Free-Electron Lasers
Free-Electron Lasers
Free-Electron Lasers
c
=
3c
3
2R
(1)
where R is the radius of curvature in the bending magnet. The radiation power is
P
rad
=
e
2
c
6
0
4
R
2
. (2)
In modern synchrotron light sources the radiation used for research is produced in wiggler or undu-
lator magnets which are periodic arrangements of many short dipole magnets of alternating polarity.
The electrons move on a wavelike orbit through such a magnet (Fig. 1) but the overall deection of the
beam is zero. Undulator radiation is far more useful than bending-magnet radiation because it is nearly
monochromatic and concentrated in a narrow angular cone with an opening angle of about 1/. The
wavelength can be estimated from the following consideration. Call
u
the period of the magnet ar-
rangement. In a coordinate system moving with the average speed of the beam, the relativistic length
contraction reduces the period to
u
=
u
/, and the electrons oscillate at a correspondingly higher
1
The total relativistic energy of the electron is denoted by W since in this article the letter E is reserved for electric elds.
477
Fig. 1: Schematic representation of undulator radiation. For simplicity the alternating magnetic eld and the
cosine-like electron orbit have been drawn in the same plane.
frequency
= 2c/
u
= 2c/
u
and emit dipole radiation. If one Lorentz-transforms this radiation
into the laboratory system one gets for the light wavelength
u
/ =
u
/
2
; for example, for a
Lorentz factor = 1000 (an electron energy of 511 MeV) the radiation wavelength is a million times
shorter than the undulator period. Moreover, the wavelength can be easily varied by changing the particle
energy.
It is interesting to note that the total energy radiated by a relativistic electron in an undulator is the same
as that in a bending magnet of equal magnetic length, however, the intensity is concentrated in a narrow
spectral range. Different electrons radiate independently in bending magnets as well as in undulators,
hence the total power produced by a bunch of N electrons is simply N times the radiation power of one
electron.
1.2 Free-electron and conventional lasers
The next big improvement is given by the Free-Electron Laser. The main component is again an undulator
magnet but by means of a clever mechanism (explained below) one forces a large number of electrons
to emit their radiation coherently. Like undulator radiation, the FEL radiation is almost monochromatic
and well collimated but the power will be N times higher if one manages to achieve full coherence in the
bunch.
A conventional laser (Fig. 2) consists of three basic components: the laser medium with at least 3 energy
levels, an energy pump which creates a population inversion, and an optical resonator. The electrons are
bound to atomic, molecular or solid-state levels, so one may call this a bound-electron laser in contrast
to the free-electron laser where the electrons move in vacuum. In an FEL (Fig. 3 ) the role of the active
laser medium and the energy pump are both taken over by the relativistic electron beam. An optical
cavity is no longer possible for wavelengths below 100 nm, because the reectivity of metals and other
mirror coatings drops quickly to zero at normal incidence. Here one has to rely on the principle of Self
Amplied Spontaneous Emission (SASE) where the laser gain is achieved in a single passage of a very
long undulator magnet. The schematic setup of a SASE FEL is shown in Fig. 4. One big advantage of an
FEL in comparison with a conventional laser is the free tunability of the wavelength by simply changing
the electron energy.
2
P. SCHM USER
478
Fig. 2: Scheme of a conventional laser (bound-electron laser).
Fig. 3: Principle of free-electron laser. For visible or infrared light an optical resonator can be used. In the
ultraviolet and X ray region one has to rely on the mechanism of Self Amplied Spontaneous Emission where the
laser gain is achieved in a single passage of a very long undulator.
2 Undulator radiation
2.1 Magnetic eld of undulator
The motion of an electron in an undulator magnet is shown schematically in Fig. 5. The undulator axis
is along the direction of the beam (z direction), the magnetic eld points in the y direction (vertical).
The period of the magnet arrangement
u
is in the order of 25 mm. For simplicity we assume that the
horizontal width of the pole shoes is larger than
u
, then the x dependence of the eld can be neglected.
The eld on the axis is approximately harmonic
B
y
(0, 0, z) = B
0
cos(k
u
z) with k
u
= 2/
u
(3)
In vacuum we have
B = 0, hence the magnetic eld can be written as the gradient of a scalar
magnetic potential
B = .
