Theory of Metals.” Cambridge University Press, London and
New York, 1936.
 OPTICAL PROPERTIES OF METALS 325

straight line does not indicate that the conductivity approaches the bulk
value for low frequencies.
Although the preceding discussion has not made use of quantum
mechanics, this extended form of mechanics is needed to give meaning to

FIG. 3. Diagram showing 2nk/i against na - k2 for evaporated aluminum. The

Drude theory predicts a straight line, such as the dashed one. The numbers beside
the experimental points represent the wavelength in microns (after Beattie and Conn).

the term effectivem a d which was introduced into Table 111. In the main
the results of the quantum-mechanical treatments of the optical prop-
erties have been the same as those obtained from the classical free electron
picture and will not be discussed here. There are, however, a few impor-
TABLE111. COMPARISON OF THE EXPERIMENTAL
ABSORPTIVITYWITH THE
VALUESPREDICTED THEORY
BY THE CLASSICAL AND THE ANOMALOUS SKIN
EFFECT WITH DIFFUSEELECTRON REFLECTION(FROM DINGLE)

Near infrared absorptivity (in %) in copper

Anomalous skin effect

Classical with diffuse electron
Temperature Experiment theory reflection

Room 17°C 1.20 0.5 0.8

Liquid oxygen - 183°C 0.8" 0.09 0.5
Liquid helium 4.2"K 0.6b 0.003 0.4

Experimental data by K. Weiss, Ann. Physik 2, 1 (1948).

* Experimental data by K. Ramanthan, Proc. Phys. SOC.(London) A66, 532 (1952).
326 M. PARKER GIVENS

tant contributions made to our understanding of the optical problems by

quantum mechanics. We shall consider these now.
IV. The Internal Photoelectric Effect
In the process known as the internal photoelectric effect, energy is
absorbed from the incident light wave to raise an electron to an excited
state which is unoccupied in the normal condition of the metal. I n the
visible part of the spectrum, a valence electron is excited into a higher
unoccupied state. in this way. The same process is found in the vacuum
ultraviolet and x-ray regions; however, the transition usually raises an
electron from a lower lying, or an inner, state to a state above the Fermi
level in the solid. The valence and higher states in a solid are grouped into
bands of appreciable width, of the order of several electron volts, so that
the absorption extends over a correspondingly broad region of the spec-
trum. The energy absorption is equivalent to an increase in the conduc-
tivity and this in turn changes the optical constants at the corresponding
frequencies. This process takes place in addition to the absorption pre-
dicted on the Drude free electron picture. Hence the absorption arising
from this process is often termed anomalous.
This problem may be treated with the use of time-dependent perturba-
tion theory12 which gives the following equation for the specific case of
absorption of electromagnetic radiation:

Here P represents the probability of a transition during the time t ,

when the light is applied at t = 0; I is the intensity of the light, v the
frequency of the light, and vm is the frequency associated with the energy
required to make the transition from the initial to the final state, i.e.,
hv, = W,,, - Wo where Wm and Wo are the energies of the final and
initial states. If the light is polarized with the electric vector in the z
direction and if the wavelength of the light is large compared to inter-
atomic distances,

Here $, and $0 are the wave functions describing the final and initial
states, respectively.
If v has a fixed value, the probability P increases with the square of
the time for short times. This behavior is characteristic of the method
I* See, for example, D. Bohm, “Quantum Theory,” Chapter 18. Prentice Hall,
New York, 1951.
 OPTICAL PROPERTIES OF METALS 327

and is usually removed by considering nonmonochromatic radiation and

integrating P with respect to frequency. The value of P is small except
when v is very close to v,. The integrated value increases linearly with
time according to the following equation when the incident radiation
includes the frequency v,:

Here I ( v ) d v is the intensity in the spectral region between v and v+ dv;

its value at v , appears in Eq. (4.3). P is zero if v = v , is not included in
the incident light.
In the case of solids, the permitted electron energies, and hence the
frequencies v,, form a practically continuous band. Thus it is necessary
to sum (4.3) over all possible pairs of states for which the energy differ-
ence corresponds to the frequency v , if one desires to determine the total
probability by this route. An alternative, and preferable approach is to
remove the t2dependence appearing in Eq. (4.1) by regarding the incident
light as monochromatic and summing (4.1) over all possible transitions
between filled and empty electron states. The states are so numerous
that this summation amounts to an integration. The results obtained
by either method require that the probability given by Eq. (4.3) be
multiplied by a factor representing the number of possible transitions
+
in the range of frequencies between vm and vm dv. It is assumed that
[arn0lis the same for all these transitions. This factor includes the product
of the densities of the initial and final states. These densities depend in
turn upon the number of atoms in the solid.
The wave functions t)mand $0 may be represented-by Bloch functions
UmLeilr.Iand Uok~eiY*‘ in which the functions U have the periodicity of the
lattice and the quantities k and k’ represent the propagation vectors of
the electron wave functions. The latter should not be confused with the
absorption coefficient k, a scalar. Using the Bloch functions in Eq. (4.2),
we find that the value of a m o is zero unless
k = k’ f 2?rh (4.4)
where h is the vector between any two points in the reciprocal lattice.
The vector h simply allows both k and k’ to be kept within the first
Brillouin zone. Equation (4.4) states that only “vertical” transitions in k
‘
space are permitted. This selection rule is equivalent to the conservation
of momentum for the electron, the momentum of the photon being
negligible in comparison.
The evaluation of Iamolrequires that the initial and final wave func-
tions be known. If this is not the case, it may still be possible to make a
328 M. PARKER GIVENS

qualitative prediction of the minimum frequency which can produce

absorption by the internal photoelectric effect. I n a metal the valence
band (see Fig. 4) is not completely filled, the electrons occupying only
the states of lowest energy up t o the Fermi level, which corresponds t o
the point A . Since only vertical transitions are permitted, an electron in a
filled state such as C may, upon absorbing light of the correct frequency,

FIG.4. Energy as a function of electronic wave vector k in a simple one-dimensional

crystal. Several powible transitions are indicated.

