theory and practice of free-electron lasersuspas.fnal.gov/materials/09unm/day 2.pdf · theory and...
TRANSCRIPT
22--11LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Theory and Practice ofFree-Electron Lasers
Particle Accelerator SchoolDay 2
Dinh Nguyen, Steven Russell& Nathan Moody
Los Alamos National Laboratory
22--22LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Course Content
1. Introduction to Free-Electron Lasers2. Basics of Relativistic Dynamics3. One-dimensional Theory of FEL4. Optical Architectures5. Wigglers6. RF Accelerators7. Electron Injectors
Chapter
ChapterChapter
ChapterChapter
Chapter
Chapter
Day 1
Day 2
Day 3
Day 4
22--33LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Chapter 4Optical Architectures
22--44LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Architectures
• Oscillator FEL• Regenerative Amplifier• Self-Amplified Spontaneous Emission• Seeded Amplifier• Optical Klystron• High-gain Harmonic Generation
22--55LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Oscillator FEL uses a combination of low-current electron beam and an optical cavity to trap the optical pulse in several passes through the wiggler. Each pass involves a new electron bunch interacting with the optical field and undergoing microbunching at the end of the wiggler. The optical intensity grows by a small amount (low gain) in each pass with an amplification factor of 1 + gss. The small-signal gain, gss, is proportional to the cube of Nw (number of wiggler periods). At high intensity, the electrons rotate in phase space and absorb light near the end of the wiggler, and the large-signal gain is reduced compared to gss. Saturation occurs when gain is equal to total cavity losses. Beyond saturation, the electrons’ synchrotron oscillations lead to sideband generation. Oscillator FEL’s extraction efficiency is approximately one divided by 2.4Nw. The most common optical cavity is the near-concentric resonator. At low gain, the optical mode is determined almost entirely by the resonator optics. The oscillator FEL cavity length has to be accurate to within a few optical wavelengths.
Oscillator FEL
22--66LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Basic Configuration of an Oscillator FEL
FEL power builds up from noise to saturation in N passes inside an optical cavity. In the Nth pass, a fresh bunch of randomly distributed electrons interacts with the optical beam and develops microbunching with period = .
Artwork of N. ThompsonIntroduction to FEL
22--77LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Small-Signal Gain
[ ]23
3
2 w wwss
b A
JJ aN Igr I
lpg
æ ö æ ö÷ ÷ç ç÷ ÷= ç ç÷ ÷ç ç ÷÷ çç è øè ø
Nw number of wiggler periods electron beam’s Lorentz factor (~ 2 x energy in MeV)[JJ] reduction factor of FEL interaction in a planar wiggleraw dimensionless wiggler parameterw wiggler periodrb electron beam radius in the wigglerI electron beam peak currentIA Alfven current; 17 kANote: Amplification factor (Pout/Pin) = 1 + gss
Small-signal gain must be higher than total round-trip losses.
22--88LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Outcoupling
• Out-coupler’s reflectivity is typically 80-99% (outcoupling ~ 1-20%).
• For maximum output power, absorption loss must be kept very low.
• The intracavity power is (1 - Routcoupler)-1 times the output power.
1 1out outcoupler absorption cavityP R P
cavityP outP
outcoupler
out out cavityP P
Wiggler
22--99LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Power Growth Curves
As the optical intensity increases, the large-signal gain decreases until FEL gain is equal to the total cavity losses and FEL power saturates.
gss = 39% total round-trip losses = 4%At pass #150 near the wiggler exit, the electrons over-bunch and absorb light as they move up the bucket. Large-signal gain is smaller than small-signal gain.
At pass #40, intracavity power increases with z3.
Pout = Pin (1 + gss)
Power and gain vs pass number Power vs z Courtesy of N. ThompsonIntroduction to FEL
22--1010LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Small-Signal Gain (revisited)
[ ]23
2 wwss
w A
JJ aN Igk I
pg s
æ ö æ öæ ö ÷ ÷÷ ç çç ÷ ÷= ÷ ç çç ÷ ÷÷ ç çç ÷ ÷÷ ççè ø è øè ø
The small-signal gain for a planar wiggler at the peak of the gain curve, assuming the electron beam radius is smaller than the optical beam, is
[ ]2 13 31
2w
w A
JJ a Ik I
rg s
æ ö æ ö÷ ÷ç ç÷ ÷= ç ç÷ ÷ç ç ÷÷ çç è øè ø
Recall the dimensionless FEL (Pierce) parameter
The small-signal gain is proportional to the cube of the Pierce parameter
( )32 2ss wg Np r=
22--1111LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Maximum Gain & Wiggler Efficiency
10 7.5 5 2.5 0 2.5 5 7.5 100.5
0
0.5
g ( )
Maximum small-signal gain at = 2.6
gss()
4 2w wEN N
E
max 2.6
maxmax
2(2.6)24 w
EE N
Saturation occurs when average electron energy evolves from = 2.6 (maximum gain) to = -2.6 (maximum absorption).