The potential fullls the Laplace equation
2
= 0 .
3
FREE-ELECTRON LASERS
479
Fig. 4: Schematic of the SASE FEL at the TESLA Test Facility.
y
x
z e
Permanent
Magnet
Pole
Gap
Fig. 5: Schematic view of an undulator magnet with alternating polarity of the magnetic eld and of the cosine-like
trajectory of the electrons. The distance between two equal poles is called the undulator period
u
.
Making the ansatz
(y, z) = f(y) cos(k
u
z)
d
2
f
dy
2
k
2
u
f = 0
we get for the general solution
f(y) = c
1
sinh(k
u
y) + c
2
cosh(k
u
y) ,
B
y
(y, z) =
y
= k
u
(c
1
cosh(k
u
y) + c
2
sinh(k
u
y)) cos(k
u
z) .
The vertical eld component B
y
is symmetric with respect to the plane y = 0 hence c
2
= 0, and
moreover k
u
c
1
= B
0
. So the potential is
(x, y, z) =
B
0
k
u
sinh(k
u
y) cos(k
u
z) . (4)
4
P. SCHM USER
480
For y = 0 the magnetic eld has also a longitudinal component B
z
.
B
x
= 0
B
y
= B
0
cosh(k
u
y) cos(k
u
z) (5)
B
z
= B
0
sinh(k
u
y) sin(k
u
z) .
In the following we restrict ourselves to the symmetry plane y = 0.
2.2 Electron Motion in an Undulator
2.2.1 Trajectory in rst order
We call W = E
kin
+ m
e
c
2
= m
e
c
2
the total relativistic energy of the electron. The transverse
acceleration by the Lorentz force is
m
e
v = ev
B . (6)
This results in two coupled equations
x =
e
m
e
B
y
z z =
e
m
e
B
y
x (7)
which are solved iteratively. To obtain the rst-order solution we observe that v
z
= z v = c = const
and v
x
v
z
. Then z 0 and
x(t)
eB
0
m
e
ck
2
u
cos(k
u
ct) , z(t) ct . (8)
The electron travels on a cosine-like trajectory
x(z) = Acos(k
u
z) with A =
eB
0
m
e
ck
2
u
.
The maximum divergence angle is
max
_
dx
dz
_
max
=
eB
0
m
e
ck
u
=
K
u
2m
e
c
. (9)
Synchrotron radiation of relativistic electrons is emitted inside a cone with opening angle 1/. If the
particle trajectory stays within this cone one speaks of an undulator magnet (Fig. 6):
Undulator:
max
1/ K 1 .
If the trajectory extends beyond the cone the magnet is called a wiggler:
Wiggler: K > 1 .
The special feature of an undulator is that the radiated eld of an electron interferes with itself along the
magnet axis. The consequence is, as we shall see, that the radiation is nearly monochromatic.
2.2.2 Motion in second order
Due to the cosine-trajectory the z component of the velocity is not constant. It is given by
v
z
=
_
v
2
v
2
x
c
_
1
1
2
2
(1 +
2
v
2
x
/c
2
)
_
.
5
FREE-ELECTRON LASERS
481
2. Physical Processes in a Free Electron Laser
B
1
/
A B
Figure 2.1.: Emission of radiation in an undulator.
In the TTF undulator, the deviation from the straight orbit is only 10 m. Syn-
chrotron radiation is emitted by relativistic electrons in a cone with opening angle
1
/. In an undulator, the the maximum angle of the particle velocity with respect to
the undulator axis = arctan(
vx
/vz) is always smaller than the opening angle of the
radiation, therefore the radiation eld may add coherently. In a wiggler,
max
> 1/,
and a broad radiation cone with lower intensity on the axis is emitted. The condition
for an undulator can be rewritten for v
z
c:
1
> arctan
v
xmax
v
z
v
xmax
v
z
Kc
c
=K < 1 (2.10)
Consider two photons emitted by a single electron at the points A and B, which
are one half undulator period apart (gure 2.1):
AB =
u
2
(2.11)
If the phase of the radiation wave advances by between A and B, the electromag-
netic eld of the radiation adds coherently
2
. The light moves on a straight line AB
that is slightly shorter than the sinusoidal electron trajectory
AB:
2c
=
AB
v
AB
c
(2.12)
2
Photons radiated by dierent electrons will however usually be incoherent.
