for the
XUV seeding experiment
at FLASH
Dissertation
zur Erlangung des Doktorgrades
des Fa hberei hs Physik
der Universität Hamburg
vorgelegtvon
Manuel Mittenzwey
Hamburg
Prof. Dr. FlorianGrüner
Guta hter der Disputation : Prof. Dr. Markus Dres her
Dr. Jens Osterho
The Free ele tron LASer in Hamburg (FLASH), operating inthe Self-Amplied
Sponta-neous Emission (SASE) mode, is urrently the most intense femtose ond light sour e in
the eXtreme-UltraViolet (XUV) regime. However, the statisti al nature of SASE leads
to intensity u tuations of the Free Ele tron Laser (FEL) pulses. Moreover, the
ele -tron a elerationpro essintrodu esarrival-timeu tuationsofthe ele tron bun hatthe
undulator entran e, whi h leads to a temporal jitter with respe t to the
syn hroniza-tion system. This jitter limitsthe resolution of orresponding pump-probeexperiments.
In order to redu e these pulse-to-pulse u tuations, a seeding se tion for the ele tron
bun h has been installed in 2009/2010. Here, XUV seed pulses from a High-Harmoni
Generation (HHG) sour e are overlapped in spa e and time with the ele tron bun hes
in new variable-gap undulators installed upstream the existing SASE undulator. Behind
the undulator se tion, the seeded FEL radiation will be dee ted by a mirror into an
experimental ontainerfor beam hara terization. Afra tion ofthe opti allaser, used as
the drivinglaser forHHG,isalsotransported intotheexperimental ontainer,thus,
tem-poraldiagnosti slike ross- orrelationsbetween thesyn hronized seededFEL andopti al
pulses arepossible. Emphasisof this workisthe onstru tionand hara terization ofthe
HHG sour e.
Referat
Der Freie-Elektronen LASer in Hamburg (FLASH) arbeit zur Zeit in einem Modus, der
die spontane Undulatorstrahlung selbst verstärkt (SASE). Die resultierende Strahlung
liegt im extremultra-violetten (XUV) Spektralberei h und istdie stärkste Li htquelle in
diesemWellenlängenberei hüberhaupt. DasVerstärkungsprinzipSASEisteineMethode,
die das Anfangsraus henverstärktund dadur hintrinsis hstatistis henFluktuationenin
der Intensität und im Spektrum unterliegt. Zusätzli h werden dur h den
Bes hleuni-gungsprozess der Elektronen Ankunftzeituktuationen relativ zu einem
Syn hronisation-slaseramUndulatoreingangeingeführt,diesi hdur hAnkunftzeituktuationender
resul-tierendenLi htpulsebemerkbarma hen. DieseS husszuS hussFluktuationenreduzieren
die zeitli he Auösung von so genannten Pump-Probe Experimenten. Um diese
Fluktu-ationen zu reduzieren, wurde 2009/2010 eine seeding-Sektion in FLASH (sFLASH)
im-plemeniert. Hierbeiwerden Li htpulse einerHohe-Harmonis hen Quelle(HHG)in Raum
und Zeit mit den Elektronenpaketen überlagert und ans hlieÿend in den sFLASH
Un-dulatoren verstärkt. Na h der Undulatorstre ke wird die verstärkte Strahlung aus dem
FLASH-Tunnel in einen Experimentier ontainer reektiert. Zusätzli h wird ein Teil des
für die HHG Quelle benötigten Fundamentallasers direkt in den Experimentier ontainer
geführt, so dass zeitaufgelöste Kreuzkorrelationsexperimentemitdem Fundamental-und
der verstärkten XUV- Laserstrahlung dur hgeführt werden können. S hwerpunkt dieser
DieseArbeitistvonmirselbstständigverfasst wordenund i hhabekeineanderen alsdie
angegebenen Quellen und Hilfsmittelverwendet.
