300 PHILIPS TECHNICAL REVIEW VOLUME28

The skin effect

Il. The skin effect at high frequencies
H. B. G. Casimir and J. Ubbink

. In Part! of this article [1] the skin effect was consid- The solution is again (cf. 1,13):
ered for various configurations of current-carrying con-
ductors without laying any restrictions on the frequen-
EIEo = JIJo = HIHo = exp [-(1 + j)zl<5k], • (3)
cy. It was taken for granted that audio or radio fre-
where the subscript 0 refers to the value at the surface,
quencies were involved. However, as the frequency or
and where 15k again represents a length, namely the clas-
the conductivity of the metal increases, complications
sical skin depth as defined in (1,1~):
arise in the theory of the skin effect, and these will be the
subject of this second part. <5k2 = 2/WI£(J. (4)
To study these complications it is useful to take a
In our treatment of the low frequency case in Part I,
simple situation, such as emerges from the following
had we wished to, we could have pictured the "current-
considerations. As the frequency increases, the electro-
carrying substance" as a continuous inertialess fluid.
magnetic fields and currents in the metal become con-
The complications arising at high frequencies are a
centrated in an increasingly thinner layer under the
direct consequence of the inadequacy of this picture:
surface, while outside the metal the finite value of the
the inertia of the electrons leads to relaxation effects,
velocity of light becomes apparent: the wave character
while the particle character of the electrons manifests
of the electromagnetic field emerges more clearly and
itself in the anomalous skin effect: the skin depth
the wavelength decreases. Now if the skin depth in the
becomes smaller than the mean free path of the elec-
material and the wavelength outside are small com-
trons.
pared with the smallest radius of curvature of the sur-
The nature of the complications becomes evident if
face, a typical "optical" situation arises, which is
the conduction process is imagined as follows. The
basically that of an electromagnetic wave incident on a
electrons move about with randomly oriented velocities
surface. It may be remarked that microwaves represent
of magnitude v. The current is carried by a relatively
an intermediate case: in microwave devices the wave-
small directed component. The resistance arises because
length is of the same order of magnitude as the device
the electron motion is interrupted by collisions; the
itself, whereas the skin depth is relatively very small (as
it is even at much lower frequencies). The simplest
situation is that of a plane electromagnetic wave inci-
dent normally on the plane surface of a metal. For a
linearly-polarized plane wave in free space, E, H and the
direction of propagation are mutually perpendicular.
In the following we shall use a co-ordinate system as
shown in fig. 1, with the z-axis normal to the surface
(metal for z > 0), E and J along the x-axis, and H
f ~E,J

along the y-axis. E, J and H then depend only on z

(and on t, as exp jwt).
The classical skin effect in this situation may be
described very simply. Writing down Maxwell's
equations (1,1) and (1,2), with J = (JE, and omitting _Z
the displacement current, we find the following rela-
tions between the fields in the metal:
-oH/az = (JE, (1)
oE/az = -jwf.1B. (2)
Fig. 1. The co-ordinate system. The (x,y)-plane is the boundary
plane between free space (on the left, z < 0) and metal (on the
Prof Dr. H. B. G. Casimir is a member of the Board of, M~,wg_e- right, z > 0). An electromagnetic wave is incident on the surface
ment of N. V. Philips' Gloeilampenfabrieken; Dr. J. Ubbink IS with from the left. E and J are directed along the x-axis, H along the
Philips Research Laboratories, Eindhoven. j-axis.
1967, No. 10 SKIN EFFECT, II 301

time between two collisions is characterized by a relax- alternating field between two collisions. The collisions
ation time 7:, and the mean free path of the electrons are now of little significanee to the alternating field,
is 1 = V7:. Let us now compare three lengths: the mean which "sees" a layer of virtually free electrons; these
free path I, the skin depth 6 and the distance uk» which electrons almost fully reflect the incident wave. This is
an electron travels in 1/2n of the period of the alternat- the reflection which is characteristic of many metals in
ing field. (It may be noted that comparing 1 with »ko is the optical region. The skin depth here is independent
equivalent to comparing W7: with 1.) both of wand of 7:. We can describe B as a case of
The various limiting cases that may occur are shown extreme relaxation, since the current, which in A is in
in fig. 2. At low frequencies and not too high a con- phase with the field, lags more and more behind the
ductivity we have 'case A: 1 « 6 and 1 «vlw, which field as the frequency increases, owing to the inertia of

-
0--.... E-- ...

A-
- B-- ....

-
Fig. 2. Diagram of electron paths in a metal in and near the skin layer, in four extreme cases.
The dashed line marks the skin layer, whose thickness may of course assume widely dif-
ferent values. The zig-zag line represents a colliding electron, the wavy lines represent the
oscillatory motion of the electron due to the alternating field, and the arrows indicate the
incident electromagnetic waves.
A) 1« 0 and 1« vlw (classical skin effect)
B) vlw« I « 0 (relaxation)
D) 0« I and 0« vlw (anomalous skin effect)
E) »k» e; 0 « I (anomalous reflection)
I is the mean free path, 0 the skin depth and ulco the distance covered by an electron in
(1/h) x one period of the alternating field.

corresponds to the classical skin effect. The electron the electrons, till in B it lags almost 90° behind the field.
experiences many collisions during the time it spends In D we have the anomalous skin effect: 1has become
in the skin layer and within one period of the alternat- greater than 6. The incident radiation is more strongly
ing field. Ifthe frequency is increased 6 and vk» become absorbed than would be calculated from the theory of
smaller, and if the conductivity is increased, I becomes the classical skin effect; this effect is to some extent
larger and 6 smaller. In both cases the region of the related to the higher resistivity in films which are
classical skin effect is ultimately left behind; we then
come either to case B, where oiox «; I «6, or to [1) H. B. G. Casimir and J. Ubbink, The skin effect, I. Intro-
case D, where 6 « 1 and 6 « uko. In B the electrons duction; the current distribution for various configurations,
'still undergo many collisions during their time in the , Philips tech. Rev. 28, 271-283, 1967 (No. 9). This article is
referred to in the following as I. The equations in I are
skin layer, but there are now many periods of the referred to as (1,1), (1,2), etc.
302 PHILIPS TECHNICAL REVIEW VOLUME28

thinner than the mean free path. In the extreme case the surface impedance Z is a suitable quantity.
events in the skin layer, and hence both the absorption The definitions of these quantities and the relation-
_and the reflection, become independent of I and hence ship between them follow most conveniently by con-
independent of the low-frequency conductivity of the sidering a plane wave incident normally upon a plane
metal. If we increase the frequency in case D, or the - surface and assuming that the electromagnetic fields
mean free path in case B, we arrive at situation E, and currents in the material itself are exponential
where ok» « (j « I. This case is closely related to B: functions of place and time.
here again the field "sees" a layer of virtually free In optics the plane wave in the direction +z is
electrons, and the reflection is almost 100 %. E is described by
referred to as the case of anomalous reflection; the resi- E = Eo exp (-2nkz/À.) exp jw(t - nz/c) =
dual absorption is a consequence ofthe collisions at the
= Eo exp jw(t - Nz/c),_ . . . . . (5)
surface and not, as in B, of the normal collisions
within the material. where
In these four cases the displacement current can be
N= n-jk. .... _ (6)
disregarded. If, starting from B or E, we increase the
frequency still further, the displacement current in the .The velocity of electromagnetic waves in the material
Maxwell equation (1,1) ceases to be negligible com- is cln, and the amplitude decreases over a distance À.
pared to J. At the "plasma frequency" D and Jare by a factor exp 2nk. In (5) c = l!V(eo,uo) is the
equal in magnitude. At still higher frequencies the velocity and À.the wavelength in free space; c = wJ,.j2n.
power not reflected at the surface is not dissipated in The complex skin depth
a thin layer but is propagated forwards as travelling
waves just as light waves in free space: transmission
+
(j = (j' j(j" . . . . . . . . . . (7)