go to state C' or state C". The lowest frequency which can be absorbed
corresponds to a transition from A to A'. There then is a continuum of
absorbable frequencies until we reach the transition B to B', which has
the largest energy of any transition between the valence and second
bands. There then is a nonabsorbing region of frequencies until the
energy reaches BB", which is the lowest value for permissible transitions
between the valence and third bands. We have assumed the third band is
steeper than the valence band. There is a minimum frequency for absorp-
tion even in three dimensions, for which the energy bands overlap, since
the energy bands are always well separated for any fixed value of k.
 OPTICAL PROPERTIES OF METALS 329
However, the minimum frequency does not, always correspond to a
transition of the most energetic electron of the valence band, i.e., the
electron at A . The second band may be concave upward and steeper
than the valence band. I n such a case, the minimum frequency corre-
sponds to a transition of the electron having k = 0.
As in the atomic case, the value of J(Y,,,o~ for any transition depends
upon the symmetry of the wave functions involved. The usual atomic
selection rules, such as the one on the aximuthal quantum number,
A1 = Ifil

which arises from spherical symmetry, cannot be applied strictly to the

case of optical transitions in solids, since most wave functions have a
mixture of azimuthal quantum numbers. In any case the value of Io,,,O[
depends upon the mixture.
WolfeI3has given a theoretical treatment of the internal photoelectric
absorption of metals containing a small amount of impurity in an other-
wise perfect crystal. The treatment indicates that the impurities make
additional transitions permissible. Presumably thermal agitation can
serve in a similar way.
V. The Anomalous Skin Wect
A comparatively recent advance in understanding the optical prop-
erties of metals has been provided by the anomalous skin effect. It was
pointed out earlier that light penetrates into the metal only a short
distance; therefore, the surface conditions are very important. The
situation in which a light wave falls upon a metal at normal incidence
is analogous to that surrounding a wire carrying a high-frequency current.
The electric and magnetic fields are parallel to the surface and the
Poynting vector delivers energy to the metal in regions where the energy
is converted to heat. Such conversion takes place in a thin layer near the
surface. Therefore it is permissible to treat optical reflection by equations
intended for the skin effect if the radius of the conductor is allowed to go
to infinity. Such a treatment does not in itself provide any new informa-
tion, but corresponds to the Drude treatment at long wavelengths, i.e.,
the Hagen-Rubens relation.
The skin effect at radio-frequencies is presented in the standard textsI4
on electricity and magnetism. It is shown that for large conductors the
13 R. Wolfe, Proc. Roy. Soc. (London) A67, 74 (1954).
14 See, for example, M. Abraham and R. Becker, “Classical Electricity and Mag-
netism.” English translation by J. Dougell, p. 199. Stechert, New York, 1932; or
W. K. H. Panofsky and M. Philips, “Classical Electricity and Magnetism,” p. 182.
Addison Wesley, 1955.
330 M. PARKER GIVENS

electric field decreases exponentially from the surface with a penetration

distance

At long wavelengths this is equivalent t o Eq. (3.17), as can be seen if we

remember that the distance (3.17) is valid for intensity whereas (5.1) is
valid for the electric field.
The anomalous skin effect, for which the classical treatment is inade-
quate, exists whenever the mean free path of the electrons is not small in
comparison with the wavelength and the penetration distance. I n this
case, the current density a t any point is determined not only by the
electric field there, but also by the motion of electrons which arrive there
from other places at a distance less than or comparable to the mean free
path. Such electrons bring with them velocities acquired in regions
where the electric field has different values. Under these conditions, it is
not possible to substitute the product of conductivity and electric field
for the current density in Maxwell's equation (2.1) as we have done in
the past.
These effects alter both the penetration distance and the reflectivity.
The conductivity, as calculated from the measured optical properties, is
no longer expected to agree with the bulk value in direct current even
if all surface defects could be removed. The discrepancy between the two
determinations of the conductivity should be increased by cooling the
metal to low temperatures since this increases the mean free path.
The situation is complicated further by the proximity of the surface.
Regardless of its mean free path, the electron is turned back when it
reaches the surface. An electron may be turned back by the simple
process of specular reflection at the surface, in which case the component
of velocity normal to the surface changes sign and the other components
remain unchanged. Another possibility is that the electron is reflected
diffusely from the surface; in this case the reflection process may bring
the electron into thermal equilibrium with the crystal and the reflected
electrons may have a Fermi distribution of velocity. Intermediate cases
in which some electrons are reflected specularly and others diffusely are
conceivable and have received quantitative attention. The calculations
for the diffuse case agree with the experimental data much better than
the results obtained from either specular reflection or the classical
treatment.
The anomalous skin effect was first suggested by Pippard.'" It was
16 R. B. Dingle, Physica 19, 729 (1953).
16 A. B. Pippard, Proc. Roy. SOC.(London) A191, 385, 399 (1947).
 OPTICAL PROPERTIES O F METALS 331

treated quantitatively by Reuter and Sondheimer" and later by other

The latter authors have incorporated the internal photo-
electric effect into their treatments; however, the initial assumptionR and
results are otherwise comparable.
For purposes of this discussion, it is convenient to introduce into the
classical picture a complex conductivity ue, where
uc = u + iwa. (5.2)
With this substitution, u,E represents both the usual density of conduc-
tion current and the polarization current density, which is 90' out of
phase with the applied field. It is assumed that the displacement current
density arising from dE/at may be neglected. It follows from Eqs. (3.7),
(3.8), and (3.9) that

+ iw.)'
QO
u, = (5.3)
(1
The classical electric wave is then represented by

E(z,t) = Egeiwte(--l+i)/I)(l~""fl (5.4)

where the relaxation time 7 = ( l / ~ ) .
Equation (5.4) is not valid under the conditions which produce the
anomalous skin effect because it depends upon the assumption that the
current density J in Maxwell's equation may be replaced by an appro-
priate constant times the electric field, which is not the case. It is neces-
sary to calculate J in another way, by choosing a model for the metal and
considering the motion of the electrons in detail. If it is possible to dis-
cover a valid relationship between J and 8 , the new equation, together
with Maxwell's equations, may be solved for &.
Following the usual treatment of metallic conduction, the conduction
electrons are assumed to be free and the electron energy E, is related to
the electronic wave vector k by the equation

