SLAC KLYSTRON LECTURES
Lecture 3
February 4, 2004
SPACE-CHARGE WAVE THEORY
George Caryotakis
Stanford Linear Accelerator Center
caryo@slac.stanford.edu
1
SPACE-CHARGE WAVE THEORY
Lecture 3
In kinematic theory, which was the earliest to describe beam interaction with
microwave fields, the fields due to space charge in the electron beam were ignored and
calculations were based only on the fields produced by the circuit with which the beam
is interacting. Energy between cavities was transmitted through ballistic electron
motion. Kinematic theory does not provide the tools necessary to analyze the smallsignal behavior of multicavity klystrons, which is based on space-charge waves.
In what follows, we shall develop the space-charge wave formulation and employ
it to derive expressions that are useful in calculating the gain and bandwidth of
multicavity klystrons, and linear beam microwave tubes, in general. We shall also
derive an equation for beam loading, which is consistent with Eq. (2-60), but cast in a
form that is both more informative, and useful for computer calculations. The theory
follows Chodorow[1], Zitelli[2], and Wessel-Berg[3]. Derivations are abbreviated. The
objective, as in Lecture 2, is to expose the student or engineer to the origins of design
equations and highlight the assumptions that are involved in their derivation. These
equations will be subsequently used in examples of actual klystron designs.
2
The initial basic assumptions are:

•
•
No electron motion or current is assumed to exist along any dimension, other than the zdimension. This requires the presence of a strong magnetic field, which is almost always the
case with modern klystrons.
All electrical quantities will be written as sums of the dc and ac terms, the ac term being small
compared to the dc term (small-signal). Cross products of ac terms will be neglected.
This will be a non-relativistic analysis. Relativistic correction formulae are provided in an
Appendix.
Following are the equations for total velocity, current density and charge density, respectively,
dc terms are denoted by the subscript 0. The ac terms u, i and r will be functions of the coordinates
transverse to the electron motion (x and y in Cartesian coordinates) and the z dependence is
exponential, indicating a traveling wave
ut  u0  ue j (t  z )
it  i0  ie
j (  t  z )

(1)
 t   0   e j (  t  z )
 In Lecture 2, the symbol used for the total current was It, and for the rf current I1. These were not necessarily small-signal
currents. Here, everything is small-signal and lower-case letters will be used. Naturally, i0 is identical to Io.
3
We shall derive expressions for space-charge waves from three
general equations, a) the wave equation for the electric field (derived from
Maxwell’s equations), b) the continuity of charge and c) the force
equation. Equations initially in Cartesian and finally in cylindrical
coordinates will be derived for the quantities above and manipulated to
find expressions for the wave motion of current between klystron cavities.
This is expressed in the form ej(ωt-γz), representing a traveling wave with a
propagation constant γ, which must be derived. It is intuitively obvious
that γ will be a function of the space charge, the voltage, the frequency and
the geometry of the drift tube. Furthermore, we should be able to derive
an expression for the rf current imparted to the beam through velocity
modulation by a cavity gap voltage V1, with the small-signal condition
V1<<V0, and with space charge taken into account.
The effect of the space charge is to act as an elastic medium
between the concentrations of electrons (bunches) which form as a result
of velocity modulation. The result is the formation of waves similar to
those formed in spring-weight systems. In the early days of klystrons, the
RF current calculated as in (2-18) was reduced by a “debunching factor”[4]
sin(hz)/hz, where z was the distance from the center of the cavity gap (with
the V1 modulating voltage) where the current was calculated, and h/2 was
known as the “wave number.” In the derivations which follow, h will be
replaced by its modern equivalent ωp, the “plasma frequency”, a function
of the charge density ρ, and modified further by a reduction factor R,
which takes into account the effect of the drift tube walls in reducing the
effects of the space charge between bunches (hence increasing the rf
current which was reduced by the “debunching factor”.
4
Let us repeat here Eq. (2-18), which is a large signal expression of the fundamental current in a
beam which has been modulated by a klystron cavity.
I1  2 I 0 J1 ( X )e j (t 0 )
the series representing the Bessel function reduces as follows, for X<<1 (V1<<V0),
X X3
X MV10
J1 ( X )  
 .. 