takes place. A skin effect no longer exists. is now defined such that the behaviour of the fields in
In the following we shall deal in more detail with the the material is given by:
effects outlined above. At this point it is useful to refer
E = Eo exp j(wt - z/(j), . . . . (8)
to figs. 11 and 12 which show the combinations of w
and .. at which the limiting cases mentioned above and comparing (5) and (8) we see that:
occur. In these figures C is the transmission region. At
lj(j = »N}c. . . . . (9)
the end of this article we shall discuss, with reference to
fig. 12, which cases fall into the category of communi- The concept of surface impedance is arrived at in the
cations engineering and which into optics. following way. It is "easy" for currents to exist in
Before examining the effects referred to, we shall metallic material but "difficult" for electric fields. The
attend to two preliminaries. In the first place, in the electric fields at the surface of a metal are weak but the
optical treatment and also in the microwave range, cer- magnetic fields may be strong. The magnetic fields
tain concepts are useful for expressing the properties of induce currents in the metal, and these (because the
the metal surface in relation to incident electromagnetic conductivity is not infinite) give rise to weak electric
waves. These concepts, such as the complex refractive fields and hence absorption. Assuming that the currents
index and the complex skin depth, will be defined in and fields vanish as z goes to infinity, then for the
some detail and we shall deal briefly with their inter- induced current:
relationship. Secondly, in order to introduce some
fJdz . Ho, (10)
terms occuring in the electron theory of conduction,
we shall present a very concise summary of this theory, where Ho is the magnetic field at the surface. This is
confining ourselves to the free electron model. found by applying Stokes's theorem to (1,1) along a
contour that first runs parallel to H at the surface and
Skin depth, refractive index and surface impedance then penetrates deep into the metal to where the fields
There are various concepts which serve to character- are zero. The displacement current is taken to be
ize a metal surface in relation to incident electro- negligible. The surface impedance Z = R + jX is now
magnetic waves. In optics one uses the optical constants by definition:
nand k, to describe the refraction and the absorption
Z = Eof fJdz = Eo/Ho, (11)
respectively of light in the material. These can be com-
bined to form a complex refractive index N. At lower where Eo is the electric field component along the sur-
frequencies, particularly in the microwave region and face and perpendicular to Ho. If the displacement cur-
in discussions ofthe anomalous skin effect, use is made rent is not negligible, the' second expression of (11) is
ofthe complex skin depth (j. At still lower frequencies the general definition of the surface impedance.
1967, No.JO SKIN EFFECT, 1I 303

From (2), (8) and (11) we have at once: In metals, for a very wide range of values of wand ..,
IZI «Zo, or cko :» 15lftr; in other words, provided
Z = wft5. (12)
that ftr does not differ too much from unity, the free-
Since only the fields at the surface appear in the space wavelength is much greater than the modulus of
definition (11) of the surface impedance, we can use the complex skin depth, and
this definition also for situations where the fields in the
a F::j 4RIZo = 4(wlc)ftr5'. (14)
bulk of the material are not exponential. Formally we
are at liberty to retain for such situations the definitions How far this range extends will be seen later. At optical
(9) and (12) for the complex refractive index and the frequencies we can put ftr = 1.
complex skin depth .. If we compare (3) with (8) we see that in the case of
. Quantities directly measurable are the absorptivity a the classical skin effect (A in fig. 2):
and the reflectivity e; e is the reflected fraction of the
incident power and a the absorbed fraction. We assume
(1 + j)/5 k = j/5,
hence
that the material extends to infinity in the direction of
5 = t(1 + j)5k, 5' = 5" = tr3k, (15)
the -l-z-axis, so that there are no complications due to
reflection from the other side. We then have a + (! = 1. where 5k is the real quantity defined by (4).
The term "transmission", already used in the intro-
The free-electron model
duetion for the effect found above the plasma frequen-
cy, means then that the radiation penetrates a consider- . As a background to the matter to be treated, we shall
able distance, i.e. many wavelengths, into the material. now present a concise summary of the electron theory
At such frequencies a finite thickness of the material of metallic conduction, in its simplest form - the free-
transmits a finite fraction of the radiation. The absorp- electron model.
tivity a should not be confused with the absorption As early as the beginning of this century Drude [2]
coefficient, i.e. the fraction of the power dissipated per put forward a theory which took account of the relax-
metre in a travelling wave; this is equal to Zosk]«. ation effects. Drude's model runs as follows. If Ve is
The absorptivity a and the reflectivity (! may be the mean velocity of the electrons, then:
expressed in terms of 5, Z or N, as follows. Outside the
mVe = eE - mud»: (16)
metal, i.e. for z < 0, the electromagnetic field is a
superposition of the incident plane wave and a m is the mass and e the charge of an electron. The term
reflected wave: on the left-hand side is the inertia term, the first term
on the right the external force, and the second term
E = ZoP exp jw(t - zie) + ZoQ exp jw(t + zie), on the right is a frictional term characterized by a
H = P exp jw(t - zie) - Q exp jw(t + zie), relaxation time x, The relaxation time is the character-
where Pand Q are complex coefficients and Zo = istic time needed to reach a static equilibrium, that is
= V(fto/eo) = 377 n, the impedance of free space. At to say the time the current takes to reach a steady state
the surface z = ° the fields are: after a change in the external field. In the simplest form
of electron theory .. is of the order of the average time
Eo = Zo(P + Q) exp jwt,
between two collisions of an electron with the lattice.
Ho = (P - Q) exp jwt.
If ne is the concentration of the electrons, then the
E and H are continuous at the surface. It follows, using current density J = neeve, from which, using (16), it
(11), that: follows that:

j = (nee2Im)E - JI..· (17)

In the d.c. case, i.e. when the field is constant (w = 0),

the left-hand side of the equation is zero, and we have:
Q Z-Zo
and hence J = (JE,
P . Z+Zo'
so that with (18)

2 For sinusoidal fields and currents j = jwJ and we find

= 1 Q \2 = 1 Z - Zo 1 = (R - Zo)2 + X2 from (17) and (18):
epI Z + Zo (R + ZO)2 + X2.
(JE
J=--- (19)
4RZo . 1 jw .. +
and a= . (13)
(R + ZO)2 .+ X2 [2) P. Drude, Ann. Physik 14, 677 and 936, 1904.
 304 PHILIPS TECHNIC~L REVIEW VOLUME28