12.4
FEL
e beam w
PP N
Wiggler efficiency of oscillator FEL
For maximum gain, one must select an energy detuning (electron beam energy minus FEL resonance energy) to be slightly positive. Conversely, at a fixed beam energy, the lasing wavelength is longer than the resonance wavelength.
Max. absorption at = -2.6
22--1212LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Visualization of Wiggler Efficiency
Wiggler entrance Wiggler exit
12.4 wN
Wiggler extraction efficiency at saturation
15 wN
12 wN
12.4 wN
Note: A common mistake is to calculate the intra-cavity power from the wiggler extraction efficiency and electron beam power. In reality, the intra-cavity power is Q times the output power, where Q is the opticalcavity quality factor (the inverse of 1 – reflectivity).
Oscillator FEL picks the wavelength for maximum gain (R = 2.4/Nw)
22--1313LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Maximum Output Efficiency & Power
12.4
outFEL
out loss wN
Output efficiency depends on wiggler efficiency, out-coupling and total losses
max 2.4out beam
out loss w
PPN
Maximum output power is electron beam power times output efficiency
12.4 2.4
beam beamcavity cavity
out loss w w
P PP QN N
Maximum intracavity power
22--1414LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Oscillator with Energy RecoveryExample: Jefferson Lab FEL
Superconducting RF Linac
WigglerBuncher
Dump
High Reflector Outcoupler
InjectorBooster
Oscillator FEL’s wall-plug efficiency can be significantly increased with energy recovery. The spent electron beams are circulated through the superconducting RF linac at the deceleration phase and dumped at lower energy to improve overall efficiency and reduce radiation hazards.
22--1515LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Energy RecoveryOscillator FEL efficiency is typically 1%. The spent electron beams still have ~99% of the initial energy. Dumping the high-power electron beam is wasteful and creates radiation hazards (Edump is beam energy before the beam dump).
2.4b
electron FELw dump
EN E
With energy recovery, the efficiency of electron-to-FEL conversion is enhanced by the ratio of beam energy at the wiggler to beam dump energy.
Without energy recovery With energy recovery
Eb Edump
Accelerator
Edump
Decelerator
Eb
~½N
Beam Dump Beam Dump
Higher energy
Accelerator
~½N
22--1616LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Resonator
• The optical resonator consists of two concave mirrors with radii of curvature R1and R2. The mirrors are separated by a distance D.
• A measure of stability is the overlap between the two radii of curvature.• The Rayleigh length is a measure of the FEL beam’s divergence. The shorter
the Rayleigh length, the larger the divergence.
Wiggler
OutcouplerTotal reflector
22--1717LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Resonator Designs
Common FEL resonators are near-concentric design, D ~ 2R.The lowest order optical mode of the near-concentric design has small mode area in the wiggler (high optical field) and large mode area at the mirrors, thanks to large divergence angle. The strong divergence reduces FEL intensity on the mirrors and minimizes risk of optical damage.
Concentric
Confocal
D = 2R
R
D = R
22--1818LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Stability, Rayleigh Lengthand Optical Beam Mode Area
Rayleigh length of a near-concentric cavity is R times square root of (1 + g)/2.
11R
gz Rg
Rayleigh length
01
4 1gDg
Gaussian mode area at waist
g1
g2
Confocal
Planar
-1
Concentric
Unstable
Unstable
Stable
Stable
-1
1
1
FEL
Fox-Li Stability Diagram
1ii
DgR
1 2 1g g Stability criterion
Symmetric resonator
1 2g g g
Note: g is resonator geometric factor, not FEL gain.