22
Fig. 6: Schematic view of undulator radiation .
We insert for v
x
= x(t) the rst-order solution, then the average z velocity is
v
z
= c
_
1
1
2
2
(1 + K
2
/2)
_
c (10)
It should be noted that the z velocity oscillates about the average
2
z(t) =
c +
cK
2
4
2
cos(2
u
t) with
u
=
ck
u
.
The trajectory in second order reads
x(t) =
cK
u
cos(
u
t) , z(t) =
ct +
cK
2
8
2
u
sin(2
u
t) . (11)
2.2.3 Lorentz transformation into a moving coordinate system
Consider a coordinate system (x
, y
, z
= (t
z/c) = t(1
2
) t/
x
= x =
cK
u
cos(
u
t)
z
= (z
ct)
cK
2
8
u
sin(2
u
t)
The electron orbit in the moving system is:
x
(t
) =
cK
u
cos(
) , z
(t
) =
cK
2
8
u
sin(2
) (12)
with
=
u
. Note that
u
t =
=
u
. Superimposed is a small longitudinal oscillation which will be ignored here, it
leads to higher harmonics in the radiation. The motion is plotted in Fig. 7. In the moving system the
electron emits dipole radiation with the frequency
=
u
and the wavelength
u
=
u
/.
2
This oscillation leads to odd higher harmonics of the undulator radiation, see for example the book by K. Wille. In a helical
undulator the z velocity is constant.
6
P. SCHM USER
482
2 0 2
1.2
0
1.2
x( ) t
z( ) t
oscillation of electron in
co-moving coordinate system
Fig. 7: Oscillation of the electron in the moving coordinate system.
Fig. 8: Radiation characteristics in the laboratory system of an oscillating dipole moving at speeds close to c.
2.2.4 Transformation of radiation into laboratory system
The radiation characteristics of an oscillating dipole moving at relativistic speed is depicted in Fig. 8.
Within increasing Lorentz factor the radiation becomes more and more concentrated in the forward
direction. We are interested in the light wavelength in the laboratory system as a function of the angle
with respect to the beam axis. The Lorentz transformation of the photon energy reads
(1
cos )
=
2c
=
2c
(1
cos ) =
u
(1
cos )
Using ,
=
_
1
1
2
2
(1 + K
2
/2)
_
and cos 1
2
/2 we obtain for the wavelength of
undulator radiation
=
u
2
2
(1 + K
2
/2 +
2
2
) . (13)
2.3 Line shape of undulator radiation
An electron passing an undulator with N
u
periods produces a wavetrain with N
u
oscillations (Fig. 9).
The electric eld of the light wave is written as
E
(t) =
_
E
0
e
i
t
if T/2 < t < T/2
0 otherwise
The time duration of the wave train is T = N
u
0
Spectral intensity for a wave train
with N
u
= 100 periods
Fig. 9: A nite wavetrain.
monochromatic but contains a frequency spectrum which is obtained by Fourier transformation
A() =
1
2
_
+
(t)e
it
dt =
E
0
2
_
+T/2
T/2
e
i(
)t
dt
=
2E
0
sin(T/2)
with =
.
The spectral intensity is
I()
_
sin
_
2
with = T/2 =
N
u
(
.
It has a maximum at =
0
Spectral intensity for a wave train
with N
u
= 100 periods
Fig. 10: Normalized spectral intensity distribution of undulator radiation in a magnet with N
u
= 100 periods.
3 Low-gain FEL
3.1 Energy transfer from electron beam to light wave
We consider the case of seeding, where the initial light wave with wavelength
is provided by an
external source such as an optical laser. The schematic setup of a low-gain FEL is shown in Fig. 11.
The light wave is co-propagating with the relativistic electron beam and is described by a plane electro-
magnetic wave
E
x
(z, t) = E
0
cos(k
t +
0
) with k
/c = 2/
.
8
P. SCHM USER
484
Fig. 11: Principle of low-gain FEL with optical resonator.
Obviously the light wave, travelling with speed c along the z axis, slips with respect to the electrons
whose average speed in z direction is
v
z
= c
_
1
1
2
2
(1 + K
2
/2)
_
< c .