. By means of this simple model Drude established (A in fig. 2) is given by (cf. eq. 15):
a relationship between the electrical properties of
metals at low frequencies and their optical properties.
l52= j/wp,(J = jAp2/W't'. (27)
This will be discussed in the next section. . The following considerations apply to metallic con-
The result of a quantum-mechanical treatment of the ductors. If we exclude semimetals and semiconductors
free-electron problem is that the Drude model remains we can treat the parameters ne, V, Wp and Ap (and also
formally applicable with the restrietion that we must m in so far as the effective mass differs from the electron
assign to the electrons a velocity V, the Fermi velocity, mass) as "insensitive" material parameters. They are
to be distinguished from the drift velocity Ve discussed virtually constants for a given metal, and are about the
above. The Fermi velocity is a randomly oriented same from one metal to another. In the free-electron
velocity of substantially the same value for all con- theory the values of V, Wpand Apare determined entirely
duction electrons; for an ordinary metal, it is of the by ne. Against these are the "sensitive" parameters:
order of 106 mis, whereas the drift velocity in an (J, I, r. These may vary by many factors of 10, depen-
ordinary metal at a current density of, say, 1 'A/mm2 ding on temperature and purity. They are proportional
is of the order of 10-4 mis: The fact that the electrons to one another, the constant of proportionality being
move in a lattice of positive ions can be taken into determined by the insensitive parameters.
account by assigning to them an effective mass m To give some idea ofwhat may be regarded as typical
instead of the free electron mass. The Fermi velocity values for the quantities we have introduced, Table I
is given by ne and m: gives values for the insensitive parameters of a "stan-
dard metal", calculated from a reasonable choice of ne,
(20)
m, er and p,r. For the same metal r and I have been
The electrons are subject to two kinds of scattering: calculated for three representative values of (J, corres-
./
by' phonons (the quanta of thermallattice vibrations), ponding to very pure copper at low temperature, copper
giving a relaxation time i dependent on the temper- at room temperature, and constantan.
ature; and by lattice imperfections, which means that
i also depends on the purity ofthe metal. The constancy
Non-instantaneous and non-local relations between J
of the Fermi velocity implies that the relaxation time i andE
is directly related to a mean free path I, We now have the groundwork for making a closer
study of the effects mentioned in the introduction. Of
1= Vi. (21)
these, the relaxation effects and the transmission can be
In Drude's theory ne, m and e only occur combined understood on the basis of Drude's theory, as we shall
in the form nee2/m. We now introduce the plasma
(angular) frequency Wp and a corresponding length Ap
as follows: Table I. Selected parameters defining a "standard metal" and
other parameters calculated from them. The sensitive parameters
I and • have been calculated for three representative values of
Wp2 =
nee2/em, (22) a. The following constants were used in the calculations:
Ap2 = m/p,nee2, (23) e = 1.60 X 10-19 C, n = 1.05 X 10-34 Is, Eo = 8.85 X 10-12 F/m,
Ilo = 1.26 X 10-6 H/m.
WpAp = l!V(ep,). (24)

2nAp is the wavelength corresponding to Wpfor electro- lie = 6.0 X 1028 m-3
Selected III = 9.1 X 10-31 kg
magnetic wàves propagating in a dielectric material parameters Er I
characterized by e and p, (the "medium", see I, Intro- {lr = I
duction). The quantities Wp and Ap will be used for'
the present only as a measure of the electron concen- v = 1.40 X 106 m/s
tration. Later they will acquire a more direct physical Calculated Wp = 1.38 X 1016 S-l
insensitive ).p = 2.18 X 10-8 m
significanee as a characteristic frequency and a char- parameters ./a = 6.0 X 1022 Dms
acteristic length. I/a = 8.3 X 10-16 Dm2

The metal is now characterized by Wp or Ap(electron

concentration) and 't' (purity and temperature). We a I'>l 1012 (Dm)-l
(pure copper, 4 OK)
·I I'>l
I'>l
5 X 10-10 S
10-3 m
then have, from (18) and (22), (23), and (4), for I
example: Sensitive
parameters
a = 6x 107 (Dm)-l
(copper, 300 OK)
·I= = 4 X 10-14 S
(J :_ eWp2't' = 't'/p,Ap2, (25) I 5x 10-8 m

l5k2 = 2Ap2/W't', (26) a = 2 X 106 (Dm)-l ·I== 2x

I X 10-15 S
(constantan) 10- m 9
and the complex skin depth in the classical region I
 1967, No. 10 SKIN EFFECT, II . ,' 305

see in the next section. On the other hand, the pheno- parameter along with a,' the conductivity at zero fre-
mena associated with the anomalous skin effect and quency; • and a are related by the expression (18) or
anomalous reflection were first investigated only com- (25). We shall still keep to the simple configuration
paratively recently (London 1940 [3l, Pippard 1947 [4l). given in fig. 1. Two modifications now arise in eq. (1).
Between the relaxation effects and the anomalous In the first place we take the displacement current into
skin effect there exists, in a certain sense, a kind of account; in the second place we do not substitute aE
symmetry. This is to be seen as follows. In describing for J but aE/(I + jcor) as given by (19). If the fields
the skin effect it' is necessary to solve Maxwell's equa- and currents have an exponential time dependence as
tions for given boundary conditions and for this it is given in (8), Màxwell's equations may then be written:
essential to know the relationship between J and E in'
aE I
the material. In the theory of the classical skin effect we jH/~ =jwBE+ _'-.-, (28)
simply write J = aE in the metal. This is a local, 1 + JW1:
instantaneous relationship: the field at a given place -jE/~ = -jwpB. (~9)
and at a given instant determines the current density
at the same place and at the same instant. In Drude's Multiplying these equations we obtain:
theory the current at a given place at the time to is jw,ua
determined by the values of the field at the same place, I/~2 = Bf-lW2 - ',. (30)
but at times of the order of. or less previous to to;
1 +jW1:

in other words, there is a local, non-instantaneous Together with (9), this expression establishes a
relation. In the theory of the anomalous skin' effect we connection between the complex refractive index
encounter a non-local relation: since the fiéld varies N = n - jk on the one hand and a and • on the
considerably over distances of the order of a mean free other. For our further discussion we write (30) with
path, the velocity of an electron at a given place is the aid of (22) to (25) in the following form: '
determined by fields in an environment of the order of
1
magnitude of l. The symmetry referred to between the (J..p/~)2 = (w/Wp)2 - 1 +. (31)
relaxation effects and the anomalous skin effëct is thus 1 + JW.

the symmetry between "non-instantaneous" and "non- giving ~ as a function of wand 1:. The co-r-plane (fig. 3),
local". in which W is plotted horizontally and. vertically, both
The relaxation effects in Drude's theory are so much on a logarithmic scale, can now be divided into three
simpler than the non-local theory because it is possible
and meaningful to restrict the treatment to a single
frequency, whereas in the non-local theory a restrietion
to a single wavelength in the material is meaningless;
moreover, the presence ofthe surface is a complicating
factor in the non-local theory. Because we may restrict T
ourselves to a single frequency in Drude's theory the
time factor may be eliminated and we thus COmeto the f
simple local relationship between J and E given by (19).
The non-instantaneous origin of (19) is still reflected
in the imaginary term in the denominator.
It should be noted, incidentally, that the explicit A
occurrence of the mean free path I in the anomalous
(non-local) theory - in addition to the parameter 1:
already encountered in Drude's theory - means that
the Fermi velocity v, a typical quantum-mechanical c ..
quantity, will now be of significance.

Relaxation effects and transmission

_w
We shall now consider the modifications to be made Fig. 3. The co-r-diagram, in which. is plotted vertically and w
horizontally, both on a logarithmic scale. With Drude's theory
to the theory of the skin effect when Drude's theory is three regions A, Band C can be distinguished in which, well
applied. We now have the, relaxation time. as a new away from the boundaries, the behaviour of the electromagnetic
waves is very simple. A is the region of the classical skin effect
(see fig, 2), B the relaxation region (see fig. 2) and C the trans-
mission region. In A, ó is given approximately by (32), in B by
[3] H. London, Proc. Ray. Soc. A 176, 522, 1940. (34) and (35), and in C by (36). The equations for the boundaries
[4] A. B. Pippard, Proc. Ray. Soc. A 191, 385 and .399, 1947. are indicated in the figure.
306 PHILlPS TECHNICAL REVIEW VOLUME 28

regions. In each of these regions (well away from the I/w-r in (31). In that case we find, for ro« Wp:
boundaries) the behaviour of fJ is particularly simple.
fJ" = Ap, (34)
A) w-r is so small that the second and third terms
fJ' = Ap/2w-r. (35)
on the right-hand side of (31) can be combined to -jw-r,
and the first term is much smaller than this new term if C) Only the first right-hand term in (31) is signifi-
(W/Wp)2 « rol' «1, i.e. w-r« 1 and t» «Wp2 r. We cant. This is the case if ro» Wp and also if w » wp2-r.
then have approximately (Ap/fJ)2 = -jw-r, or: Then approximately l/fJ2 = sp,w2, and hence