2 16
2
4V0
The small-signal current, reduced by the “debunching factor”, would then be, empirically,
i
I 0 M  0 sin(  p l ) j (t  e z )
e
V1
2V0
 pl
After some algebra, using 0  ez and e/p = /p, and recognizing that the original modulation voltage
V1 in Eq. (2-18), was actually V1sint and that the equivalent in complex notation is –jV1 (Re[jV1ejt]=V1sint), we finally have an important equation relating the small-signal current to the
modulating cavity voltage (time dependence omitted) .
ij
I0

M
sin(  pl )e je zV1
2V0  p
(2)
We shall now derive the above formula and show that p must be modified by the presence
of a drift tube. The mathematics follows. The electric field in the drift tube is,
E  E( x, y)e j (t  z )
As indicated above, γ is the unknown propagation constant, whose value is to be
determined. We begin with the inhomogeneous wave equation (i.e., a differential equation
containing “sources,” involving the charge and the current to the right of the equal sign). It is the
equation that governs the operation of every electron device.
5
1 2 E 1
i
 E  2 2    0
c t
0
t
2
(3)
where c is the velocity of light, 0 and 0 the permittivity and permeability of free space, and k
= ω/c is the wave number in free space.
We will work with the z-components of the vectors E, i and  . The other components can be
derived from Ez. We then have (in Cartesian coordinates),
 2 Ez  ( 2  k 2 ) Ez 
where,
1
0
 z   j0iz

dz
 2 Ez  2 Ez
2
  Ez 

x 2
y 2
(4)
z  
(5)
The objective in what follows is to eliminate ρ and iz from (4) and obtain an equation in Ez
alone. The equation of continuity is,
iz 

0
z t
(6)
which can be written (from Eq. 1)
 j iz  j  0
(7) 6
We write current and space charge as the sums of dc and small-signal rf quantities, and neglect
products of rf current and rf charge. In what follows the z-subscripts will be omitted for current
and velocity.
it  ( 0   )(u0  u)  i0  0u  u0
(8)
i  0u   u0
(9)
and
The force equation is (neglecting the term u  B), which is negligible, except for relativistic
velocities), and with = e/m
 Ez 
du
dt
(10)
The total velocity derivative with time in the equation above needs to be distinguished from
the partial derivative used to observe the behavior of the beam at a fixed point. The total
derivative of u(z,t) is equal to the sum of changes, with respect to both time and distance.
du u u dz u
u



 u0
dt t z dt t
z
E 
u
u
 u0
 j (  u0 )u
t
z
(11)
(12)
7
We now need to eliminate ρ and i from (4) and reduce it to a homogeneous form. To do
that, we require two more relationships. From (7) and (9)

0
  u0
u
(13)
and
i
0


u

  u0
(14)
Substituting for u from Eq. 12 into both equations above from (12) we have

0
Ez
j (  u0 ) 2
(15)
and
i
0
Ez
2
j (  u0 )
(16)
8
Substituting these into (4), we obtain, omitting much of the arithmetic, and using
c2 = 1/ε0μ0 (velocity of light),
2
2
2

2
2
2   / c 
 Ez   (  k )  P
E 0
2  z
(


u

)
0


2

which can be finally written as

P2  2  k 2 
2
2
 Ez   (  k )  2
E 0
2 z
u
(



)
0
e


2

(17)
We have re-introduced the electron wave number e = ω/u0, and we define the plasma frequency,
p  
0
0
(18)
We thus have obtained a homogeneous wave equation of the form
 2 Ez  T 2 Ez  0
(19)
and defining p = ωp/u0 as the plasma wave number,
  p2