The question is then: how can there be a continuous energy transfer from the electron beam to the light
wave? The electron energy W = m
e
c
2
changes in the time interval dt by
dW = v
Fdt = ev
x
(t)E
x
(t)dt .
The x component of the electron velocity v
x
and the electric vector E
x
of the light wave must point in the
same direction to get an energy transfer from the electron to the light wave. To determine the condition
for energy transfer along the entire trajectory we compute the electron and light travel times for a half
period of the undulator:
t
el
=
u
/(2 v
z
), t
light
=
u
/(2c) .
Figure 12 illustrates that after a half period v
x
and E
x
are still parallel if the phase of the light wave has
slipped by , i.e.
(t
el
t
light
) = .
(Remark: also 3, 5 . . . are possible, leading to higher harmonics of the radiation). This condition
allows to compute the light wavelength:
=
u
2
2
_
1 +
K
2
2
_
which is identical with the wavelength of undulator radition (in forward direction).
3.1.1 Quantitative treatment
The energy transfer from an electron to the light wave is
dW
dt
= ev
x
(t)E
x
(t)
= e
cK
sin(k
u
z)E
0
cos(k
t +
0
)
9
FREE-ELECTRON LASERS
485
z
v
x
v
x
v
x
v
x
E
x
E
x
electron trajectory electron trajectory
light wave
Fig. 12: Condition for energy transfer from electron to light wave.
=
ecKE
0
2
[sin((k
+ k
u
)z
t +
0
) sin((k
k
u
)z
t +
0
)]
We consider the rst term. The argument of the sine function is called the ponderomotive phase:
(k
+ k
u
)z
t +
0
= (k
+ k
u
)
ct
t +
0
. (14)
There will be a continuous energy transfer from the electron to the light wave if is constant (indepen-
dent of time) and in the range 0 < < , the optimum value being =
0
= /2. The condition
= const can be fullled only for a certain wavelength.
= const
d
dt
= (k
+ k
u
) v
z
k
c = 0 . (15)
Insertion of v
z
permits to compute the light wavelength:
=
u
2
2
_
1 +
K
2
2
_
. (16)
The condition for resonant energy transfer all along the undulator therefore yields exactly the same light
wavelength as is observed in undulator radiation at = 0. This is the reason why the spontaneous undu-
lator radiation can serve as a seed radiation in the SASE FEL.
Now we look at the second term. Here the phase of the sine function cannot be kept constant since from
(k
k
u
)
ct
t +
0
= const
we would get
k
(1
) = k
u
= 2/
< 0
which of course is unphysical. Hence the second sine function oscillates rapidly and averages to zero.
3.2 The pendulum equation
We assume seeding by an external light source with wavelength
r
m
e
c
2
is dened by the equation
=
u
2
2
r
_
1 +
K
2
2
_
. (17)
Let the electron gamma factor be slightly larger, >
r
, and call = (
r
)/
r
the relative energy
deviation. We assume
0 < =
r
r
1 .
10
P. SCHM USER
486
0
Rotation
Oscillation
r
)
Fig. 13: Phase space curves of a mathematical pendulum.
The energy deviation
r
m
e
c
2
and the ponderomotive phase will both change due to the interaction
with the radiation eld. The low-gain FEL is dened by the condition that the electric eld amplitude
grows slowly such that E
0
const during one passage of the undulator.
The time derivative of the ponderomotive phase is no longer zero for >
r
:
= k
u
c k
c
1 + K
2
/2
2
2
.
We subtract 0 = k
u
c k
c (1 + K
2
/2)/(2
2
r
), see Eq. (16), and get
d
dt
=
k
c
2
_
1 +
K
2
2
__
1
2
r
2
_
.
From this follows
d
dt
2k
u
c
k
u
c) . (18)
The time derivative of is
d
dt
=
eE
0
K
2m
e
c
2
r
sin . (19)
Combining Eqs. (18) and (19) we arrive at the so-called Pendulum Equation of the low-gain FEL
+
2
sin = 0 with
2
= eE
0
Kk
u
/(m
e
2
r
) . (20)
3.3 Phase space representation
There is a complete analogy with the motion of a mathematical pendulum (Fig. 13). At small amplitude
we get a harmonic oscillation. With increasing angular momentum the motion becomes unharmonic. At
very large angular momentum one gets a rotation (unbounded motion). The phase space trajecory for an
electron in an FEL can be easily constructed by writing the coupled differential equations (18) and (19)
as difference equations and solving these in small time steps. The trajectories for 20 electrons of different
initial phases
0
are shown in Fig. 14 for =
r
and >
r
. In the rst case the net energy transfer is
zero since there are as many electrons which supply energy to the light wave as there are which remove
energy from the wave. For >
r
, however, the phase space picture clearly shows that there is a net
energy transfer from the electron beam to the light wave. This will be computed in the next section.