(32) fJ' = I/roV(sp,), fJ" = O. (36)

This is the expression for the classical skin effect (27). The solutions (8) now represent undamped travelling
The region A is therefore the region of validity for the waves of phase velocity 1!V(sp,) and wavelength 2:n;fJ'.
classical skin effect. From (9), (6) and (25) it follows The fields are therefore in principle no longer limited to
that (27) is equivalent to: a thin layer at the surface, and transmission occurs. We
are then really outside the region with which we are
n =k = V(a/2sw). (33)
concerned: a skin effect no longer occurs. The para-
This is the Hagen and Rubens relation [51, the classical meters that define the behaviour of the electrons have
skin effect expressed in optical terms. vanished. The electrons have no longer any effect on
B) The middle right-hand term of (31) is the most the electromagnetic field (again, of course, only in the
significant when rol'» 1 and t» «wp. Then approxi- zero order approximation).
mately fJ2 = _Ap2, or fJ' = 0 and fJ" = .Ap.The electro- In a previous section we gave an approximate ex-
magnetic field is now an oscillation without wave pression (equation 14) for the relation between the
character, which decays along the z-axis as exp (-z/.Ap) complex skin depth and the absorptivity; we can now
(sometimes called an "evanescent wave"). The pene- state the region in which this expression is valid. The
tration depth is independent of frequency and of d.c. condition for the validity of (14) was cko ;» IfJl, or
conductivity (i.e. of relaxation time). Between two w2/C2 « 1/lfJ12 (putting p,r = 1). This means that the
collisions the electrons undergo many cycles of the first term on the right-hand side of (30), and thus the
alternating field. (1'» 1/ w) so that what the electro- first term on the right-hand side of (31), must be small
magnetic field primarily "sees" is a layer of entirely compared with the other terms; in other words, coming
free electrons which screen but do not absorb because from A or B we must not approach the boundary with
the mean velocity of the electrons, and hence the cur- C too closely.
rent, is always 90° out of phase with the field. Given The relation between these limiting cases and the
w-r» 1 this 90° phase lag follows immediately from transition from the one wave behaviour to the other is
(19). B is the relaxation region. The incident wave is found by constructing the curve representing fJ in the
fully reflected. The absorption, however, is zero, and. complex fJ-plane, as ro goes from zero to infinity (fig. 4);
the reflection complete, only to the zero order approx- and further by considering the character of the wave
imation. The first order approximation is found by field corresponding to various points in the complex
also taking into account terms of the first order in fJ-plane (fig. 5). Fig. 4a shows the locus of fJ in the

Q
Fig. 4. The path followed by <5 in the complex plane, according to Drude's theory, as W
goes from 0 -+ 00, a) in the three limiting cases A, Band C of fig. 3, b) for wp.» 1, c) for
Wp. = 2, and cl) for Wp. « 1. The arrow indicates the direction of increasing co,
 1967, No. 10 SKIN EFFECT, II 307

limits A, Band C as to varies (increasing in the direction and time dependence of the field as given in (8):
of the arrow). If we examine qualitatively how b varies
in accordance with the complete expression (31), we exp j(wt - z/b) = exp (-b~'z/lbI2) ~xp j(wt =: b'z/lbI2).
arrive at fig. 4b for wp?:» 1, fig. 4c for Wp?: = 2 and We shall confine ourselves to waves travelling to the
fig. 4d for Wp?: « 1. For wp?:» 1, b describes a loop right (or to the evanescent- case), b' ;;:::0, i.e. to the
first and fourth quadrants of the complex b-plane. In
the first quadrant we have attenuated waves (bil> 0),
in the fourth quadrant waves of increasing amplitude
(bil < 0), alongthe real axis unattenuated waves and
along the imaginary axis the evanescent case (non-
propagating oscillations) either decaying exponentially
(è3" > 0) or augmenting exponentially (bil < 0) with z.
In these figures there are two further points to note.
a) In fig. 4 we see that b is always in the first quadrant;
we therefore always have attenuated travelling waves,
and energy is always dissipated in the material. In the
limiting cases Band C the dissipation is zero: in C the
waves are unattenuated, and in B no waves are propag-
ated and there is therefore no energy transport. b) In
the "plasma transition" B -- C at large values of Wp?:
(fig. 4b) b first goes off along the imaginary axis
(b' R:j0, bil increases) and later returns along the real
axis (bil ~ 0, b' decreases). At first, therefore, the field
penetrates further and further, while the fields at
different points have virtually the same phase; later
phase differences occur over large distances, i.e. we
have long travelling waves, which become shorter and
shorter and are virtually not attenuated.
To what extent does Drude's theory cover the exper-
imental facts? In general terms we may say the follow-
ing.
For our standard metal the "plasma boundary"
to = Wp is at a-wavelength of2nÀp = 2~x2.18 X 1O-8m
= 0.137 (.I.m,i.e. in the far ultra-violet (see fig. 6).
The value of r at the point where the three' regions meet
is I/wp = 0.725 X 10-16 s; this is a factor of 500
lower than the ?: for copper at room temperature r
?:Ir = 3.6 X 10-14 s. The wavelength at which copper,

at room temperature, passes the boundary on: = 1 is

Fig. 5. The wave character of the field in the metal appropriate 2nc?:k = 68 (.I.m.
to various regions of the complex ö-plane. For waves travelling
to the right (i.e. along the positive z-axis) 15 lies in the first or the
Confining ourselves at first to "optical" frequencies
fourth quadrant (15' > 0). We have attenuated waves in the first (the visible range and the near infra-red and ultra-
quadrant, waves of increasing amplitude in the fourth quadrant,
unattenuated waves on the ö'-axis (ö = 0), and on the öH-axis
H violet) we find from these rough estimates that only
(15' = 0) non-propagating oscillations having no wave character metals which are poor conductors will be in region A;
("evanescent waves"), either decaying (ö > 0) or augmenting
H

(ö < 0) along the z-axis.

H for metals which are good conductors optical frequen-
cies lie in region B.
o Constantan is an example of a poor conductor, where
,between Band C; this loop disappears for small wp?:. the Hagen-Rubens relation, which is appropriate to
The intermediate case with a cusp occurs for Wp?: = 2 region A, is still reasonably valid. For the good con-
(this cusp has no obvious physical significance). For 0
ductors (Cu, Ag, Au, AI) the expressions (34) and (35)
very large Wp?: the loop is traversed in a very short for region B give a good qualitative description of the
co-Interval. . experimental behaviour of the optical constants as
Fig. 5 shows the wave fields corresponding to certain
points in the b-plane. These fields follow from the space [SI E. Hagen and, H. Rubens, Ann. Physik 11, 873, 1903.
308 PHILIPS TECHNICAL REVIEW VOLUME28

expected plasma frequency. At these frequencies ab-

sorption bands are found which. are attributed to
quantum jumps of electrons induced by the radiation.
-1& These are not accounted for in Drude's theory.
r=3.6xl0 s
In metals such as gold, copper and silver the plasma
bouridary is thus obscured by these absorption bands.
This does not apply to the alkali metals. In 1933
Wood [8] discovered that these become transparent in
the ultra-violet. Zener [9] interpreted the wavelength
at which this occurs as the plasma boundary. The
theoretical and experimental transition wavelengths are
compared in Table lIl.