17 G. E. H. Reuter and E. H. Sondheimer, Proc. Roy. SOC.(London) A194 336 (1948).

l* T. Holstein, Phys. Rev. 88, 1427 (1952); 96, 535 (1954).
R. B. Dingle, Physica 19, 311, 348, 729, 1187 (1953).
20 V. L. Ginsburg and G. P. Motulewitsch, Uspekhifiz. Nauk 66, 469 (1955).
31 R. Wolfe, Proc. Roy. SOC.(London)A68, 121 (1955).
83 M. Ya. Azbel and E. A. Kmer, Zh. dksper. leor. Fiz. 29, 876 (1955); also Soviet
Phys. J E T P 2, 749 (1956).
88 M. I. Kaganov and M. Ya. Azbel, Dokl. Akad. Nauk SSSR lOa, 49 (1956).
1 4 J. G. Collins, Applied Scientific Research 7, 1 (1958).
332 M. PARKER GIVENS

in which m* is the effective mass of the electron. The electrons in equilib-

rium have a Fermi distribution function f o

where Er is the Fermi energy, k is Boltzmann's constant, and T is the

absolute temperature. As a result of the electromagnetic wave, the dis-
tribution function is altered to a new form f where

f = fo + fl. (5.7)
Here f o depends only on E, or IvI, whereas f l depends upon both v and z ;
v is the velocity of the electron and z is the distance from the surface of
the metal. f i is an unknown function which must be determined; it is
expected t o be small compared with unity.
A steady state is established under the combined action of the applied
electromagnetic field and the collisions of the electrons with the lattice.
The distribution function in the steady state is determined by the follow-
ing equation:

df
at
= 'd+ (8 + 'T) gradkf - v * grad,f. (5.8)

The first term on the right is a result of collisions of the electrons with
the lattice, the second term originates in the action of the electromagnetic
field on the electrons, and the last term is a consequence of the diffusion
of electrons from one place to another. The medium is assumed to be
isotropic in the sense that 7 depends upon the absolute value of v only;
f is considered to be a function of the wave vector k and the space vector r,
both of which are related simply to v and z of the previous paragraph.
We shall use the relation hk = 2?rm*v and shall neglect the magnetic
forces, which are much smaller than the electrostatic forces. Moreover,
we shall replace gradk f by gradk fo, which is permissible if f~ is small.
We shall also assume that f l is a simple periodic function of the time
like the electric field. For normal incidence f l is independent of x and y,
so that Eq. (5.8) reduces to

for light polarized with the electric vector in the y direction. A solution
must be found for this equation using the appropriate boundary condi-
tions a t z = 0.
 OPTICAL PROPERTIES O F METALS 333
The first term in Eq. (5.9) arises from the motion of the electrons
in the z direction, which is ignored in the classical treatment. If the
classical solutions are regarded aa first approximations, it is possible to
estimate the conditions under which the first term is really negligible,
that is, the conditions for which the classical treatment is adequate.
The function f l should have the same dependence upon z as the field &.
I n the classical case, this is given by Eq. (5.4). It follows that, in order
of magnitude, C3fl/C3z is f l / (6(1 +
o ~ T ~ The
) ~ ) second term of Eq. (5.9)
+
.
is of the order fl(l o ~ T ~ ) + / Z where I = T[vI. Therefore, the first term of
Eq. (5.9) is negligible compared to the second if
1
-
6
<< (1 + o272)f (5.10)

6 itself is dependent on frequency [see Eq. (5.1)]. This inequality is

valid at very low frequencies because the penetration depth 6 is large
compared to the mean free path, the right-hand side remaining essentially
unity. At very high frequencies, the right-hand side becomes (wT)~. Even
though the penetration depth decreases with increasing frequency, the
left-hand side increases only as d.Therefore, the inequality is valid a t
very high frequencies. At these frequencies, the distance traveled during
one period of the electric field is small compared with the penetration
depth. Between these two extremes, there is a broad region of the spec-
trum which cannot be treated adequately by the classical method. The
results obtained in this region are of considerable interest.
The boundary conditions which must be imposed upon the solutions
of Eq. (5.9) may be represented as follows. Denote byflc1) that part of the
function f l which applies to electrons moving toward the surface (v, < 0)
and by fl(z) the part which applies to electrons moving away from the
surface (v, > 0). The boundary condition a t z = 0 required for specular
reflection is
f1(2)(vz, vy, v,, 2 = 0) = f1(1)(vn, vy, -v,, z = 0) (5.11)
and that for diffuse reflection is
f1(2)(v,, vr, v,, z = 0) = 0. (5.12)
The details of the solution may be found in the article by Reuter
and Sondheimer” and will not be reproduced here. The results indicate
that the electric field for intermediate frequencies does not die off expo-
nentially in the metal but takes some more complicated form. For this
reason the usual definitions of the optical constants are not applicable
and the surface impedance of the metal is calculated instead. The im-
pedance concept has been applied to optical problems by many previous
334 M. PARKER GIVENS

author^.^^-^' The surface impedance Z is defined as

(5.13)

in which the electric and magnetic fields are evaluated a t the surface.
Here d is the current passing through a plane surface which is perpendicu-
lar to E and is bounded by a line of unit length in the surface and two
lines normal to the surface extending indefinitely into the interior of the
metal; B is a complex quantity which includes phase information as well
as magnitude. I n most cases, numerical values of 2 can be obtained, for
example by numerical integration, much easier than the functional
dependence of E ( z ) can be determined.
Just aa the reflection from the end of an ac transmission line is deter-
mined by the input impedance of the device which terminates the line,
the reflection from the surface of a metal may be expressed in terms of its
surface impedance by the following equation

(5.14)

The term k / c represents the impedance of free space. This result may
also be obtained directly by imposing the usual boundary conditions
upon E and 3C a t the surface of the metal and treating (5.13) as an auxil-
iary condition relating 8 and X just inside the metal. Equation (5.14) is
equivalent to the classical formula [Eq. (1.5)] if
47r
n = -Real Part Z-l
C
(5.15)
it = Imag. Part 2-1.
C