22--1919LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Beam Mode Areas
01~
2gR
0
21
R
g
Wiggler center At the mirror
Optical mode diameters at the wiggler entrance and exit have to be smaller than the wiggler or wiggler tube apertures.
2
0
12
w w
R
Lz
Confocal resonatorOptical beam >> electron beam
Concentric resonatorOptical beam > electron beam
Short Rayleigh lengthOptical beam ~ electron beam
At the entrance and exit of the wiggler
22--2020LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Angular Sensitivity ofNear-concentric Resonators
1 34 4
1 1g gD
Angular sensitivity for R1 = R2 (g1 = g2)
342 x
D
For near concentric resonators,g = –1 + x where x << 1
Example: For an FEL with near-concentric resonator with g = 0.98 (x = 0.02), with = 1 and D = 20 m, the mirror angular sensitivity is 11 rad.
x xR
22--2121LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Pulse & Cavity Length
• Cavity length must be such that the optical pulses overlap with the electron bunches in N passes in order to reach saturation (N could be in the hundreds or in the thousands).
• Mirror separation must be equal to the spacing between electron bunches to an accuracy of the FEL detuning length (about 2 - 3 wavelengths).
Nth pass
1st passCavity length too long Cavity length too short
22--2222LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Slippage & LethargyWiggler entrance
• In the small-signal regime, the trailing edge is amplified more than the leading edge. The group velocity of the optical pulse is less than c. Gain is maximized at slightly shorter cavity length.
• Near saturation, gain is reduced and the optical pulse slips ahead by Nw(the slippage length). Power is maximized at a cavity length close to the electron bunch spacing.
g c
g c
Small-signal
Near saturation
Wiggler exit
Slippage length s wl N
22--2323LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Cavity Length Detuning
Saturation power (black line) and FEL net gain (red dots) versus cavity length detuning for the ELBE oscillator FEL experiment at 20 . The cavity length must be initially shortened to maximize gain. After oscillation is established, the cavity is lengthened to maximize output power. When the cavity length exceeds the electron bunch spacing (c divided by f), FEL power drops to zero.
22--2424LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Sideband Generation
FEL efficiency can be increased by a factor of 2-4X with sideband generation.
Efficiency without sidebands
Efficiency with sidebands
AB
C
D
c (cm) Wavelength ()
Sidebands are generated at zero cavity detuning after the FEL saturates. The FEL pulses develop ultrashort spikes and its spectrum shows additional peaks (sidebands)
22--2525LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Phasor Diagram of Sidebands
t
t
t
0
At saturation, synchrotron oscillation causes the optical intensity to be modulated in time. This modulation in turn causes the electron motions in phase space to be modulated. The FEL micropulses break up into individual spikes whose width is on the order of a slippage length. Sidebands are spectrally separated from the central frequency by c divided slippage length. The low-frequency sideband is more intense than the high-frequency one.
Phasor diagram
Real axis
Imaginary axis
Frequency
s
cl
22--2626LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Regenerative Amplifier FEL
Regenerative Amplifier FEL (RAFEL) uses a combination of high-current electron beam and a low-Q optical resonator to reach saturation in a few passes through the wiggler. Each pass involves a new electron bunch interacting with the optical field and bunching near the end of the wiggler. The optical intensity grows exponentially (high gain) in each pass until it reaches the saturation intensity. Exponential gain depends on emittance, energy spread and peak current. At saturation FEL power oscillates between high (optimum feedback) and low (over-bunched).
RAFEL optical feedback varies depending on optical guiding in the wiggler. The FEL optical mode is determined mainly by the electron beam. The resonator cavity length detuning is a fraction of the electron bunch length. Maximum power occurs on both sides of zero detuning length.
22--2727LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Regenerative Amplifier Feedback
Output Coupler Reflector
Paraboloid Reflector
Wiggler
Paraboloid Reflector
Electron BeamSpent Beam
Schematic of the Regenerative Amplifier FEL
1 exp9
wss
G
LGL
Gain per pass
net FB ssG R G
Net gain
3FB OC M mR R R f
Optical feedback
Optical feedback is less than the reflectivity of mirrors because only a small fraction, fM of the returning optical beam matches with the guided mode thatcan be amplified in the next pass.