11
FREE-ELECTRON LASERS
487
/2 0 /2
0.06 %
0.04 %
0.02 %
0
0.02 %
0.04 %
0.06 %
a)
/2 0 /2
0.06 %
0.04 %
0.02 %
0
0.02 %
0.04 %
0.06 %
b)
/
r
/
r
Fig. 14: Phase space trajectories for 20 electrons of different initial phases
0
. Left picture: =
r
. The electrons
with
0
< 0 withdraw energy from the light wave while those with
0
> 0 supply energy to the light wave.
Obviously the net energy transfer is zero for =
r
. Right picture: >
r
. One can easily see that the net energy
transfer is positive.
3.4 Computation of the FEL gain, Madey theorem
The energy (per unit volume) of the light wave is
W
light
=
0
2
E
2
0
.
The energy increase and relative gain caused by one electron are
W
light
= m
e
c
2
r
G
1
=
W
light
W
light
=
2m
e
c
2
0
E
2
0
.
Here is the change of the relative energy deviation of the electron upon passing the undulator. We
use Eq. (18) to compute this change:
=
2k
u
c
.
Summing over all electrons in the bunch (n
e
per unit volume) the total gain becomes
G =
m
e
c
r
n
e
0
E
2
0
k
u
<
> . (21)
Here <
> denotes the change of the time derivative of the ponderomotive phase, averaged over all
electrons. Hence it is this quantity we have to compute.
3.4.1 Phase change in undulator
We multiply the pendulum equation
+
2
sin = 0 with 2
and integrate over time
2
2
2
cos = const
(t)
2
=
2
0
+ 2
2
[cos (t) cos
0
] .
From Eq. (18)
0
=
(0) = 2c k
u
we obtain then
(t) =
_
1 + 2(/
)
2
[cos (t) cos
0
] . (22)
12
P. SCHM USER
488
For a weak laser eld one nds (/
)
2
1, so we expand the square root up to second order
1 + x = 1 + x/2 x
2
/8 . . .
and get
(t) =
+
2
0
=
0
= 0.
First order: the phase (t) in rst order is obtained by integrating
0
:
1
(t) =
0
+
0
t =
0
+
t .
We insert this in Eq. (23) to get
in rst order
1
(t) =
+ (
2
/
)[cos(
0
+
t) cos
0
] . (24)
The ight time through the undulator is T, so the change of
1
when the electron passes the undulator is
1
= (
2
/
)[cos(
0
+
T) cos
0
] .
According to Eq. (21) the gain is obtained by averaging
over all particles in the bunch which means
that one has to average over all initial phases
0
. The result is
<
1
>= 0 . (25)
The FEL gain is zero in rst order. The physical reason is the nearly symmetric initial phase space dis-
tribution.
Second order: Equation (24)) is integrated to get in second order:
2
(t) =
0
+
t
. .
1
(t)
+(/
)
2
[sin(
0
+
t) sin
0
t cos
0
]
. .
2
(t)
(26)
This is inserted in Eq. (23) to compute
at t = T in second order
2
(T) =
+ (
2
/
)[cos(
0
+
T +
2
) cos
0
]
4
/(2
3
)[cos(
0
+
T +
2
) cos
0
]
2
(27)
2
1 cos(
0
+
T +
2
) cos(
0
+
T)
2
sin(
0
+
T)
cos(
0
+
T +
2
) cos(
0
+
T)
(/
)
2
sin(
0
+
T)[sin(
0
+
T) sin
0
T cos
0
]
Averaging over all start phases
0
yields
< cos(
0
+
T +
2
) >= (1/2)(1 cos(
T)
T sin(
T)).
From this we get
<
2
>= (
4
/
3
)[1 cos(
T) (
T/2) sin(
T)] .