Since the optical constants have no physical meaning in the present

discussion, the values given by Eq. (5.15) represent the apparent values
determined by reflection measurements.
I n the experimental tests of this theory, it is customary t o compare
the measured absorptivity, A , with the value predicted from the theoreti-
cal surface impedance, Z . A is defined by the relation
A = l - R . (5.16)
The following results are predicted by the theory.
*6 S. A. Schelkunoff, “Electromagnetic Waves.” Van Nostrand, New York, 1943.
2e P. J. Leurgans, J . Opt. Soc. Am. 41, 714 (1951).
27 B. Salzberg, J . Opt. SOC.Am. 40, 465 (1950).
 OPTICAL PROPERTIES OF METALS 335
a. The absorptivity is greater than that determined by the classical
Hagen-Rubens relation in the infrared. The result is

(5.17)

where A , is the absorptivity predicted classically and v' is the Fermi

velocity in the metal. This ratio is larger than unity a t room temperature
and is of the order of several hundred a t liquid helium temperatures.
Diffuse reflection of electrons has been assumed here.
b. The temperature dependence of the absorptivity a t low tempera-
tures and in the near infrared is
A = - (e X ) 4
(5.18)
ao(T) rm*
when specular reflection of electrons prevails, or

(5.19)

when diffuse reflection prevails. If A is plotted against a0-I (a function

of T),the slope of the curve provides an estimate of no/m*. Moreover,
the first term of (5.19) is usually much larger than the second for good
conductors, so that the absorptivity is almost independent of temperature
in the diffuse case and is strongly dependent upon temperature in the
specular case. The experimental data support the diffuse case; the classi-
cal theory and the specular case are in substantial agreement. Table I11
gives a comparison of the experimental values, the predictions of the
classical theory, and the recent predictions based on the assumption of
diffuse electron reflections and the dc conductivity measured a t the
temperatures in question. The experimental data agree with the present
theory much better than with the classical theory.
This theory also predicts a slightly different dependence on wave-
kength in the infrared than is predicted by classical theory.
Benthem and KronigZ8have extended the theory to include an energy
loss arising from interaction among the electrons. Such a correction is
important a t low temperatures and improves the agreement between
theory and experiment. Their work has been criticized by Ginsburg and
Silin.29
Note. I n a recent paper, Abelesaohas stated that he has an article in
presss1in which he discusses the optical properties from a purely quantum-
*8 C. W. Benthem and R. Kronig, Physiea 20, 293 (1954).
V. L. Ginsburg and V. P. Silin, Soviet Phys. JETP 2, 46 (1956).
80 F. Abeles, J . Opt. SOC.Am. 47, 473 (1957).
81 F. Abeles, Cahiers Phys. to be published.
336 M. PARKER GIVENS

mechanical viewpoint in the hope of removing the remaining differences

between theory and experiment.
VI. Experimental Methods
1. DRUDE'S
METHOD
I n addition t o contributing to the theoretical understanding of the
optical properties of metals, Drudea2played a leading role in developing
the early experimental techniques for measuring the optical constants.

x
MONOCHROMATIC
LIGHT \

POLAR 12ER

FIQ.5. The Drude method of measuring the optical constants.

His method deserves consideration not only because of its historical

interest, but because it is still used with only minor variations. See Fig. 5.
I n this method, a collimated beam of linearly polarized light having
its electric vector oriented a t 45' to the plane of incidence falls on the
metal surface to be studied a t an angle of incidence 0. The reflected light
is examined to determine the difference in phase between the components
for which the electric vector is parallel and perpendicular to the plane of
incidence. This phase difference is called A. The relative amplitude of the
two components is also determined and an angle #, called the angle of
restored azimuth, is defined in such a way that
E
tan # = 2.
E,
82 P. Drude, Ann. Phys. u. Chem. 82, 584 (1887).
 OPTICAL PROPERTIES O F METALS 337

Here E, is the amplitude of the electric vector perpendicular to the plane

of incidence and E, is the amplitude of the electric vector parallel to the
plane of incidence.
If 8, A, and # can be determined experimentally, the optical constants
of the metal are calculated from the following equation^^^^^^ which are
derived readily from electromagnetic theory
cos2 2# - sin2 21) sin2 A
122 - k2 = sin2 0 tan2 0
(1 +
cos A sin 21,b)~
+ sin2 8 (6.2)
sin 21) cos 2# sin A
nk =
(1 + sin2 8 tan2 8.
cos A sin 21,b)~ (6.3)

In these equations, the convention of sign has been selected so that

A = 180” at normal incidence and goes to zero (through the first two
quadrants) for large angles of incidence; # and 2# are always in the first
quadrant.
Other equations of a similar nature have been presented recently by
Ditchburnaband Price.3s
The phase difference A is 90’ for a value of 8, usually between 65”
and 75”. This value of 0 is called the “principal angle of incidence” and
is denoted by 8. The value of # associated with the condition is 8 = 8 is
called the “principal angle of restored azimuth” and is denoted by J..
If 8 and J. can be measured or interpolated graphically from measure-
ments at nearby angles, the calculating equations (6.2) and (6.3) can be
simplified somewhat by substituting sin A = 1 and cos A = 0. It is
usually impractical to determine the principal angle of incidence when
many measurements are to be made. However, it is desirable to work at
angles near 8, because the calculated values of the optical constants are
most sensitive to small changes in the experimental quantities 8, A, and $
in this region; therefore, greater accuracy can be obtained by working
near 8.
Approximate formulas are often q~oted.~78~8 The following are typical:
sin 8 tan 0 cos 2#
n =
1 4-sin 21) cos A
sin 8 tan 8 sin 2$ sin A
k =
1 +
sin 21) cos A
a3 H. Geiger and K. Scheel, Handbuch der Physik 20, 240-250 (1928).
34 J. Bor, Proc. Phys. SOC.(London) B66, 753 (1952).
36 R. W. Ditchburn, J . Opt. SOC.Am. 46, 743 (1955).
36 D. J. Price, Proc. Phys. SOC.(London) 68, 704 (1946).
37 F. A. Jenkins and H. E. White, “Fundamentalsof Optics,” 3rd ed., p. 523. McGraw-

Hill, New York, 1957.