22--2828LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
RAFEL Power Growth per Pass
n tRFRF wave train
FEL micropulses
Small-signalregime
1 0 0 ew
g
LL
n n n w n nP R P L R P
22--2929LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Feedback Cavity
Ring cavity length is set to be synchronous with electron bunch arrival.Optical mode is guided by the electron beam in the wiggler. Without optical guiding, the optical mode is not reproduced from pass to pass.
Wiggler Outcoupler
Spherical mirror
Cylindrical mirror
Injection mirror
Spherical mirror
Cylindrical mirror
22--3030LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Guiding in FEL
Electron beam
Naturally diffracted1/e2 radius Refractively guided
1/e2 radiusOptically guided1/e2 radius
Refractive guiding arises from phase velocity reduction of a Gaussian beam near the center of the electron beam. Refractive guiding depends on the real part of the complex refractive index.
Gain guiding is due to selective amplification of the central part of the optical beam that overlaps with the high-current electron beam. Gain guiding depends on the imaginary part of the complex refractive index.
The combination of diffraction and gain guiding leads to mode selection.
22--3131LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Modes
Laguerre-Gaussian (LG) Modes Hermite-Gaussian (HG) Modes
Cylindrical ( ) Coordinates Cartesian (x y) Coordinates
The lowest order mode (LG00 or HG00) has the strongest on-axis intensity to interact with the electron beam, and thus sees the highest FEL gain.
22--3232LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Electron Beam’s Refractive IndexElectron beam’s complex refractive index
-Imaginary term = gain
Real term = refractive focusing
Electrons tend to bunch around slightly positive of 0, thus the sin (gain) term is positive and the 1 + cos (refractive index) term is slightly larger than unity. The FEL optical beam sees a refractive index difference from vacuum, n – 1, that is proportional to electron beam peak current. This refractive index is higher at the center of the electron beam and lower at the edge, similar to an optical fiber.
( )
( )
1 sinIm
1 cosRe 1
s
A
s
A
da Ink dz I
d Ink dz I
qg
f qg
æ ö÷ç ÷- = µç ÷ç ÷çè øæ ö÷ç ÷- = µç ÷ç ÷çè ø
0
200
400
600
800
1 103
I r( )
r
radius (mm)
curre
nt (A
)
n - 1
0 0.1 0.2 0.3-0.3 -0.2 -0.1
22--3333LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Beam Guided Mode Radius2 2
2 2 3 20
3 G b
d r rdz r L
Natural diffraction Focusing from gain/refraction
1/423 G bg
Lr l sp
æ ö÷ç ÷=ç ÷ç ÷÷çè ø
Guided mode radius
0 .0 2
0 .0 4
0 .0 6
0 .0 8
0 .1
0 .1 2
0 .1 4
0 2 4 6 8 1 0
P lo t o f 1 /e 2 r a d iu s v e r s u s d is ta n c e
1 /e 2
R a d iu s
W ig g le r L e n g th (m )
Diffraction > focusing
Diffraction < focusing( )
1/42 22
2 2
2 1w bg ww al s
g p r
æ ö÷ç ÷= +ç ÷ç ÷çè ø
Guided 1/e2 mode radius in 1D
Plot of 1/e2 radius vs. distance in wiggler
1/e2
radius
wiggler length (m)
wg
22--3434LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical feedback is large at small signals to provide high net gain
Optical feedback is reduced at high intensity and most of FEL power exits through the large hole on the annular mirror.
Variable Feedback in RAFEL
Outcoupling Annular Mirror
Reflective surface
Outcoupling hole
Feedback at low power >75% Feedback at high power = 3%
22--3535LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Output Power, Gain vs Input Power
1.E+06
1.E+07
1.E+08
1.E+09
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Pin (W)
Pou
t (W
)
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Pin (W)G
ain
Maximum power and maximum gain occur at different input power. Maximum gain occurs at low Pin. Maximum power occurs at intermediate Pin. Output power oscillates between high and low because Pin varies from pass to pass.
22--3636LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Self-Amplified Spontaneous Emission SASE FEL uses a combination of high-current electron beam and a long wiggler (undulator) to reach saturation in a single pass. SASE starts up from the random positions of electrons in the bunch, i.e. noise. The initial interaction is linear, that is, electrons outside the separatrix interacting with the optical field become bunched and amplify the optical field. SASE power grows exponentially (exponential gain) along the wiggler with a characteristic e-folding length called the power gain length. At higher intensity, the electrons become trapped in the growing bucket and the FEL interaction becomes non-linear. Saturation occurs when the over-bunched electrons at the bottom at the bucket rotate upward. SASE extraction efficiency is approximately the same as the gain coefficient . SASE is the baseline design for many x-ray FEL around the world, since no materials exist with sufficient reflectivity as x-ray mirrors. SASE FEL is fully coherent transversely and only partially coherent longitudinally. The spectral bandwidth is narrow in the exponential gain region but broadens at saturation. Shot-to-shot energy fluctuations are large in the exponential gain region.