T  (  k ) 
 1
2
(



)
 e

2
2
2
(20)
9
The propagation constant γ continues to be the unknown to be determined.
To solve (20) we must specify appropriate boundary conditions. Depending on the coordinate
system, the solutions will be trigonometric or Bessel functions. They will be summations over the Tn’s,
with four separate n’s to be calculated from each Tn, since (20) is quartic in . Before deriving a
useful solution, let us consider the case of an infinite beam, where Ez does not depend on coordinates
that are transverse to z. Then, (19) becomes
 2 Ez  0
(21)
which requires T = 0, which results by inspection of (20) to two pairs of propagation constants.
The first pair of roots is
 1,2   k
The second pair is,
(22)
 3,4   e   p
(23)
The first two solutions are obviously plane waves in free space, traveling with the velocity of
light. The second two are plasma waves with propagation velocities above and below the beam
velocity. There are no transverse boundary conditions to be considered since the transverse space is
infinite. The overall solution of equation (21) must include all four solutions, properly matched to
the initial conditions at z = 0. The velocity solution has the same form as that for Ez:
n4
u / u0    n e  j n z
n 1
The Λs (lamdas) are to be determined from the boundary conditions at z = 0.
(24)
10
Since, from (14)
e
i

u
i0  e  
(25)
The current can be written as,
n4
i / I0  
e
n 1 e   n
ne
 j n z
(26)
At the boundary z = 0, the total velocity was first approximated in Eq. 1-4. The current is
zero. Using the definition for the ns in (22)
and
u
u0
i
I0
MV1
2V0
(27)
e

3  e  4  0
p
p
(28)
 1   2  3   4 
z 0
 1   2 
z 0
The equations above have the solutions
1   2  0
11
and
3   4 
MV1
4V0
(29)
The rf current in the beam can then be written as
i
MI 0V1   j ( e   p ) z  j ( e   p ) z
(e
e
)
4V0  p
and finally
MI V 
ij 0 1
sin(  p z )e je z
2V0  p
(30)
This is close to the desired result, except of course we need an expression that is valid for a
cylindrical or sheet beam, not one of infinite dimensions.
For a pencil beam, not filling a cylindrical drift tube, the solution in cylindrical coordinates
contains ordinary Bessel functions (J0) in the beam and modified Bessel functions (I0, K0) in the
region between the beam and the drift tube.
This equation has, as before, four solutions, of which two correspond to perturbed waveguide
modes, and the other two to space-charge waves. We will make the assumption that the former
have velocities almost equal to velocity of light, and hence that βe>>k. Eq. (20) then becomes,
 m 21,2  k 2 
Tm2
 p 
1 


 e
2
(31)
12
which are perturbed waveguide modes, and reduce to these if p=0
For the plasma wave modes, we can set γ  e , everywhere, except where for the difference
between the two appears.
Then,
 m3,4  e 
and, finally,
p
 T 
1   m 
  e 
 qm
R

p

2



 e   qm
1
 Tm  
1    
   e  
2
(32)
where the ratio of the “reduced plasma frequency” to the plasma frequency is defined as “R”,
the “plasma reduction factor”. We must first solve the wave equation (19) for the particular
geometry of beam and drift tube. To determine Rm, we shall neglect all higher order Tm’s and use
only the T0 eigenvalue. Branch and Mihran[5] have published solutions and graphs for a number
of geometries. For the pencil beam in a cylindrical drift tube, an approximate solution of (19) can
be determined from the transcendental equation below, which results from matching solutions, at
the boundary r = b, inside the beam (J0’s and J1’s) and in the space between beam and drift tube
(K0’s and K1’s). The approximation  =e is used. (See Fig. 1 for a graphical solution of (34)
T0
J1 (T0b)
K (  a ) I (  b)  K1 (  eb) I 0 (  e a )
 e 0 e 1 e
J 0 (T0b)
K 0 (  eb) I 0 (  e a )  K 0 (  e a ) I 0 (  eb)
(33)
13
Fig. 1 Graphical solution of Eq. (34)
From the above, the operating formula used in design is,
.
R
1
T 
1  0 
 e 
2
(34)
Using a similar procedure as in the case of no drift tubes, it should be obvious that the
expression for the current is the same as in Eq. (30), with q substituted for p. The drift tube is
assumed to be cut off. Eq. (35) is the fundamental small-signal transconductance equation
relating the rf current i to the voltage V1 across a cavity gap, at distance z down the drift tube
from the center of that gap.
MI 0V1 
ij
sin( q z )e je zV1
2V0 q
(35)
14
R=
Figure 2: Plasma frequency reduction factor, vs. ea for a beam of radius b
inside a drift tube of radius a, with the inverse of the beam-filling factor b/a as a parameter
15
A similar equation can be used to determine the plasma reduction factor for a sheet beam
of thickness t centered between parallel plates spaced by a distance W. In rectangular
coordinates, the transcendental equation corresponding to (34) is
 Tt   t
Tt
 (W  t )
tan    coth
2
2
2 2
(36)
It should be remembered that Equations (34) and (36) contain the approximation
2
o <<c and are therefore not relativistic. A simple relativistic correction is not readily
available, but the correction is not thought to be significant.
u2
Having derived expressions for the plasma reduction factor R, we can now proceed to
analyze the interaction between beam and rf circuit in one dimension, using plasma theory. We
shall consider only the z-dimension, along the path of the electron beam. This is not restrictive,
because formulae exist, e.g. Eq. (2-59), which averages beam parameters over the beam, when
transverse dimensions matter. Computer simulations do the same thing, even more accurately.
Higher-order modes in the drift tubes or in the circuits with which the beams interact are also
neglected. Drift tubes are cut off below the operating frequency. Beams are assumed “cold”,
with no thermal velocities.
The mathematical treatment follows Wessel-Berg. We shall skip some steps to show the
method, and jump to the final, very useful, beam-loading formula. This reduces, for zero space
charge (p = o) to the Gb/G0 Eq (60) cited in Lecture 2.
The objective is to derive equations relating in detail circuit fields to the current of
the beam interacting with these fields. The concept of positive and negative energy space-charge
waves provides important insight to the operation of the klystron.
16
We proceed as earlier in this section by using the equations for continuity (6),
current (8) and force (10). Wessel-Berg defines the “kinetic voltage” U(z)
U ( z) 
u0u