3* R. W. Ditchburn, “Light,” p. 447. Interscience, New York, 1953.
338 M. PARKER GIVENS

These equations may be simplified further a t the principal angle of

incidence. The approximation used in obtaining these equations from the
+
exact ones assumes that sin2 0 << (nz k2). This is only a fair approxi-
mation, as can be confirmed readily by comparing the optical constants
crtlciilated by the exact and aptxoximate equations from a given set of
experimental data.
Experimentally, the sample is mounted on a spectrometer table so
that the axis of the instrument lies in the plane of the surface. Collimated
monochromatic light is provided and the angle of incidence 0 (near 8) is
measured in the usual way. A compensator of either the Babinet or
Soleil type is inserted in the reflected beam with the axes of the compensa-
tor parallel and perpendicular to the plane of incidence. When properly
calibrated and equipped with an analyzer, these compensators provide a
direct measure of the phase shift A. When the phase shift A has been
compensated in this may, the emergent light is linearly polarized. The
angle between the normal to the plane of incidence and the electric vector
of the linearly polarized light is the angle of restored azimuth $. This is
determined by observing the azimuth of an analyzer which has been set
to extinguish the light emerging from the compensator [0 I $ I ( ~ / 4 ) ] .
This is the method introduced by Drude and used with minor variations
by VoigtlasMinor40 and many, more recent workers.

2. AN ALTERNATE
METHOD
An alternate experimental method, which has become popular because
it is adaptable to photoelectric methods, consists in measuring the orien-
tation and axial ratio of the elliptically polarized light which is produced
by reflection from the metal surface. The compensator and analyzer of
the Drude method are replaced by a photoelectric cell in front of which
is placed a simple analyzer. As the analyzer is rotated about the direction
of the light as its axis, the position of the analyzer corresponding to
maximum output of the photocell gives the orientation of the major axis
of the ellipse. If the output of the photocell is linear in intensity, the ratio
of minimum to maximum output is the square of the axial ratio of the
ellipse. The method, simple in principle, is subject t o varying degrees of
refinement and a u t ~ m a t i o n . ~ ~ . ~ ~
From measurements such as those just described, the experimental
quantities appearing in Eqs. (6.2) through (6.5) may be determined by

89 W. Voigt, Physik. 2.2, 303 (1901).

40 R. S. Minor, Ann. Physik 10, 581 (1903).
41 J. F. Archard, P. L. Clegg, and A. M. Taylor, Proc. Phys. SOC.
(London) B66, 758
(1952).
 OPTICAL PROPERTIES OF METALS 339
the following relations
tan A =
tan B (tan2q + 1)
tan q (tan2B - 1)
tan I) =
cos2 q + tan2 p sin2 q’
In these equations tan fl is the axial ratio of the ellipse and has values
between zero and one. The angle between the major axis of the ellipse
and the normal to the plane of incidence is denoted by q, which has
values between *u/4. The convention of signs concerning q must be
selected in such a way that q > 0 for small angles of incidence (0 < 8)
and q < 0 for large angles of incidence (0 > e).
METHOD
3. SCHULZ’S
A new method of measurement, independent of the Drude method
has been described by S c h u l ~ . ~I n~ this
- ~ ~method two separate experi-
ments are performed. One of these gives accurate values of the index n,
and a relatively inaccurate value of k. The other gives accurate values
of the extinction coefficient k, when values of n are known. In the second
experiment, n must have been determined previously (e.g., from the
first experiment), except in the infrared where the value of k becomes
insensitive to n. The two experiments will be discussed briefly here;
further details and modifications can be found in the original papers.
The index of refraction n is determined from measurements of the
reflectivity of a dielectric-metal interface. The dielectric is usually taken
to be glass. In the case of a dielectric other than air that has an index nd,
Eq. (1.5) for normal incidence becomes
(n - nd)2 + k2
Ro = (n nd)2 + + k2’
Here Ro represents the ratio of reflected and incident intensities. In case
k 2 is large compared with the values of the indices, this equation may be
used to calculate n from experimentally determined values of ROand a
“reasonable” value of k. The accuracy of the value of n is fixed largely
by the accuracy of the quantity (1 - R ) and is affected only slightly by
errors in the value of k. It should be emphasized that Eq. (6.8) applies
to the case of normal incidence which is not easy to realize experimentally.
Schulz chose to make measurements of the reflectivity a t an angle of
43 L. G . Schulz and E. J. Scheibner, J . Opt. SOC.Am. 40,761 (1950).
**L.G. Schulz, J . Opt. SOC.Am, 41, 1047 (1951); 44, 357 (1954): Cu, Ag, Au, Al, Hg,
Ga.
44 L. G. Schulr and F. R. Tangherlini, J . Opt. SOC.Am. 44, 362 (1954).
340 M. PARKER GIVENS