22--3737LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Linac Coherent Light Source
InjectorInjectorat 2at 2--km pointkm point
1 km S1 km S--band band linaclinac
ee Transport (340 m)Transport (340 m)
UndulatorUndulatorExperiment HallExperiment Hall
22--3838LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
LCLS Accelerator Layout
SLAC linac tunnelSLAC linac tunnel Courtesy of P. EmmaCourtesy of P. Emma
LinacLinac--00L L =6 m=6 m
LinacLinac--11L L 9 m9 mrf rf 2525°°
LinacLinac--22L L 330 m330 mrf rf 4141°°
LinacLinac--33L L 550 m550 mrf rf 00°°
BC1BC1L L 6 m6 m
RR5656 39 mm39 mm
BC2BC2L L 22 m22 m
RR5656 25 mm25 mm DL2 DL2 L L =275 m=275 mRR56 56 0 0
DL1DL1L L 12 m12 mRR56 56 0 0
undulatorundulatorL L =130 m=130 m
6 MeV6 MeVz z 0.83 mm0.83 mm 0.05 %0.05 %
135 MeV135 MeVz z 0.83 mm0.83 mm 0.10 %0.10 %
250 MeV250 MeVz z 0.19 mm0.19 mm 1.6 %1.6 %
4.30 GeV4.30 GeVz z 0.022 mm0.022 mm 0.71 %0.71 %
13.6 GeV13.6 GeVz z 0.022 mm0.022 mm 0.01 %0.01 %
LinacLinac--XXL L =0.6 m=0.6 mrfrf= =
21-1b,c,d
...existinglinac
L0-a,b
rfrfgungun
21-3b24-6dX 25-1a
30-8c
undulatorundulator
beam parked here
beam parked here
22--3939LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
LCLS Undulator Hall
Courtesy of P. EmmaCourtesy of P. Emma
22--4040LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
LCLS Power Growth
= 1.5 = 1.5 ÅÅE = 14 E = 14 GeVGeVx,yx,y = 0.4 = 0.4 m (slice)m (slice)IIpkpk = 3.0 kA= 3.0 kAEE//EE = 0.01% (slice)= 0.01% (slice)
LLGG 3.3 m3.3 m
Courtesy of P. EmmaCourtesy of P. Emma
22--4141LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Start-up power without a seed laser depends on the number of correlated electrons, characterized by the initial bunching coefficient, and the resulting equivalent start-up power.
For SASE, the initial bunching is due to shot noise so the square of b1is simply 1/Ne and the SASE start-up is spontaneous emission. For pre-bunched electrons, the start-up power is
Start-up Power
2
G
zL
seed startup
dP dP dP e ed d d
w
ws
w w w
æ öD ÷ç ÷ç- ÷ç ÷÷çè øæ ö÷ç ÷ç= + ÷ç ÷÷çè ø
Spectral power grows exponentially with z from start-up power
01
1
1 ek
Ni t
ke
b eN
210.22 b
prebunchIEP be
2 21startup e eP b N P
22--4242LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Exponential Gain Length
34w
GL 1D Power gain length
1D Field gain length
Requirements for SASE
Normalized emittance
Energy spread
Diffraction
4n
G RL Z
2 3w
gL lp r
=
Wiggler length over which power grows by a factor of 2.718.
Wiggler length over which field grows by a factor of 2.718 (twice as long as power gain length).
Gain length will increase if any of the above requirements is not met.
22--4343LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Three-Dimensional Effects
• Diffraction
• Emittance
• Energy spread
First calculate 1D gain length, then use M. Xie’sparameterization to calculate 3D gain length, longer than 1D gain length due to 3D effects.