Table m. The plasma' boundary, i.e. the transition wavelength

at which a metal ceases to reflect and becomes transparent:
experimental values, after Wood [BI (Äexp) and theoretical values
(2:n:Äp)after Zener [91. nsuln« is the number of electrons per atom
I I
À=68p.m À=137nm at which the theoretical plasma wavelength would coincide with
the experimental value.
Fig. 6. Some characteristic values of. and Ä in the co-r-diagrarn
(see fig. 3) for the standard metal. At the point where the three Äexp 2:n:Äp neiln«
regions meet (00 = Wp, 't' = I/wp), • = 0.7 X 10-16 s, and
Ä = 137 nm. At • = 3.6 X 10-14 s (corresponding to the value Li 205 nm 155 nm 0.54
of a for copper at 300 OK) the boundary line ar: = 1 corres- Na 210 210 1.00
ponds to a wavelength Ä = 68 (.I.m. K 315 290 0.85
Rb 360 320 0.79
Cs 440 360 0.67

functions of w, provided w is not too large [6]. To

obtain numerical agreement however it is necessary to
insert a value for 7: which is appreciably smaller than In the low-frequency and radio-frequency regions the
the value corresponding to 0', the d.c. conductivity (see classical skin effect theory provides in general a good
Table 11). The results of measurements in the optical description of the phenomena. In his measurements of
region would thus be in agreement with the theory if the high-frequency resistance of superconducting tin in
these metals were worse conductors than they really 1940 H. London [3] found, however, as an incidental
are. A possible explanation for this is that the concen- result, that at low temperature the apparent resistance
tration of impurities in the thin layer under the surface, ofthe normally conducting metal was greater at micro-
in which the optical effects take place, is greater than wave frequencies than the d.c. resistance. He then put
in the metal as a whole. This also explains the dispari- forward a suggestion that this was due to the fact that
ties in the results of various workers and explains why the skin depth becomes smaller than the mean free path
the results depend to such a marked extent on the ofthe electrons, and in so doing laid the foundation for
method of preparing the surface [6]. the theory of the anomalous skin effect.
The values of w at which there ceases to be qualita-
The anomalous skin effect
tive agreement are in general much lower than the
In calculating the classical skin effect we implicitly
assumed that we could define the current density J and
Table n. Comparison of optical and low-frequency electrical
properties of copper, silver and gold, after Försterling and the field strength E at any point in the metal, and that
Freedericksz [71•• (opt) is the relaxation time necessary to describe these quantities were always connected by the relation-
the results of the optical measurements by Drude's theory;
.(a) is the relaxation time derived from a. We have converted
ship J = O'E.Now if the skin depth - a distance over
the a-values given by the authors to. by means of eq. (25). The which the fields vary considerably - becomes smaller
factor lJ.Îtp2 which occurs in this conversion was derived for the
free-electron model from the density and the atomic weight, than the mean free path (see D in fig. 2) then the velocity
assuming one electron per atom .• is expressed in units of 10-14 s of the electron (and hence its contribution to the cur-
and l1.Îtp2 in units of 10-22 Hm.
rent density) is determined not only by the local field
Cu Ag Au but also by all the fields to which the electron has been
I subjected since its last collision.
.(opt) 0.46 0.95 1.7
A simple picture of the situation - which can also be
.(a)
I 2.5 3.9 2.6
used to obtain a reasonable estimate of the effect - has
p,Äp2 4.2 6.1 6.1
I been given by Pippard [4] with his "ineffectiveness con-
 ..
1967, No. 10 SKIN EFFECT, II 309

cept". An electron moving approximately at right -reil is equal to the time -r between two collisions. In the
angles to the surface will be subject to the field of the case we are considering, however, the period of un-
skin layer only for a small part of its transit time be- interrupted interaction is marked by the electron enter-
tween two collisions. lts interaction with the field is ing or leaving the skin layer and a collision with the
therefore very slight and it is therefore not effective in surface: -reil is now approximately equal to the time
the screening and absorption of electromagnetic waves. needed to pass through the skin layer. If we put
Screening and absorption, and therefore the formation -reil = bê]», and, by analogy with (26), 152= 2Äp2jw-rell,
of a skin layer, are mainly due to the few electrons we arrive immediately at (38). The reason for the fact
which graze the surface at small angles, since these that 15is independent of the sensitive parameters might
spend a large part of their free transit time in the skin therefore be summarized as follows: the thickness of
layer. The fraction ofthe electrons in the skin layer that the skin layer is determined by the effective interaction
are effective depends on the skin depth whose value, between the electrons and the field, and this interaction,
however, calculated in the classical way with a reduced provided I» 15,is in turn determined by the thickness
number of effective electrons, is in turn dependent on of the skin layer; the sensitive parameters have thus
this fraction. If 15is the skin depth, then the ratio ofthe been completely eliminated.
number of effective electrons (the electrons whose path A rigorous theory of the anomalous skin effect has
between two collisions is almost entirely inside the skin been given by Reuter and Sondheimer [lOl, and a
layer) to the total number in the skin layer is roughly somewhat less rigorous theory by Chambers and
bjl, where I is the mean free path ofthe electrons. In the Pippard [11l. Where a comparison is possible, both
skin layer there is therefore an effective conductivity treatments give the same result. In the theory three
aell = babj!, where b is a number approximately equal steps can be distinguished.
to 1, or somewhat greater because the non-grazing 1) The main problem is to find the current density
electrons do still make a slight contribution. With this occurring at a point in the material as a result of the
effective conductivity we again calculate the skin depth, field, not only at the point itselfbut also within a radius
using equation (4), and find: of the order of magnitude of I. Chambers and Pippard
postulate that an electron "remembers" the fields,that
152= 2jwftaell = (2jwfta)ljM = bk2ljbb,
it has traversed in the past, but that its memory fades
hence
in the course of time as exp (-tj-r). A calculation along
(37)
these lines leads to the following non-local relation
This expression is instructive inasmuch as it shows that,
apart from a factor Ijbl/3, the skin depth is the geo-
metric mean of the classical skin depth 15k(doubly
weighted) and the mean free path I. But at the same
J=-
3a
4nl
exp (-rjl)
1'4
r:
between J and E (see fig. 7):

dV. (39)

time it is perhaps misleading, as it suggests that 15 The current density J is a volume integral; r is the
depends on the sensitive parameter I. That this is not radius vector between a volume element d V where a
the case follows if we substitute for bk2 the expression field E prevails and the point where J is calculated.
(26); with I= v-r we obtain:
(38)
We can also arrive at this result by a line of thought
which we shall use again at a later stage. This is based
on the introduetion of an "effective relaxation time"
-reil. This is the time during which the field and the
electron are in uninterrupted interaction with one an-
other. It determines the effect of the electron on the . Fig. 7. Referring to eq. (39) giving the current density at P.
field (screening, absorption). In the classical case .the A contribution dJ comes from electrons from the element d V
at Q that reach P without colliding, and which have been subject
field remains the same between two collisions, so that to the field E in Q; r is the radius vector from Q to P.

[6] M. Parker Givens, Solid State Physics 6, 313, 1958.