incidence of 45”. Under these conditions, the reflectivity is dependent

upon the polarization. We shall designate by R, and R, the reflectivities
when the electric vector is perpendicular and parallel to the plane of
incidence, respectively. It may be shown from electromagnetic theory
that
R, = R,2 (6.9)
when the angle of incidence is 45”. This relation is used to check both the
accuracy of the measurements and the quality of the surfaces. Failure of
the equality by more than the experimental uncertainty implies rough-
ness of the surface or another inhomogeneity in the sample employed.
Defective samples are discarded. Failure of this equality after an exten-
sive heat treatment indicates oxidation or deterioration of the surface.
The value of n is obtained from the experimental values of R, and R,
at the 45’ angle of incidence in the following way. The average reflec-
tivity B given by
= &(R. + R,) (6.10)
is determined experimentally. Electromagnetic calculations indicate
that R is approximately equal to Ro and thus may be used in place of Ro
to calculate an approximate value of n from Eq. (6.8). The approximate
value of n and the assumed k are then used to calculate, from theoretical
curves, the difference, AR = Ro - E . This information is used to get
more nearly correct values of Ro and n. The correct value of n is obtained
by successive approximations.
The extinction coefficient k is determined by constructing a small
interference filter, such as has been used by T o l a n ~ k yand~ ~is shown in
Fig. 6. A thin sheet of dielectric material is coated on one side with silver
to a thickness which transmits only 1% to 2% of the incident light.
Mica is used because it is cleaved easily into thin sheets having parallel
faces. Half the other side of the dielectric (top half in the figure) is
covered with a similar coat of silver and the remaining half is covered
with a layer of the metal under investigation which we shall designate
as 2. If the filter is to be used in reflection, the metals on the second sur-
face may be opaque; if it is to be used in transmission, they must have a
transmission of 1% t o 2%. The dielectric is from 1p to 30p thick. This
device is essentially a Fabry-Perot interferometer and transmits a num-
ber of discrete and well-separated wavelengths. When used in reflection,
all wavelengths are reflected except for a number of narrow well-separated
bands. If white light is incident normally on this filter, the wavelengths
45 S. Tolansky, “Multiple Beam Interferometry,” Chapter 8. Oxford University
Press, Glwgow, 1948.
 OPTICAL PROPERTIES O F METALS 341

present in transmission or absent in reflection are those satisfying the

equation
2nd +-
61X
27r
+-
27r
62X
= NX (6.11)

where nd is the index of the dielectric, t is the thickness of the dielectric;

61 is the phase shift upon reflection by the first dielectricmetal (silver)

FIG.6. The Schulr method of measuring k. A. The interferometer with a mica

spacer and the reflecting layers of silver and the metal under study aa shown. B. A
channeled spectrum produced by this interferometer and a spectrograph.

interface, 6 2 is the phase shift upon reflection by the second dielectric-

metal face, N is an integer, and X is the vacuum wavelength. The wave-
lengths present in the light coming through the lower half of the filter
differ from those coming through the upper half because b2 is different
for silver and the metal under investigation.
An image of the filter may be focused on the entrance slit of a spec-
trograph, so that light going through the Ag-mica-Ag filter enters the
spectrograph through the lower half of the slit and the light passing
through the Ag-mica-s filter enters through the upper half of the slit.
Under these conditions, lines which have a break or wavelength shift
in the middle will appear in the spectrograph as in Fig. 6. Measurements
made in the upper and lower half of such a spectrum provide the wave-
lengths which satisfy Eq. (6.11) for the Ag-mica-Ag and Ag-mica-z
filters. Such wavelengths, along with Eq. (6.11) and suitable corrections
for the dispersion of mica and silver, provide numerical values of 6rs - 6,
342 M. PARKER GIVE NS

a t a number of wavelengths throughout the spectrum. Concordant values

of the optical constants of silver have been obtained by the various
experimenters, so 6r, is known. Thus 6, may be determined from this
experiment .
From these data and the index n, the extinction coefficient k is cal-
culated by the equation
(6.12)

This method has advantages which offset the rather awkward manner
in which the data are reduced. The most important of these is the internal
check upon the measurements provided by Eq. (6.9). Other advantages
are (1) that the method does not require a Babinet or Soleil compensator,
items not usually found in the laboratory; and (2) the measurements are
made on dielectric-metal interfaces which are smoother and less subject
to oxidation than air-metal interfaces. The chief disadvantages are (1)
separate samples must be prepared for the determinations of n and k ;
and ( 2 ) the method cannot be used in those portions of the spectrum in
which the dielectric is opaque.

4. OTHERMETHODS
Other methods include (1) the direct determination of k from meas-
urements of the transmission of thin films as a function of t h i c k n e s ~ , ~ ~ , ~ ~
all measurements being made a t normal incidence; and ( 2 ) the calcula-
tion of n and k from measurements of the transmission and reflection
of thin films a t various angles of incidence and for both directions of
polari~ation.~~~~*

VII. Experimental Results

1. GENERALREMARKS
The experimental results have been presented in a variety of forms.
I n his earliest experiments, D r ~ d simply
e ~ ~ reported the optical constants
of metals for visible light and made no attempt to determine the depend-
ence upon wavelength. Drude and his followers soon corrected this,ag,40~60
reporting full dispersion ,curves for the visible and near ultraviolet. In
later experimentslS1even very recent ones, the experimenters have been
46 M. Garnett, Phil. Trans. Roy. SOC.A203, 385 (1904).
47 L. G.Schuls, Phil. Mag. Suppl. 6, 102 (1957).
48 J. R. Collins and R. 0. Bock, Rev. Sci. Instr. 14, 135 (1943).

49 P. Drude, Ann. Physik 39, 481 (1890).

60 W.Meier, Ann. Physik 31, 1017 (1910):Au, Ni, Fe,Pt, Se.
61 G.Sabine, Phys. Rev. 66, 1064 (1939): Reflectivity Measurements.
 OPTICAL PROPERTIES OF METALS 343

interested in measuring either the reflecti~ity6~*6~ or the absorption

coefficient and have not determined the optical constants.
Discrepancies appeared among the results obtained by various
experimenters and it was soon determined that these were the result of
differences in the methods used to prepare the samples. The early samples
were polished mechanically with a variety of polishing agents. Later,
electropolished samples were used and still later samples mere prepared
by cathode sputtering and vacuum evaporation.
The optical properties of copper are very sensitive to the method used
in preparing the sample. This dependence has been studied extensively
by Lowery et al.64 Schulz66 has also observed the change in the opti-
cal constants of evaporated copper as a result of annealing. HassSsand
Hunter have shown that the reflectivity in the ultraviolet of evapo-
rated aluminum films is dependent upon the rate a t which the metal is
deposited upon the substrate. This lends some support to a remark
made by “If the optical properties of films are as sensitive to
residual gas in the evaporator as their electrical properties, the hope
of getting precisely reproducible results in ordinary vacuum systems
seems rather remote.”
I n 1904 Garnett46 demonstrated that there is a correlation between
the particle size and optical properties of finely dispersed metals. Since
very thin evaporated films are discontinuous, s8 forming little aggregates
of metal on the substrate, Garnett’s treatment is applicable to samples a
few atomic layers thick, which have an optical behavior quite different
from the metal in bulk. Sennett and have made an experimental
study of this phenomenon.
I n the paragraphs which follow, the experimental results will be given
for a few typical metals. Moreover, references to additional measurements
on these and other metals will be presented.