( )3 1 1D DL L= +L
1 4Dn
b
L
1 4D
w
L
1Dd
R
LZ
0.57 1.6 2 2.9 2.4 0.95 3 0.99 1.1
0.76 2.3 2.7 2.1 2.9 2.8 0.43
0.45 0.55 3 0.35 51 0.62
5.3 120 3.7d d d
d d d
European XFEL gain length vs. emittance
22--4444LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Gain Spectrum of High-Gain SASE
Unlike the low-gain FEL which has zero gain at resonance, high-gain SASE FEL has maximum gain at zero detuning (resonance energy).
(a) (b)
22--4545LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
SASE Animation
22--4646LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Rate of Energy Loss and Increase
• Rate of energy loss due to FEL interaction
• Increase in energy spread due to FEL interaction
• Increase in energy spread due to spontaneous emission recoil, also known as energy diffusion
Increase in energy spread reduces FEL gain and causes saturation. At Eb > 35 GeV, energy diffusion makes SASE FEL inoperable.
2
32e C w w
d r a kdz
2 223
b e wdE r Bdz
2
20
23
e w
c
r Bddz m c
22--4747LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Transverse Coherence
Incoherent Spontaneous Emission SASE in exponential region SASE at saturationCourtesy of I. Zagorodnov
22--4848LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Saturation reduces optical guiding
The guided mode has constant FWHM. However, the rms radius increases after saturation because more power is added to the wings after the center saturates. Saturation increases the higher order mode content of the SASE optical beam.
Courtesy of Zhirong Huang
22--4949LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Limit-Cycle Oscillations
Evolution of electron (blue dash) and FEL (red solid) profiles at FLASH along the wiggler length. The initial profiles are smooth but as the optical pulse slips ahead of the electron bunch, the temporal profile breaks up into sub-pulses.
Exponential Region
SaturationEnd of exponential region
Beginning of exponential region
Courtesy of M. Dohlus
22--5050LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Power Growth and Fluctuations
1II MD
= b
c
Mls
=Bunch length
Coherence length
Shot-to-shot fluctuations
Shot-to-shot FluctuationsPeak Power (W)
22--5151LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Temporal & Spectral Profiles
Temporal profileSpectral profile
Spectral profile
SASE temporal profile consists of several correlation lengths. The corresponding Fourier transform consists of many spectral features. The spectral FWHM is inversely proportional to the spikes’ temporal width while the narrow spectral features are the inverse of the full temporal pulse.
22--5252LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Longitudinal Coherence
SASE in exponential growth has narrower linewidth than spontaneous noise. At saturation, SASE bandwidth increases to , about the same as 1/Nw. The temporal profile develops “spikes” with short coherence length (large ).
Incoherent Spontaneous Emission SASE near Saturation
(nm)
1
wNllD
=l
rlD
=
(nm)
Courtesy of Phil Sprangle
22--5353LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Improving Longitudinal Coherence
SASE coherence can be improved by dividing the wiggler into two sections. The first wiggler produces partially coherent FEL light that is made more coherent (monochromatic) with a diffraction grating. The monochromatic light is injected into the second wiggler where it seeds the amplification to generate a fully coherent x-ray beam.
22--5454LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Harmonic Power Growth in SASE
The fundamental optical power causes electrons to bunch. As electrons develop microbunching, the 3rd and 5th harmonics experience higher gain, their power grows more quickly and saturates earlier than the fundamental.
22--5555LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Seeded Amplifier
Unlike SASE which starts from spontaneous noise, seeded amplifiers use either a coherent optical input (seed laser) or an information-encoded electron beam (pre-bunched beam radiation) to start the amplification process. The seed laser power is initially reduced by a factor of 9. The seed power only has to exceed the spontaneous noise emitted into the coherent bandwidth and angle. Seeding improves the FEL longitudinal coherence and reduce the shot-to-shot energy fluctuations. Slippage in a long wiggler must be considered, as it can reduce the optical pulse energy. In a high-gain FEL, the FEL pulse travels slower than c and slippage effects are reduced compared to the low-gain cases.
22--5656LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Seeded Amplifier Power Growth
GLzexpP)z(P 09
1
Power grows exponentially with z
The factor of 1/9 arises because only 1/3 of optical field contributes to the exponentially growing mode.
Seed power must be 9 times the spontaneous power emitted into coherent angle and bandwidth.