We will be omitting z-subscripts since z is the only dimension. We can then write, after some
algebra


j


 e z  U  Ez  Eb  Ec
(37)
i0
 1

j


i

j

U
e
 e z 
2
V0
(38)
where Es is the field due to the space charge and Ec is the circuit field. These will be used
to eventually derive expressions for energy transfer and beam loading. The assumption is
made that superposition applies and that the E-field in the total system is the sum of the circuit
and beam sub-systems, with coupling between the two removed.
Now, notice that Es can be related to the current i through one of the Maxwell equations:
 H  i 
D
 i  j 0 Es
t
(39)
Taking the divergence, which is zero, we obtain for the truly one-dimensional case (beam
of infinite cross-section), where there are no transverse fields and the z-dependence of field
and current are the same,
  Eb 
Eb
1 i

z
j 0 z
(40)
17
Integrating, and since i(0) = 0,
Eb  
1
j 0
(41)
i
Recall that in the earlier three-dimensional derivation, where there were no circuit
fields and Ez was equivalent to the present Eb, we obtained Eq. (16). This can be
rewritten as,
Ez  Eb  
1
j 0
(  u0 ) 2

2
p
(42)
i
In the case where the beam does not fill the drift tube, we write, as before,    e   q
Then,
1  q
Es  

j 0   p
2

1
R 2i
 i  
j 0

Comparing (43) with (41), we see that for a finite beam, the effect of the radial field
leakage to the walls of the drift tube is to reduce the longitudinal space-charge field Ez by R2,
a number less than unity, previously derived. The effect of the walls is illustrated in Figure 3
below, which shows how the field between bunches is weakened by the drift tube wall.
(43)
18
Figure 3. Fields from space charge bunches, a) infinite beam; b)
finite beam in freespace; c) finite beam in conducting cylinder.
These fields are considered to be the net change in space charge
(as a result of bunching) from the space charge of the
unmodulated beam.
Wessel-Berg now defines a quantity W as the “beam characteristic impedance”
W 
2V0  q
I0 e
(44)
Then, after manipulating (43), (44), and the expressions for  p and R
Es  j  qWi
(45)
19
Equations (37) and (38) become


j


U  j  qWi  Ec
e

z 

1

j


i

j

U
q
 e z 
W
(46)
(47)
Wessel-Berg uses both of these equations in analogy to transmission line equations to
further develop details of the space-charge theory. They are combined to form a second order
equation for the current i. The partial signs are dropped, since z is the only variable.
d 2i
di
1
2
2