2. COPPER,SILVER,AND GOLD
The noble monovalent metals have been studied extensively, not
only because their electronic structure is simple but also because they
62 D. Fabre and J. Romand, Compt. rend. 242, 893 (1956): U.V. Reflectivity Measure-
ments.
68 S. Kandare, Compt. tend. 244, 571 (1957): U.V. Reflectivity Measurements.

64 H. Lowery, H. Wilkinson, and D. L. Smare, Phil. Mag. [7] 22, 769 (1936).
66L. G. Schulz, J . Opt. Soc. Am. 44, 540 (1954).
66 G. Hass and W. R. Hunter, J . Opt. Soc. Am. 46, 1013 (1956).
67 G. D. Scott, J . Opt. Soc. Am. 46, 178 (1955).
68 0. S. Heavens, “Optical Properties of Thin Solid Films,” Chapter 3. Butterworth,
London, 1955.
R. 5. Sennett and G. D. Scott, J . Opt. Soc. Am. 40, 203 (1950): Very thin films of
Ag, Au, Cu, Al, Sb. Cr,and Pd.
344 M. PARKER GIVENS

FIG.7. Dispersion curve for the conductivity of silver.

FIQ.8. Dispersion curve of (1 - E) for silver. Notice the change in the vertical
-
scale at 1 E = 10.

are easy to handle and evaporate. The data shown in Figs. 7 through 13
were taken from the work of Beattie and Conn,60S c h u l . ~ ,Bor
~ ~ el
. ~aZ.,61
~
Forsterling and Freedericksz,* Meier,60 and Hodgson.62 Other experi-
80 J. R. Beattie and G. K. T. Conn, Phil. Mug. [7] 46, 989, 222 (1955): Cu, Ag, Au,
Ni, Al in the infrared.
6 1 J. Bor, A. Hobson, and C. Wood, Proc. Phys. Soc. (London) 61, 932 (1939).

62 J. N. Hodgson, Proc. Phys. Soc. (London) B68, 593 (1955): Cu, Ag, Au, Zn, Sn, Al

in the infrared.
 OPTICAL PROPERTIES O F METALS 345

mental data are also a ~ a i l a b l e . ~The

- ~ ~data
- ~ ~in the graphs are generally
presented in such a way as to show the conductivity (nkv) and the
quantity (1 - e) where e is the dielectric constant (nz- k 2 ) . These
quantities are of interest in the theory of solids. For comparison, part
of the data for silver are also presented in such a way that n and k are

OPT I C A L CONSTANTS
20 -2.0 OF SILVER

I
-SCHULZ
15-1.5 -MINOR
-A-* F~STERLINGan
FREEDERICKSZ
t t
X I = n

A (PI--
FIG.9. The optical constants of silver.

given as functions of the wavelength. The solid lines in the curves describe
the results of Drude’s theory. The parameters were adjusted to provide a
good fit, and were not taken from the measured dc value of the conduc-
tivity. The sudden change in the conductivity of copper and gold between
500 m p and 600 mp and of silver between 300 m p and 400 mp is generally
attributed to the low-frequency limit of the internal photoelectric process.
This process appears as an increase in the conductivity of the metal since
it absorbs energy from the light wave. I n copper, a second rise begins
P. L. Clegg, Proc. Phys. Soc. (London) B68, 774 (1952) : Ag, Au, Sn, Ir.
64 R. Kretzman, Ann. Physik 37, 303 (1940). (This contains a list of many older
references.)
66 A. Q. Tool, Phys. Rev. 31, 1 (1910).
O6 F. GOOEJ,2.Physik 106, 606 (1937): Ag, Au.
346 M. PARKER GIVENS

h Ip)-

FIQ.10. Dispersion curve for the conductivity of gold.

FIG.11. Dispersion curve of (1 - C) for gold.

was made to explain the absorption anomalies in silver and gold, the
principal quantum numbers being increased by 1 and 2, respectively.
This is reasonable since the three metals have very similar electronic and
6’ N. F. Mott and H. Jones, “The Theory of the Properties of Metals and A~oYs,”
p. 118. Oxford University Press, London and New York, 1936.
 OPTICAL PROPERTIES O F METALS 347
crystalline structures. It is not clear why silver and gold exhibit only
one absorption edge, whereas copper has two. The colors of copper and
gold are a consequence of the presence of these absorption bands in the
visible.
OR

CONDUCTIVITY OF COPPER

FdRSTERLlNG and
FREEDERIKSZ.
lor fi 0 3 8 8 I I I I I 1 1 1

Fro. 13. Dispersion curve of (1 - e) for copper.

The results obtained in the near infrared by Schulz,66when corrected

for the anomalous skin effect, agree with the Drude theory [Eqs. (3.7)
through (3.10)]if one assumes there is a single free electron per atom and
takes the effective mass m* of the free electron to be 0.98 m, 0.97m, and
1.45 m for gold, silver, and copper respectively. These values agree very
348 M. PARKER GIVENS

