RF cavities

(Lecture 25)

February 2, 2016

319/441
Lecture outline
A good conductor has a property to guide and trap
electromagnetic field in a confined region. In this lecture we will
consider an example of a radio frequency (RF) cavity, and discuss
some of its properties from the point of view of acceleration of
charged particles. We then discuss the electromagnetic pressure
and derive Slater’s formula.

320/441
Waveguide, TM modes
Let us consider a cylindrical waveguide of radius a made from a
perfect conductor. Such a waveguide has a number of
electromagnetic modes that can propagate in it. We will focus first
our attention here on so called TM modes that have a nonzero
longitudinal component of the electric field Ez , with Bz = 0. To
find the distribution of the electric field in the waveguide for a
mode that has frequency ω, we will assume that in cylindrical
coordinates r , φ, z,

Ez (r , φ, z, t) = E(r )e −iωt−imφ+iκz , (25.1)

use Eq. (13.4) for Ez

m2 ω2
 
1 d dE 2
r − 2E+ − κ E = 0. (25.2)
r dr dr r c2

321/441
Waveguide, TM modes
The solution of this equation is given by
E = E0 Jm (k⊥ r ) , (25.3)
−1
√
where Jm is the Bessel function of m-th order and k⊥ = c ω2 − c 2 κ 2 .
The boundary condition Ez = 0 at r = a requires that k⊥ r be equal to a
zero of Jm . For each function Jm there is an infinite sequence of such
zeros, which we denote by jm,n with n = 1, 2, . . ..

1.0

0.8

0.6
J0 ,J1 ,J2

0.4

0.2

0.0

-0.2

-0.4
0 2 4 6 8 10 12 14
x
322/441
Waveguide, TM modes
Hence k⊥ = jm,n /a and recalling the definition of k⊥ we find that
1/2
ω2 jm,n
2

κm,n = ± − . (25.4)
c2 a2

In order for a mode with indices m and n to have a real value of κ,

its frequency should be larger than the cut-off frequency cjm,n /a.
The plus sign defines the modes propagating in the positive
direction, and the minus sign corresponds to the modes in the
opposite direction. If ω < cjm,n /a, then we deal with evanescent
modes that exponentially decay along the z-axis (and,
correspondingly exponentially grow in the opposite direction).

323/441
Waveguide, TM modes
Given Ez (r , φ, z, t) as defined by (25.1) we can find all other
components of the electric and magnetic fields using Maxwell’s
equations. They will all have the same dependence e −iωt−imφ+iκz
versus time, angle and z. The radial distribution of the four
unknown components Eφ , Er , Bφ and Br (remember that Bz = 0)
are found from the four algebraic equations, which are r and φ
components of the two vectorial equations ∇ × E = iωB and
c 2 ∇ × B = −iωE .

324/441
Waveguide, TM modes
Here is the result
iκm,n a 0  r  −iωt−imφ+iκm,n z
Er = E0 Jm jm,n e (25.5)
jm,n a
mκm,n a2  r  −iωt−imφ+iκm,n z
Eφ = −E0 2
J j
m m,n e (25.6)
rjm,n a
mωa2  r  −iωt−imφ+iκm,n z
Br = E0 2 2 Jm jm,n e (25.7)
c rjm,n a
iωa 0  r  −iωt−imφ+iκm,n z
Bφ = E0 2 Jm jm,n e , (25.8)
c jm,n a

where Jm0 is the derivative of the Bessel function of order m and we

dropped the indices m, n on the left sides. These modes are
designated TMmn or Emn . Example: TM01 .

325/441
Waveguide, TM modes
Note that in addition to vanishing Ez on the wall, which we have
satisfied by choosing k⊥ = jm,n /a, we should also require Eφ = 0
on the surface of the wall (because it is tangential there). This
however is automatically satisfied because the radial dependence of
Eφ in (25.6) is the same as Ez in (25.3).

The physical meaning has the real parts of Eqs. (25.5). Since the
longitudinal wavenumbers (25.4) do not depend on m, the modes
with positive and negative values of m (assuming m > 0) are
degenerate—they have the same values of κm,n . A sum and
difference of m and −m modes, which convert e imφ and e −imφ into
cos mφ and sin mφ, are often used as another choice for the set of
fundamental eigenmodes in circular waveguide.