13
FREE-ELECTRON LASERS
489
.. , 10 9.97 8
I ( )
sin( )
2
G( )
. .
2
sin( )
2
cos( )
.
2
sin( )
2
3
spectral line of undulator gain of FEL
10 0 10
0
0.5
1
I( )
10 0 10
1
0.5
0
0.5
1
G( )
0
Fig. 15: The normalized lineshape curve of undulator radiation and the gain curve (arbitrary units) of the low-gain
FEL as a function of = N
u
(
)/
.
Remembering that T = N
u
u
/c is the ight time through the undulator and =
2
> =
3
[1 cos(2) sin(2)]
=
N
3
u
3
u
4
8c
3
d
d
_
sin
_
2
The FEL gain function (21) is hence
G() =
e
2
K
2
N
3
u
2
u
n
e
4
0
m
e
c
2
3
r
d
d
_
sin
2
2
_
(28)
We have proven the Madey Theorem which states that the FEL gain curve is given by the negative
derivative of the line-shape curve of undulator radiation. This is shown in Fig. 15.
4 High-gain FEL
4.1 General principle
The essential feature of the high-gain FEL is that a large number of electrons radiate coherently. In that
case, the intensity of the radiation eld grows quadratically with the number of particles: I
N
= N
2
I
1
. If
it were possible to concentrate all electrons of a bunch into a region which is far smaller than wavelength
of the radiation then these N particles would radiate like a point macroparticle with charge Q = Ne,
see Fig. 16. The big problem is, however, that this concentration of some 10
9
electrons into a tiny volume
is totally unfeasible, rather even the shortest particle bunches are much longer than the FEL wavelength.
The way out of this dilemma is given by the process of micro-bunching which is based on the following
principle: those electrons which lose energy to the light wave travel on a cosine trajectory of larger
amplitude than the electrons which gain energy from the light wave The result is a modulation of the
longitudinal velocity which eventually leads to a concentration of the electrons in slices which are shorter
than
. The result of a numerical simulation of this process is shown in Fig. 17. The particles within a
micro-bunch radiate coherently. The resulting strong radiation eld enhances the micro-bunching even
further. The result is a collective instability, leading to an exponential growth of the radiation power.
The ultimate power is P N
2
c
where N
c
is the number of particles in a coherence region. A typical
value is
N
c
10
6
P
FEL
= 10
6
P
undulator
.
14
P. SCHM USER
490
Fig. 16: Radiation from a point-like macroparticle
x
/
m
m
0 2 4 2
0.2
0.1
0
0.1
0.2
x
/
m
m
0 2 4 2
0.2
0.1
0
0.1
0.2
x
/
m
m
0 2 4 2
0.2
0.1
0
0.1
0.2
a) b) c)
Fig. 17: Numerical simulation of microbunching. (Courtesy S. Reiche).
4.2 Approximate analytical treatment and experimental results
An approximate analytic description of the high-gain FEL requires the self-consistent solution of the
coupled pendulum equations and the inhomogeneous wave equation for the electromagnetic eld of the
light wave. In the one-dimensional FEL theory the dependencies on the transverse coordinates x, y are
disregarded. The wave equation for the radiation eld E
x
reads
2
E
x
z
2
1
c
2
2
E
x
t
2
=
0
j
x
t
where the current density
j is generated by the electron bunch moving on its cosine-like trajectory. In
addition, one has to consider the longitudinal space charge eld E
z
which is generated by the gradually
evolving periodic charge density modulation. After a lot of tedious mathematical steps and several
simplifying assumptions one arrives at a third-order differential equation for the amplitude of the electric
eld of the light wave:
d
3
E
x
dz
3
4ik
u
d
2
E
x
dz
2
+ (k
2
p
4k
2
u
2
)
d
E
x
dz
i
3
E
x
(z) = 0 . (29)
15
FREE-ELECTRON LASERS
491
Fig. 18: Observation of microbunching at the 60 m FEL Firey.
Here we have introduced the gain parameter and a parameter k
p
=
_
0
K
2
e
2
k
u
n
e
4
3
m
e
_
1
3
, k
p
=
_
4
2
c
3
K
2
(30)
and assumed that the electron beam has negligible energy spread.