2
j


(



)
i

j

Ec
e
e
q
q
2
dz
dz
W
(48)
Using Laplace transforms, Equation (48) can be integrated subject to the initial conditions,
at z = 0, i= 0 and di/dz = 0. The result is:
z
1
i( z )  j  Ec ( )e je ( z  ) sin  q ( z   )d
W 0
(49)
20
Assume now a two-cavity klystron with two interaction regions, symmetrical about suborigins ξ = 0, with lengths d1 and d2. Except for their symmetry, these regions are quite general,
with arbitrary fields. The integration only needs to be performed the two regions, where the
circuit fields are Ec1 and Ec2. The fields are zero elsewhere.
We need to introduce some shorthand to make the equations that follow manageable. Define:
F (  e ) 
1
 F (  e   q )  F (  e   q ) 
2
(50)
It can be shown that:
  aF (  e )  bF (  e )   aF (  e )  bF (  e )
 dF (  e ) 
d


 F (  e ) 

d

d

e
e


  F (  e ,  )d    F (  e ,  )d 
(51)
21
M1 
d1
2
1
j (  e   q )
E
(

)
e
d
c
1

V1 d1

Slow wave:
M1 
Fast wave:
2
βe- = βe + βq
1
V1
d1
2
E
c1

( )e
j (  e   q )
d
d1
2
(52)
βe+ = βe – βq
It should be clear that the + signs stand for a higher wave number and hence a slow wave,
while – signs indicate fast waves. Clearly, the equations above are the definitions for the slow
and fast wave coupling coefficients. Now, as in Lecture 2, write for the fields in the two
regions,
Ec1,2  V1,2 f1,2 ( )
(53)
and normalize them as in the equation below:
d
2
d  f1,2 ( ) f1,2 ( )d  1

(54)
d
2
It follows that
d
2
V1,2V 1,2  d  E ( )E  ( )d

(55)
d
2
Where the asterisks signify complex conjugates. Eq. (55) clearly represents the square of the
magnitude of the gap voltages
22
From the foregoing, it can be shown that the following equation follows,
V2  2
V1
j  e ( l  2 )
i ( 2 )    e
M 2 (  e )     f ( 2 )e j e2 d  2
W
W d2

(56)
2
Expanding the first term of (56), one obtains,
V
i(2 )   1
2W
d
d
 j ( e  q )( l  )
 j ( e  q )( l  ) 

2
2
 M 1 (  e   q )e
 M 1 (  e   q )e



(57)
The significance of the above equation is the following: The current in the interaction Region 2 is
composed of a superposition of fast and slow space-charge waves originating within Region 1, as a result
of the voltage V1 there, plus currents originating within Region 2
The complex power imparted to, or extracted from the beam at Region 2, is given by the following
expression,
P2 
1
2
d2
2
1 
*
i
(

)
E
(

)
d


V2
2
2
2
d
2
2

2
d
2
 i( ) f ( )d
2
2
2
d

2
23
This, and a great deal of algebra leads the final result , (in the coming weeks the entire derivation will be posted)
Gb 
And since,
1
4W
 M    2  M    2   1 e I 0  M  2  M  2 
 e q   8  V 
q
  e

q
0
(55)
 M   2  M    2
 e q
 e q  
 M    2  M      2 
 e q  
q
q
  e
2 q
2
It follows that,
lim Gb  
 q 0
Hence,
e I 0 
M
4 V0  e
Gb
 M
 e
G0
4  e
2
2
IIn other words, the expression above, which was stated without proof in the first lecture
(Eq. n60), as the result the result of kinematic analysis, is equivalent to (55) in the absence of space
Charge ( q→0).
o
Equation (55) is fundamental to the design of extended interaction klystrons.
t
Equations
(51) and (52) form the basis for calculating beam loading by using simulated electric
h
fields
e and performing numerically the integrations indicated. A negative value for Gb can imply a
“monotron”
instability. (Energy in the fast wave is lower than that ion the slow wave). This is a
r
condition where the exchange of energy between beam and circuit is such that net energy flows to
the w
circuit, which will develop excessively high fields, i .e. an oscillation, unless it is sufficiently
loaded.
o
24
25