326/441
Waveguide, TE modes
Phase velocity of this modes is larger than the speed of light:
ω ω
vph = =c >c (25.9)
κ 2 2 2 /a2 )1/2
(ω − c jm,n

This means that we cannot use EM field in waveguides for

acceleration of charges. Disk loaded structures are used for that
purpose.

327/441
Waveguide, TE modes
TE modes have nonzero longitudinal magnetic field Bz with Ez = 0.
Their derivation follows closely that of TM modes. A simple observation
of special symmetry of Maxwell’s equations allows one to obtain the
fields in TE modes without any calculation.
Indeed, assuming the time dependence ∝ e −iωt for all fields, Maxwell’s
equations in free space are

∇ × E = iωB , c 2 ∇ × B = −iωE , ∇ · E = 0, ∇ · B = 0.
(25.10)

Note that a transformation

(E , B ) → (c B , −E /c) (25.11)

converts (25.10) into itself. This means that having found a solution of
Maxwell’s equation one can be obtain another solution by means of a
simple transformation (25.11). The only problem with this approach is
that one has to make sure that the boundary conditions are also satisfied.

328/441
RF modes in cylindrical resonator
Cylindrical resonator is a cylindrical pipe with the ends closed by
metallic walls. Various modes of electromagnetic field that can
exist in such a resonator are characterized by their frequency. The
resonator modes can be easily obtained from the waveguide modes
derived above.
In comparison with waveguides, a resonator requires one more
boundary condition—vanishing tangential electric field on the end
walls. Let’s assume that the resonator left wall is located at z = 0,
and the right wall is located at z = L. Start with TM modes. To
satisfy the boundary condition Er = Eφ = 0 at z = 0 we choose
two TM modes with the same frequency and the same m and n
indices but opposite values of κm,n (that is two identical waves
propagating in the opposite directions) add them and divide the
result by 2.

329/441
RF modes in cylindrical resonator
Using
1 iκm,n z
(e + e −iκm,n z ) = cos(κm,n z), (25.12)
2
1
(κm,n e iκm,n z − κm,n e −iκm,n z ) = iκm,n sin(κm,n z),
2
it is easy to see that both Er and Eφ = 0 acquire the factor
sin(κm,n z) and hence satisfy the boundary condition at z = 0. In
order to satisfy the boundary condition at the opposite wall, at
z = L, we require κm,n L = lπ, where l = 1, 2, . . . is an integer
number.

330/441
RF modes in cylindrical resonator
The result is
 
 r lπz
Ez = E0 Jm jm,n cos e −iωt−imφ (25.13)
a L
 
κm,n a 0  r lπz
Er = −E0 Jm jm,n sin e −iωt−imφ
jm,n a L
imκm,n a2 
 
r lπz
Eφ = −E0 2
Jm jm,n sin e −iωt−imφ
rjm,n a L
mωa2
 
 r lπz
Br = E0 2 2 Jm jm,n cos e −iωt−imφ
c rjm,n a L
 
iωa 0  r lπz
Bφ = E0 2 Jm jm,n cos e −iωt−imφ .
c jm,n a L

331/441
RF modes in cylindrical resonator
Eq. (25.4) should now be interpreted differently: we replace κm,n
by lπ/L, square it, and find the frequency ω of the mode
2
ω2 2

lπ jm,n
= + . (25.14)
c2 L a2

The modes given by (25.13) and (25.14) are called TMmnl modes.

332/441
RF modes in cylindrical resonator
A similar procedure can be done with the TE modes, but instead
of adding, we need to subtract the mode with negative κm,n from
the mode with the positive κm,n and divide the result by 2i.
An important quantity associated with the mode is the energy W
of the electromagnetic field. This energy is given by the integral
over the volume of the cavity of (0 /2)(Ez2 + c 2 Bθ2 ), where one has
to take the real parts of the fields before squaring them.