This third-order differential equation can be solved analytically. For the case =
r
one obtains
E
x
(z) = A
1
exp (iz) + A
2
exp
_
i +
3
2
z
_
+ A
3
exp
_
i
3
2
z
_
. (31)
The second term exhibits exponential growth as a function of the position z in the undulator. The electric
eld grows exponentially as exp(
3
2
z), the power grows as exp(
pierce
=
u
4
L
g
=
1
3
. (32)
The above calculations, which have been sketched only very briey, indicate that there is an onset of
an instability, leading to a progressing microbunching and an exponential increase in radiation power
along the undulator. A quantitative treatment requires elaborate numerical simulations (see Fig. 17).
Microbunching has been experimentally observed at the 60 m FEL Firey at Stanford University, see
Fig. 18. The exponential growth of radiation power and the progressing microbunching in a long undula-
tor are depicted in Fig. 19. One characteristic but quite undesirable feature of a SASE FEL is the presence
of uctuations which are due to the stochastic nature of the initiating undulator radiation. The light pulse
energy uctuates from pulse to pulse, and the same applies for the wavelength and the time structure. As
an example I show in Fig. 20 the measured wavelength distribution of several FEL pulses together with
the average over 100 pulses. The origin of the large uctuations is that small statistical uctuations in
the incoming undulator radiation are strongly amplied in the exponential growth region (see Fig. 19).
There are so-called seeding schemes under development where radiation of the desired wavelength is
produced in a short undulator, passed through a monochromator and then used as seed radiation for the
FEL process. Here one can expect much higher monochromaticity of the nal FEL radiation. In spite
of the uctuations the SASE radiation is highly coherent as demonstrated by the double-slit interference
patterns in Fig. 21.
Bibliography
K.Wille, The Physics of Particle Accelerators, Oxford University Press 2001
E. L. Saldin, E. A. Schneidmiller, M. V. Yurkov, The Physics of Free Electron Lasers, Springer 2000
16
P. SCHM USER
492
Fig. 19: The exponential growth of radiation power as a function of undulator length. The data at = 100 nm
have been obtained at the SASE FEL of the TESLA Test Facility at DESY. The progressing microbunching is
indicated schematically.
4. The TTF Accelerator and SASE-FEL
4.5.2. Measurements of the spectrum
The spectral distribution of the radiation pulse could be measured with a spectrom-
eter [GFL
+
01]. It consisted of a normal incidence grating with 1 m focal length; the
image which was focused on a uorescent screen was imaged with tandem optics pro-
viding a very large aperture. This set-up was able to capture 5.3% of isotropically
emitted light from the screen. The image intensier was coupled to a CCD, which is
digitised directly in the device. It was controlled and read out by a bre optical link
from the control room. A wavelength resolution of 0.2 nm has been measured with
this spectrometer, while the imaged range is of 7 nm [GFL
+
01]. Single-shot spectra
could be recorded using the short exposure times of the image intensier (gure 4.19).
94.5 95 95.5 96 96.5 97 97.5 98 98.5 99
0
10
20
30
40
50
60
Wavelength / nm
I
n
t
e
n
s
i
t
y
/
a
r
b
.
u
n
i
t
s
\
| Individual spectra
/
Average of 100
Figure 4.19.: Measured spectra of FEL pulses. Individual electron bunches produce
dierent spectral distributions. An averaged spectrum is also shown.
94
Fig. 20: The measured spectra of three pulses in the SASE FEL at DESY. Also shown is the average spectrum of
100 pulses.
7.1. Measurements of the FEL in saturation
x / mm
y
/
m
m
a)
3 2 1 0 1 2 3
3
2
1
0
1
2
3
x / mm
y
/
m
m
b)
3 2 1 0 1 2 3
3
2
1
0
1
2
3
x / mm
y
/
m
m
c)
3 2 1 0 1 2 3
3
2
1
0
1
2
3
x / mm
y
/
m
m
d)
3 2 1 0 1 2 3
3
2
1
0
1
2
3
Figure 7.2.: Measured horizontal double slit diraction patterns at 100 nm FEL
wavelength. a) 0.5 mm, b) 1 mm, c) 2 mm and d) 3 mm slit separation.
125
Fig. 21: Measured double-slit diffraction patterns at 100 nm FEL wavelength. The separation of the horizontal
slits is 0.5 resp. 1 mm.
17
FREE-ELECTRON LASERS
493