[7] K. Försterling and V. Freedericksz, Ann. Physik 40, 200, The result (39) can be obtained as follows, The field in the
1913. direct environment of Q (jig. 8a) affects the current in P because
[8] R. W. Wood, Phys, Rev. 44, 353, 1933. P is traversed by a number of electrons which have come from Q
[0] C. Zener, Nature 132, 968, 1933. without colliding on the way. We note in the first place that a
[10] G. E. H. Reuter and E. H. Sondheimer, Proc. Roy. Soc.
A 195, 336, 1948. field in Q perpendicular to r has no effect on the current density
[11] A. B. Pippard, Advances in Electronics and Electron Physics in P: electron paths leading from Q to the environment of Pare
6, 1, 1954; Reports on Progress in Physics 23, 194,218, 1960.. twisted slightly by such a field; in the presence of such a field
310 PHILIPS TECHNICAL REVIEW VOLUME28

'the electrons that pass through P are not the same as those when completely wiped out upon reflection. The integral in
there is no field, but their number and mean velocity is identical (39) then has to be taken only over the volume of the
in both cases (fig. 8b, c). Let us now consider all the paths in a
metal. For specular reflection the calculation proceeds
solid angle d.Q of the electrons that pass P from the Q-direction
in a given short time interval; their number is proportional to
as if the electrons were able to pass through the surface
d.Q (fig. 8d). If we follow these paths backwards in time, that is undisturbed, and as if outside the metal the electric
to say, in the direction of Q, then every now and then one will, field were a mirror image of that inside the metal; the
as it were, be "knocked out of the cone". The chance of this integral is then taken over the entire volume. In its
happening in, an element dr is equal for each path and propor- effect on a reflected electron prior to the reflection, the
tional to dr. The number of "undisturbed" paths thus decreases
exponentially in the direction of Q and the number of relevant
field in the metal is simulated by the image field.
3) Substitution ofthe integral expression found for J
J
in Maxwell's equations (omitting the displacement cur-
rent) yields, after elimination of H, an equation for E '
p
r Q which has been solved by the above-mentioned
authors [10] [11], for the cases where all the electrons
undergo either specular or diffuse reflection at the sur-
E face.
Q
The result (38), based on simple considerations,
~P
agrees surprisingly well with the rigorous theory.
According to the latter, equation (38) - after the
correct value for b has been inserted - is' a valid
expression for the complex skin depth -in the anomalous
limit without relaxation (b « I, b « v]», D in fig. 2),
except for one detail. The considerations that lead to
(38) are too oversimplified for a proper analysis of the
complex character of b. It is found, in fact, that the
rigorous theory simply adds a minus sign to equation
(38). In other words, in the anomalous limit (without
relaxation) the complex skin depth is given by:
Fig. 8. Calculation of the effect of the field at Q on the current (j3 = -(2/b)(v/W)Ap2. (40)
in P via electrons originating directly from Q.
(jmust lie in the first quadrant of the complex b-plane
(passive material, waves travelling to the right). The
paths in the element dr around Q is proportional to exp(-rll)dr.
If we now assume that the average contribution to the current
minus sign then implies that [12]
intensity in P due to the field in Q is proportional to the para11el
b" = ö'V3. (41)
component of the field in Q (in doing so allowance is made for
the collisions), then the contribution d.lto the current density The value of b depends on the manner in which the
in P is proportional to (rlr)Er exp(-rll)dr d.Q, where Er is the electrons are reflected at the surface. For diffuse re-
field component in Q in the direction ofr. The factor r]» indicates
that dJ has the direction of r, since EJ. has no effect. Substitution
flection b = 4n/V3 ~ 7.3; for specular reflection b =
of Er = (E.r)lr and dr d.Q = d Vlr» gives the exprèssion (39), 352-7nV3 ~ 10.3. Here, therefore, the character of the
apart from the proportionality factor. The integration is easily reflection has only a minor effect. As we shall see later,
carried out for a. homogeneous field E; if in this' case we put this is no longer the case if »[o: « b (E in fig. 2).
J = aE, we find immediately the value of the factor of propor-
The rigorous theory, of course, provides more in-
tionality.
In (39) the relaxation effects are neglected. These can be taken
formation. In the first place the complex skin depth is
into account by substituting for E the "retarded field" [E], i.e. also calculated in the transition region where I is not
the field in d V. at the instant to - rl» at which the electrons much larger or smaller than bk. Secondly, the field con-
from d V contributing to J at to passed through d V. figuration in the metal is calculated completely. The
fields are no longer exponential functions of place. For
a given relative decrease at the surface the fields extend
.2). Next, the effect of the surface has to be taken into much deeper into the metal than would be the case with
account. It is necessary to make an assumption about an exponential function. (Even so, the complex skin
the manner in which the electron is reflected at the sur- depth is still a useful concept for describing the surface
face; the cases which have been calculated completely in relation to incident waves; see the remark following
are those of specular reflection and diffuse reflection. equation 12.)
In the latter case the "memory" of the electrons is In order to find the boundary in the zo-r-plane (cf.
1967, No. 10 SKIN EFFECT, II 311

fig. 3) between the regions of validity of the classical

skin effect (A) and of the anomalous skin effect (D) 1/R
we put 1= 1c51. Calculating c5 in the classical limit as
given by (27), we then find: Î
(42) o Ag
n Au
This new boundary is shown in fig. 9 together with the " Pb
boundaries indicated in fig. 3.
For our standard metal (Table I) at the new boun-
dary we have (lrr3 = 2Ax 10-28 S2, which means that
for the values 7: = 5 X 10-10 sand 7: = 4 X 10-14 s from
Fig. 10. Surface conductivity llR of silver, gold and lead as a
function of Va at 3600 MHz, after Chambers. [13). The points
represent measured values obtained at various temperatures. The
drawn curve is calculated from the theory of the anomalous skin
effect with diffuse reflection of the electrons at the surface. For
each material the scale is chosen so as to give the best fit between
experimental values and theoretical curve. (On the basis of specu-
lar reflection the best fit that can be obtained is not nearly so
good.) This matching process gives all, a quantity which is
independent of the sensitive parameters (in our notation all =
f IIp,vÄp2). The initial portion corresponds to the classical limit,
where the relationship is a linear one; the right-hand portion
corresponds to the extreme anomalous limit, where R is indepen-
dent ofa.

figure shows the surface conductivity (ljR ex: Ijc5') for

a given frequency, plotted against the root of the con-
ductivity a (on arbitrary scales). According to the
classical theory these quantities are proportional to one
c another: IjR = V(2ajf.tw). The most striking feature
of c5' in the anomalous limit is its independenee of the
sensitive parameters, e.g. a (see equations (40) and (41)).
_W
This is neatly demonstrated in fig. 10.
Fig. 9. The co-r-diagram as in fig. 3, showing the new boundary
between the classical skin-effect region (A) and the region of the
extreme anomalous skin effect (D). In the latter region the skin
depth is given approximately by eq. (40). According to (40) and (41) in the anomalous limit and at a
given frequency v' is entirely determined by lie (through v and
Äp). Rigorous application of the free-electron theory gives in this
Table I the boundary is reached at the values t» ~ 2 S-l limit:
and W 4x 1012S-l respectively. We must therefore
f::::j
lIe~ = 3n2(liI4bp,e2w15'3)3. ....• (43)
expect that the "best" metal, e.g. pure copper at 4 "K,
Chambers used this to find the number of free electrons per
will have already become anomalous at very low fre-
atom for various metals; see Table IV [14).
quencies and that at millimetre waves (w 1012S-l)
f::::j
The fact that values near unity are found for the simple metals
even a very "moderate" metal, like copper at 300 "K, may be seen not only as a confirmation in general terms of the
will show anomalous behaviour. At the point where the theory outlined above, but also as an indication that the surfaces
new boundary runs into the old one, W7: = 1, we have used by Chambers were treated with great care. If the surface is
not smooth and free from stresses the values found for. v' can
7: = Apjv which, for our standard metal, is 1.6 X 10-14 s.
easily be on the large side, resulting (as lie OC 15'-9/2) in values
A metal which is a poor conductor, such as constantan for lie which are far too Iow. Moreover, the free-electron theory
(i 10-15 s) will therefore not enter the anomalous
f::::j can be expected to be a reasonable approximation only for the
region D as the frequency is increased, but the relax- simplest metals. For these reasons these results should be seen
ation region B. rather as a confirmation of the theory than as an accurate deter-
minatien of lic.
As an elegant confirmation of the theory of the
anomalous skin effect we reproduce in fig. 10 some Table IV.
experimental results obtained by Chambers [13]. This
Cu 1.0
Ag 0.68
[12) This relation can also be derived directly from the Kramers- Au 0.60
Kronig relations, if it is assumed as given that V oc W-1/3 Sn 1.10
(see Pippard [l1). Hg 0.23
[13) R. G. Chambers, Proc. Roy. Soc. A 215, 481, 1952. AI 0.37
(14) E. H. Sondheimer, Phi!. Mag. Supp!. I, I, 1952.
'J