333/441
TM010 mode
For illustration, let us calculate the energy of TM010 mode.
ω2010 2
j0,1
= , (25.15)
c2 a2
where j0,1 = 2.4.
 r
Ez = E0 J0 j0,1 e −iω010 t (25.16)
a
iω010 a 0  r 
Bφ = E0 2 J0 j0,1 e −iω010 t .
c j0,1 a
The calculation can be simplified if one notices that although Ez
and Bθ depend on time, the energy W does not. Because there is
a phase shift of π/2 between these fields, one can find a moment
when Bθ = 0, and then
Z
0 0
W = dV |Ez |2 = πE02 a2 LJ12 (j1 ) , (25.17)
2 2
R1
where we used 0 J02 (bx)xdx = 12 J12 (b).
334/441
TM010 mode
With account of the finite conductivity of the wall, one finds that an
initially excited mode decays with time because its energy is absorbed in
the walls. This damping results in the imaginary part γ in the mode
frequency, ω = ω 0 − iγ, where ω 0 and γ are real and positive. The
imaginary part of the frequency can be calculated with the help of the
Leontovich boundary condition.
A related quantity is the quality factor Q of the cavity equal to
ω0
Q= . (25.18)
2γ
The quality factor for the TM010 mode of the cylindrical cavity
aL
Q= , (25.19)
δ(a + L)
where δ is the skin depth at the frequency of the cavity. A crude estimate
of the quality factor is Q ∼ l/δ, where l is a characteristic size of the
cavity (assuming that all dimensions of the cavity are of the same order).
Typical copper cavities used in accelerators have Q ∼ 104 ;
superconducting cavities may have Q ∼ 109 .
335/441
Electromagnetic field pressure
Electromagnetic field terminated by a conducting wall exerts a
force on this wall.

When electric field lines are terminated on metal surface

E
B

z z

there are image charges with the surface density equal to 0 En , (n

is the normal to the surface of the metal).

336/441
Electric field pressure
Consider in more detail distribution of the electric field inside the
metal. The metal occupies the region z > 0. The charge density
inside the metal is ρ(z), and the electric field is Ez (z).

dEz ρ(z)
= . (25.20)
dz 0
The force per unit area is given by the integral
Z∞ Z∞
(E ) dEz 0
fz = dzρEz = 0 dzEz = − En2 . (25.21)
0 0 dz 2

The minus sign means that the electric field has a “negative
pressure”—it pulls the surface toward the free space.
Numerical example: for E = 35 MV/m the pressure is about 0.5
N/cm2 .

337/441
Magnetic field pressure
Tangential magnetic field also exerts a force on the surface.
Assume that the magnetic field By (z) is directed along y , and
varies along z due to the current jx (z) flowing in the x direction.
dHy
= −jx (25.22)
dz
and the force per unit area is
Z∞
(M)
fz = dzjx By . (25.23)
0
We have
Z∞
(M) dHy 1 2
fz =− dzBy = B , (25.24)
0 dz 2µ0 t
(M)
We see that fz is positive—it acts as a real pressure applied to
the surface.

338/441
Side comment on magnetic field pressure
We briefly talked about a pressure tensor in force equations. For a
uniform magnetic field in the z^ direction, the tensor describing the
magnetic pressure away from any boundary is
 2 
B /2µ0 0 0
 0 B 2 /2µ0 0  . (25.25)
0 0 2
−B /2µ0

This has significant consequences for magnetized plasmas. The

negative, longitudinal term is called magnetic tension.

339/441
Electromagnetic field pressure
The effect of the electromagnetic pressure causes a so called
Lorentz detuning in modern superconducting cavities which should
be compensated by a special control system.

340/441
Slater’s formula
What happens to the frequency of a cavity, if its shape is slightly
distorted?

The frequency of modes changes. To calculate the frequency

change we compute the work against the electromagnetic field
needed to change the cavity shape.

341/441
Slater’s formula
Assume that a mode is excited in the cavity and the distortion of
the cavity shape occurs slowly in comparison with the frequency of
the mode. This work, with a proper sign, is equal to the energy
change δW of the mode. Since the distortion is small, we can take
the unperturbed distribution of the electric and magnetic fields on
(E ) (M)
the surface, compute fz + fz and average over the period of
oscillations. This averaging introduces a factor of 12 .
Z  
1 1 2 0 2
δW = dS h B − En , (25.26)
2 2µ0 t 2

where h is positive in the case when the volume of the cavity

decreases, and it is negative in the opposite case, Bt and En are
the amplitude values of the field on the surface.

342/441
Slater’s formula
We know from quantum theory that the number of quanta does
not change in adiabatically slow processes. Hence W /ω = const,
δω δW
= . (25.27)
ω W
This gives us
Z
δω 0
dV c 2 Bt2 − En2 ,


= (25.28)
ω 4W ∆V

where the integration in the numerator goes over the volume of

the dent, and the integration in the denominator goes over the
volume of the cavity. This is often called Slater’s formula.

The perturbation must be small and cannot have any sharp

features, otherwise one cannot use the unperturbed fields.

343/441