2
2
1137 1
w bnoise
w
a h IPa e
næ ö÷ç ÷= ç ÷ç ÷ç +è ø
Exponential growth
Seeded amplifier peak power versus length
Lethargy
Exponential growth
Saturation
108
106
104
102
10 2 4 86 10
Wiggler length (m)
Pow
er (W
)
22--5757LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Seeded amplifiers have higherlongitudinal coherence than SASE
Longitudinal coherence and fluctuations can be improved by injecting a coherent (long pulse, narrow linewidth) seed to start up the amplification.
Seeded Amplifier SASE
TemporalProfile
Spectrum
Courtesy of Henry Freund
22--5858LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Slippage in a Seeded Amplifier
Wiggler entrance
Slippage in a seeded amplifier varies depending on gain. In the small-signal high gain regime, slippage is reduced due to “lethargy” effect. At saturation, lethargy vanishes and the optical pulse slips over the electron bunch one wavelength every wiggler period.The amplified FEL pulse can undergo limit-cycle oscillations if electron bunch length is much less than the slippage length (N). To avoid limit-cycle oscillations that could reduce optical pulse energy, both the electron bunch and optical pulse length must be longer than the slippage length.
g c High gain
Wiggler exit
22--5959LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Optical Klystron and HGHG
Optical klystron FEL uses the same process of energy modulation followed by density modulation in a klystron except the bunching happens at optical wavelengths instead of microwave. Optical klystrons are used mainly to shorten the wiggler length for the same gain (or increase gain for the same length). The optical klystron consists of two wiggler sections separated by a dispersive section. The electrons interact with a seed laser in the first wiggler, undergo synchrotron motion in longitudinal phase space and develop periodic density modulation in energy-phase. The modulation period is the seed laser wavelength. This energy-phase modulation transform into density-phase modulation in the dispersive section, which speeds up the high energy electrons and slows down the low energy ones.
The microbunched electron beams are made to radiate in a second wiggler, either at the fundamental wavelength (conventional OK FEL) or at the harmonics, a process known as High-Gain Harmonic Generation (HGHG).
22--6060LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Principle of an Optical Klystron
Modulatore-beam
Seed laser
modulatede-beam
Radiator
Energy modulation
Dispersive Section
Density modulation
d/d
z
z = 1.44 m
d/d
z
z = 2.28 m
22--6161LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Magnetic Field and ElectronTrajectory in an Optical Klystron
Magnetic Field in Vertical Direction Single Electron Motion in Horizontal Direction
Lower Energy
Higher Energy
Dispersive SectionModulator Radiator
22--6262LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Energy Modulation & Bunching
221
s w
R w
a aa
2
Ls
eEakmc
Energy modulation at optical field as
Dimensionless field as
02L LE Z ILaser electric field
R
(rad) (rad)
22--6363LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
High-gain Harmonic GenerationHGHG is a way to generate short wavelength radiation using a lowenergy electron beam and a long-wavelength seed. HGHG requires high quality electron beams (ones with low emittance and energy spread).
2112 1
2w
s a
22 1
2wh
FEL ha
22--6464LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Harmonic Microbunching
Energy vs. phase
Density vs. phase
Energy modulation Density modulation
212
56
n
n nb J nkR e
Property of dispersive sectionEnergy modulation
Bessel function of nth orderGaussian spread in energy
Bunching coefficient at the nth harmonic
22--6565LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
Harmonic Bunching vs Seed Power
Courtesy of N. ThompsonSimulating HGHG with Genesis
22--6666LA-UR 09-02731US Particle Accelerator School 2009US Particle Accelerator School 2009
University of New Mexico University of New Mexico -- Albuquerque NMAlbuquerque NM
References
• “Review of X-ray Free-Electron Laser Theory,” Z. Huang and K.J. Kim, Phys. Rev. ST-AB, 10, 034801 (2007)
• “Review of Existing Soft and Hard X-ray FEL Projects,” A. Renieri and G. Dattoli, JACoW Proceedings of the 27th FEL Conference.
Publications
URL
• “Refinement of FEL 1-D Theory” by M. Dohlus, J. Dohlus and P. Schmüserhttp://www.springerlink.com/content/3016v01520372111/fulltext.pdf
• “Simulating HGHG with Genesis 1.3 – An Unofficial Guide” N. Thomsonhttp://www.eurofel.org/sites/site_eurofel/content/e160/e167/e190/e193/infoboxContent200/N.Thompson-SimulatingHGHGwithGenesis.pdf