312 PHILlPS TECHNICAL REVIEW VOLUME28

Anomalous reflection approximately where the mean free path is equal to the
The question that immediately arises on considering penetration depth: I = Ap, or:
fig. 9 is whether relaxation effects arise in metals that
•= Ap/v. (45)
are in the anomalous region if the frequency is in-
creased still further. Conversely, we may ask whether In region E - in contrast to D - it makes a great
anomalies comparable with the anom.alous skin effect difference whether the reflection is specular (and there-
occur if, at a given frequency, the mean free path fore imperceptible to the alternating field since the
increases (e.g. because the temperature is reduced) in a tangential velocity is unchanged) or diffuse (and there-
metal in the relaxation region B. And in either case, fore equivalent to a collision). To make this plausible,
what may we expect to happen to the penetration depth let us consider how energy is transferred from the
and the surface impedance? alternating field to an electron. Suppose that an electron
Let us first increase the frequency, starting from the is subject, between collisions, to a homogeneous
anomalous region D. The characteristic difference be- alternating field. For cases A and B this is a reasonable
tween regions A and B was connected with the number approximation u « b). The velocity of the electron
of periods of the alternating field which the electron immediately after a collision is completely arbitrary
experiences between two collisions: this was much and on average makes no contribution to the transfer
smaller than 1 in A (r « l/w) and much greater than of energy from the field to the electron. The additional
1 in B (.» l/w). In the extreme anomalous region, velocity imparted to the electron as a result of the
however, the relaxation time 7: is no longer significant
and therefore no effect can be expected when the boun-
dary • = l/w is passed. Instead, what now matters is
whether the electron undergoes many periods of the
alternating field or only a fraction of a period during 0 -R
:,,; E
It 3'"-
the time it spends in the skin layer. This time is about 3'"- It
3
Ibl/v, and the new boundary is therefore Ibl/v = 'ï]». ï 3-
Obtaining Ibl from (40) and disregarding the factor 2/b __ :.:!:pi':___
this is equivalent to: 1 8
w/wp = o]c. (44)
Coming from D and passing this boundary we enter
the region E (fig. 11) which we shall call the region of
anomalous reflection (see also E in fig. 2). Here, as in B,
the field sees a layer of virtually free electrons, and the
penetration depth, as in B, is equal to Ap. The time spent c
in the skin layer is now approximately Ap/v; putting this
equal to 1/w we again find the boundary (44) between D
andE. _W

What difference now remains between regions E Fig. 11. The co-r-dlagram as in fig. 9, with the new region E, the
and B? To the zero order of approximation there is no region of "anomalous reflection". Both in E and B eq. (34) is
valid to the zero order approximation. To a first approximation,
difference: in both cases the layer of free electrons has the absorption is given in B by eqs. (35) and (14) and in E by (46).
a screening effect,(penetration depth Ap) but it does not
absorb, since the velocity of an entirely free electron in
an alternating field of constant amplitude is always 90° acceleration by the alternating field always results
out of phase with the field. The difference is of the first initially in a positive transfer of energy from the field
order. In B some absorption takes place because the to the electron. If w. « 1, then at a given field ampli-
free motion of the electrons 'is interrupted occasionally tude this energy transfer between two collisions in-
by a collision. In E the reflection of the electron at the creases as • increases. On the other hand, if w.» 1,
surface must be held responsible for the absorption (at the total energy transfer between two collisions will be
least if the reflection is not specular); a small contribu- independent of' r, because, after the initial transfer of
tion (becoming smaller with increasing frequency) to energy during about one period, energy will be trans-
the absorption in E arises because the electron is ferred alternately from the electron to the field and
moving through the skin layer and thus sees an alternat- vice versa, with zero net transfer of energy. The energy
ing field of varying amplitude (Holstein (151). transfer per second therefore decreases as. increases,
. The boundary between Band E (see fig. 11) lies Going now from A to D, we may make use of an
 1967, No. 10 SKIN EFFECT, 11" 313

"effective relaxation time", for in case D the electrons Table V.

enter the skin layer suddenly. They begin by taking up
energy as they enter (disregarding their random velocity
• in 10-14 s Cu Ag
I Au

upon entry), and long before the field has completed .(opt) 0.46 0.95 1.7
one period the energy exchange is interrupted, either at .(a) 2.5 3.9 2.9
'E 3.1 4.2 4.2
the surface (diffuse reflection) or upon emerging from
the skin layer (specular reflection). The ratio of the
absorptions for diffuse and specular reflection would be
expected to be about 2 at the most. To verify the theory in the E-region, and in partienlar
In going from B to E the situation is different. In equation (46), it is necessary to carry out absorption
case E the electron enters the skin layer very "gradual- measurements at high values of -c, that is to say on pure
iy" with respect to the alternating field: the electron metals at low temperatures, paying particular attention
experiences many periods in the time that it sees the to keeping the surface clean. A detailed comparison of
field grow from zero to its value at the surface. There is published _experimental material with the Reuter and
no distinct initial energy transfer, and positive and Sondheimer theory has been given by Dingle [18]. By
negative contributions continue approximately to com- way ofillustration we shalljust mention Ramanathan's
pensate one another. If at the surface the reflection is results [19] and those found by Biondi [20].
specular the situation remains unchanged, and the total For copper at 1.4 (l-m and 4.2 "K Ramanathan
absorption is much smaller than in B. If however the measured an absorptivity of 0.006. For our standard
reflection is diffuse, then immediately after a reflection, metal, equation (46) gives an absorptivity of 0.0035.
just as after a collision in B, there is a distinct transfer The agreement is very reasonable, particularly com-
of energy from field to electron. pared with the value of a R! 0.00003 predicted for this
We can now make an estimate [16] ofthe absorptivity case by the classical theory, from (35).
for region E with diffuse reflection by using the absorp- Biondi has measured the absorptivity of copper and
tivity for region B but with an effective x, this being the silver, in the wavelength region 0.3 to 3.3 (l-mand at
time spent in the skin layer per collision: -Celt = 2Àp/v. about 4.2 "K, with -c values (derived from the d.c.
Using (35) and (14) we find that the absorptivity in B resistance) of about 10-11 s. In addition to a few
is given by a = 2/wp-C. It then follows for region E that expected absorption bands (cf. page 308) at the shorter
a = 2/wp-Celt = v]c. This differs only by a factor of 4/3 wavelengths, he found for À >.1.5 (l-mthat, in quali-
from the absorptivity found by Holstein [15] in an tative agreement with (46), the absorptivity was inde-
exact calculation for diffuse reflection: pendent of the wavelength (and also independent of
temperature, although the temperature interval covered
a = tv/c. (46)
was small, being 3.4-4.2 OK). Numerically, Biondi's
Holstein also made the calculation for specular re- conclusion is that, even with a virtually ideal surface
flection, but since all experimental results indicate that and far from the absorption bands, equation (46)
the reflection is entirely diffuse, we shall not consider accounts to a considerable extent but notfully for the
this here. Holstein's results are implicit as a limiting absorption measured in the extreme E-region. From a
case in the Reuter and Sondheimer theory [17]. further theoretical study ofthe processes possible in the
E-region, a "volume-absorption process" in the skin
For our standard metal the boundary (45) lies at 1.6 X 10-14 s,
layer: has been put forward (as opposed to the surface
so that the measurements by Försterling and Freedericksz (see process of (46» and this is held to be responsible for the
Table 11) lie more or less in the boundary region. In Table Y remaining absorption.
we have added the values for rn = 8Àp/3v to the data ofTable II: Table VI presents the theoretical and experimental
'E is the effective r which, with (35), gives the absorption in E.
values given by Biondi.
It is clear that the theory for the E-region does not explain these
'E
optical results: in the first place in the E-region ought to be
lower than the actual r ; moreover 'E is even further away from
• than .(opt). For Iow values of rfopt) therefore, we still have to Table VI. The absorption factor of copper and silver, after
resort to an explanation like that mentioned on page 308. Biondi [20].