well with the values of m*/m obtained from mea~urements6~~6~ of the

electronic specific heat of the metals. This improved agreement is a
consequence of the care with which the samples were prepared and the
fact that the measurements were made on a surface unexposed to the air.
Schulz measured the extinction coefficient k by both reflection and trans-
mission. The surface (or reflection) value was found to be smaller than
the volume (or transmission) values by 5 % or 10%.
3. THE ALKALIMETALS
Lithium, sodium, and potassium follow the free electron theory of
Drude and Zener very closely. No evidence has been found for internal
photoelectric absorption. These metals become transparent in the ultra-
violet as has been mentioned previously (see Table I). The optical con-
stants of sodium and potassium have been measured by Ives and Briggs?"
who found that they agree with the values predicted by the free electron
theory for sodium and potassium if m* = 1.42m. Their results for sodium
are reproduced in Fig. 2.
Cesium and rubidium have also been examined by Ives and Briggs?'
who found some absorption, presumably as a result of internal photo-
electric absorption which begins in the quartz ultraviolet (A > 2000 A).
Recent work by Frohlich and P e l ~ e indicates
r~~ that plasma oscillations
may play an important role in the optical properties of these metals.
4. DIVALENT
METALS
A number of divalent metals have been examined; many show
anomalous absorption in the red and near infrared. The absorption bands
are a result of internal photoelectric absorption which probably takes
place between the valence band and the band just above it. Since there
are two electrons per atom in these metals, the valence band is nearly
filled. Hence the minimum energy required for photoelectric absorption
is less than for the monovalent metals if the band structure has the form
represented in Fig. 4. The experimental curves for zinc and beryllium
are shown. See Figs. 14 through 17. Zinc is a typical divalent metal;
beryllium has the simplest electronic structure of any of the divalent
metals. The data for beryllium do not extend very far into the infrared.
Apparently the anomalous absorption band has its low-frequency limit

68 J. Rayne, Phys. Rev. 96, 1428 (1954).

69 W. S. Corak, M. P. Garfunkel, C. B. Satterthwaite, and Aaron Wexler, Phys. Rev.
98, 1699 (1955).
70 H. E. Ives and H. H. Briggs, J . Opt. SOC.Am. 27, 181 (1937); 26, 238 (1936).
71 H. E. Ives and H. H. Briggs, J . Opt. Soe. Am. 27, 395 (1937).
74 H. Frohlich and H. Pelzer, Proc. Phys. SOC.(London) A68, 525 (1955).
 OPTICAL PROPERTIES OF METALS 349
near 2 p, rises to its maximum a t 0.3 p, and begins to fall again. The value
of 1 - e has a maximum near 0.25 p , the fact that this is a t a wavelength
shorter than the peak of the conductivity curve indicates that no absorp-
tion band of higher frequency is contributing to the absorption. This

h C)-

FIQ.14. Dispersion curve for the conductivity of zinc.

Fra. 15. Dispersion curve of (1 - E) for zinc.

absorption is probably a result of transitions between the valence (2s)

band and the 2 p band.
The interpretation of the curves for zinc is similar.
The data given here are from the work of Meier,60 Hodgson,a2 Bor
350 M. PARKER GIVENS

e l UZ.,~' Bock,la and Givens.14Other measurements on the divalent metals

are also a~ailable.'.7~
5. ALUMINUM
Trivalent aluminum has been studied extensively because it handles
well and has practical value in the production of mirrors and interference
filters. See Figs. 18 and 19.

FIQ.16. Dispersion curve for the conductivity of beryllium.

FIG.17. Dispersion curve of (1 - e) for beryllium.

The observed values follow the form of the Drude theory if one uses
a value for the dc conductivity which is 0.23 times its actual value and
assumes about 1.3 valence electrons per atom. As mentioned earlier, the
optical properties in the ultraviolet are sensitive to the rate of evapora-
'8 R. 0. Bock, Phys. Rev. 68, 210 (1945):Be, Mg,Zn.
74 M. P. Givens, Phys. Rev. 61,626 (1942).
76 H.M.O'Bryan, J . Opt. Soe. Am. 26, 122 (1936):data on Be, Mg, Ce, St, Ge, La,
Sr, Mn.
 OPTICAL PROPERTIES OF METALS 351
tion. The data reproduced here are from Hodgson,62Beattie and Conn160
and Schulz.4a

Fro. 18. Dispersion curve for the conductivity of aluminum.

QI 10
. 10 100
A(P)-=

FIQ.19. Dispersion curve of (1 - e) for aluminum.

6. THE LIQUIDMETALS
Mercury and gallium have been studied recently in the liquid state
by S c h ~ l z 'and
~ earlier by MeierSO and others. The measured optical
constants are found to agree with the Drude theory within experimental
error.
7. THEANOMALOUS SKINEFFECT
The version of the theory of the anomalous skin effect which assumes
diffuse reflection of electrons from the surface has been checked experi-
'OL. G. Schulz, J . Opt. SOC.Am. 47, 64 (1957).
352 M. PARKER GIVENS

mentally and found satisfactory. The experiments which test this theory
must be made a t low frequencies or low temperatures or both. I n addition
to the near infrared work of Weiss and Ramantham which was cited in
part V, there is the work of Biondi,’? who made more refined measure-
ments of the absorptivity of copper and silver in the region from 1.5
to 3.3 p and a t liquid helium temperature (4.2’K). His results are shown
in Table IV. He used etched single crystals.

TABLEIV. COMPARISON
OF THE EXPERIMENTAL RESULTSOF BIONDI’WITH
THE THEORETICAL
PREDICTIONS OF HOLSTEIN* FOR THE ABSORPTIVITY
AT 1.5 TO 3.3 p A N D AT T = 4.2”K

Theoretical absorptivity
Experimental
Metals Surface Volume Total absorptivity

Copper 0.0029 0.0020 0.0049, 0.0050

Silver 0.0036 0.0009 0.0045 0.0044

0 M. Biondi, Phys. Rev. 102, 964 (1956).

T. Holstein, Phys. Rev. 96, 535 (1954).

The column headed “surface” under the designation “theory,” is

derived from the theory of the anomalous skin effect; the column headed
“volume” is derived from a quantum-mechanical treatment by Hol-
stein.18 He assumed that an electron near the surface absorbs a photon
and diffuses into the volume of the lattice where it generates a phonon
instead of giving up its energy to the lattice by collision with the sur-
face. This effect is not important a t room temperatures, but is obviously
important here.
The anomalous skin effect has also been confirmed in the microwave
region by Chambers78 who measured the absorptivity of a number of
metals a t 1200 and 3600 mc/sec, and a t temperatures ranging from 2 to
90°K. He found good agreement between theory and experiment for
copper and small discrepancies for silver and gold. Theoretical predictions
are not available for the other metals he examined.
77 M. Biondi, Phys. Reu. 102, 964 (1956).
7s R. G. Chambers, Proc. Roy. SOC.(London) A216, 481 (1952): Cu, Ag, Au, Sn, Cd,
Pb, Al.