Cu Ag
[15] T. Holstein, Phys. Rev. 88, 1427, 1952.
[16] See also Holstein's estimate, given by Biondi [20].
Surface-effect (eq. 46) 0.0029 0.0036
[17] R. B. Dingie, Physica 19, 311, 1953. Volume-effect 20 09
[18] R. B. Dingie, Physica 19, 348, 1953.
un] K. Ramanathan, Proc. Phys. Soc. A 65, 532, 1952.
Total ----:w ~
Experimental 50 44
(20) M. A. Biondi, Phys. Rev. 102, 964, 1956.
314 PHILIPS TECHNICAL REVIEW VOLUME 28

Briefly, the volume-absorption process may be described as for a number of monovalent metals. These have been
follows. In pure metals the absorption correlated with the calculated from the expressions from the free electron
"normal" relaxation time T (in A and B) is due to electrons theory, in which ne is determined from the atomic
colliding with phonons (the quanta of the thermal lattice vibra-
weight and the density with the assumption of one con-
tions). At very low temperatures no phonons are present (T -+ 00).
However, an absorption process of the following kind is still duction electron per atom.
possible: the electron absorbs a photon (quantum ofthe alternat- It can be seen from this figure that the limiting cases
ing field) and simultaneously generates a phonon as a result of of the free electron model that interest the communi-
its interaction with the crystal lattice. cations engineer differ from those that interest the opti-
cal physicist. In communications we are not as yet con-
Concluding remarks cerned with waves shorter than 1 mm, and are therefore
Fig. 12 shows once more the w-i-diagram, now with interested only in the normal skin effect (A) and the
a scale in seconds along the r-axis and a scale in anomalous skin effect (D). In optical physics, in the
radians/second along the w-axis. A scale for the wave- visible spectrum (hatched area on A-scale), or in the
length }, is also added. The boundaries between the near infra-red, we are interested only in the relaxation
regions A, B, C, D and E are indicated in the diagram region, with normal reflection B or anomalous reflec-

10-8Sr- ~--~bTTc----------_T~IT_--_,

I
I
I
I
I
I
o I E
anomalous skin effect I anomalous
I reflection
I '//////////f/

I 5 I
I I
I I

---- '<-,
....___....___
I
II
3$ ~
.___.___~-
I
-----1
-Cu 300 oK
2 c
///////////////// I trans-
--.::~t======~==umission
A
normal skin effect B. I
relaxation I
I
I

cm lcm

Fig. 12. The w-T-diagram for various monovalent metals. The electron concentration ne has
been obtained from density and atomic weight on the assurnption that there is one electron
per atom; the Fermi velocity v and plasma frequency Wp have been derived from this using
the expressions for the free-electron model. a) Caesium, b) the standard metal, c) copper.
The values of ne for the other alkali metals and for silver and gold lie between those for
caesium and copper, and thus correspond to intermediate diagrams. A classical skin effect
region, B relaxation region with normal reflection, C transmission region, 0 anomalous skin
effect region, E relaxation region with anomalous reflection.
The finely-shaded strips indicate the w-T-region where measurements have been carried out
by: J Hagen-Rubens (51, 2 Försterling-Freedericksz [71, 3 Wood [~J, 4 Chambers [[31,
5 Biondi [201. (The r-value of 3 is uncertain; Wood gave no values for conductivity.)
A wavelength scale is also given along the horizontal axis; the hatched area corresponds
to the visible spectrum.
1967, No. 10 SKIN EFFECT, J[ 315

tion E. In optics we are concerned with the A-region tion mechanism we have described, are swamped by the
only for the infra-red (see for example the hatched absorption bands, which are of different origin. This
area 1, corresponding to the experiments of Hagen and happens to an increasing extent at higher frequencies,
Rubens mentioned on page 307). The region D is right where an increasing number of processes can be ex-
outside the domain of optics. E differs from B in that cited by the radiation. Consequently, the transmission
the absorption is considerably higher than would be region C, which on the basis of the free-electron model
calculated with the expression appropriate to B, the for metals would be expected in the ultra-violet or
reason being that the collisions at the surface begin to X-ray regions, is in practice of no importance for met-
become relatively more significant than the collisions als. Nevertheless, the fairly sharp transition from B or
with the lattice. E to C, the plasma boundary, has been convincingly
The difference between D and A is of a similar nature demonstrated by Wood's experiments [8].
to that between E and B: the absorption in D is greater The simplified theory outlined in the foregoing is of'.
than would be derived from the expressions for A. We course generally too approximate for quantitative and
shall illustrate this by taking a microwave cavity reso- detailed study of the phenomena. Let us quote two of
nator whose Q (quality factor) is to be improved by the shortcomings that very soon appear upon closer
using a materialof higher conductivity. Let the cavity examination. In the first place it is too naïve to assume
be cylindrical, with equal height and diameter (2a) and that only one relaxation time is sufficient for the whole
let it be excited in the TEoll mode (the wavemeter range of frequencies. Secondly, in nearly all metals the
mode). In this case a = 0.66 A (where A is the wave- velocity of the electron (the Fermi velocity) depends
length in free space) and Q = a/20'. For A = 3 cm upon the direction in which it moves in the crystal
and copper at room temperature (a = 6 X 107 Q-1m-1 structure.
and. = 4 X 10-14 s) the classical skin effect theory is In spite of these and other shortcomings, we hope
still valid (see fig. 12), so that according to eq. (15) that the present article gives some idea of the relation-
0' = t Ok. Using Ok2 = 2Ap2/0J?: (eq. 26) we find ships that exist between a variety of phenomena.
0' = 0.3 [Lmand Q = 30 000 (for Ap, and for v below,
we take the value for the standard metal, given in
Table I). Now for a metal where. is 104 times greater
(e.g. very pure copper at 4 OK), we would expect, on
Summary. The skin effect at high frequencies is examined for the
the basis of the classical expressions, 0' to be 100 times case of an electromagnetic wave at- normal incidence upon the
smaller, and Q therefore 100 times greater. Fig. 12 plane surface of a metal. In this case simple relations can be
given between the concepts "skin depth", "refractive index" and
shows, however, that the anomalous limit should be a "surface impedance". The complications which arise at high
better approximation than the classical limit for this frequencies are discussed with the aid of the free-electron model
for a metal. In Drude's classical model it is found that at in-
value of x, Using 0' = tlol and 1013 = (2/b)(v/W)Ap2 creasing frequencies the frequency range of the classical skin
- see eqs. (40) and (41) - we find 0' Rj 0.07 [Lm, effect is followed first by a "relaxation range" (with total reflect-
ion); at still higher frequencies the metal becomes transparent.
which is smaller by a factor 4. Thus, the Q of the In pure metals, taking account of the mean free path of the
cavity is only 4 times, rather than 100 times, greater electrons, and in particular of the fact that this can be greater
than the skin depth, one finds an "anomalous skin effect" at
than it was for copper at room temperature. frequencies between the classical range and the relaxation range.
In optical physics it is usually the case that the free Moreover, in the relaxation range the absorption becomes
anomalous. In both cases the absorption is greater than would
electron effects in Band E, and in particular the absorp- be calculated from the classical theory.