1 Introdu tion 1
1.1 Highharmoni generation . . . 1
1.2 Self-amplied spontaneous emission . . . 2
1.3 HHG vs. SASE . . . 3
1.4 Seeded FELs . . . 3
1.5 Thesis stru ture . . . 4
2 Fundamentals 5 2.1 Des riptionof laser pulses . . . 5
2.1.1 Geometri al opti s . . . 5
2.1.2 Gauss opti s . . . 6
2.2 XUV opti s . . . 8
2.3 HighHarmoni Generation . . . 9
2.3.1 Three-step model . . . 9
2.3.2 Quantum me hani al des ription . . . 12
2.3.3 Phase mat hing . . . 14
2.3.4 Properties of the harmoni s . . . 16
2.4 Free Ele tron Laser . . . 17
2.4.1 Ele tron sour eand a eleration . . . 18
2.4.2 Undulator radiation. . . 18
2.4.3 High-gain free-ele tron laser . . . 20
2.4.4 FEL operationmodes. . . 21
3 Experimental environment 25 3.1 seeded Free ele tron LASer inHamburg (sFLASH) . . . 25
3.1.1 FLASH fa ility . . . 26
3.1.2 sFLASH . . . 28
3.2 HighHarmoni Generation . . . 32
3.2.1 The laser system . . . 32
3.2.2 High harmoni generation sour e . . . 33
3.3 Diagnosti s . . . 39
3.3.1 Chara terization of the driving laser . . . 39
3.3.2 XUV opti s . . . 39
3.3.3 Separation of XUV radiation. . . 42
3.3.4 Chara terization of XUV radiation . . . 43
4.1 Fundamentallaser . . . 51
4.1.1 Laser pulse duration . . . 51
4.1.2 Laser beam proleand intensity . . . 51
4.1.3 Polarization . . . 55
4.1.4 Wavefront . . . 56
4.2 Calibration of the spe trometer . . . 56
4.2.1 Online alibration . . . 56
4.2.2 Quasi online alibration . . . 56
4.2.3 Absolute alibration . . . 57
4.3 HighHarmoni Generation . . . 59
4.3.1 XUV beam propagation . . . 60
4.3.2 Harmoni energy and onversion e ien y . . . 64
4.3.3 Harmoni spe trum- plateauand uto . . . 75
4.3.4 Wavefront . . . 79
4.3.5 Stability . . . 79
4.3.6 Quasi phase mat hing . . . 79
4.3.7 Two olor high harmoni generation . . . 81
4.4 Seeding FLASH . . . 83 5 Con lusion 85 5.1 Outlook . . . 88 5.1.1 Seed power . . . 88 5.1.2 Inje tion beamline . . . 89 5.1.3 Alternativeseeding . . . 90 A Abbreviations I
B Propagation of un ertainties III
C Te hni al details V
C.1 Software environment . . . V
C.2 Hardware . . . VII
D Code of pra ti e XI
D.1 The laser system . . . XI
D.2 Highharmoni generation sour e . . . XII
1.1 Peak brillian eof dierentlight sour es . . . 2
2.1 Cal ulated beam width . . . 7
2.2 Simple man'smodel. . . 10
2.3 Typi alHHG spe trum . . . 16
2.4 S hemati representation of anundulator . . . 19
2.5 GENISIS simulationof the mi robun hing pro ess . . . 21
2.6 Exponentialgrowth of the FEL pulse energy . . . 22
3.1 Free Ele tron LASer Hamburg at DESY . . . 25
3.2 S hemati representation of FLASH . . . 26
3.3 Overview of sFLASH . . . 28
3.4 Cross se tion of the inje tion beamline . . . 30
3.5 S hemati sket h of the laser system . . . 32
3.6 CAD model of the pit . . . 34
3.7 CAD model of the HHGsour e . . . 35
3.8 CAD model of the HHGtarget hamber . . . 36
3.9 AlternativeCAD model ofthe HHG target hamber. . . 38
3.10 Ree tivity of B
4
C for p-and s- polarized light. . . 403.11 CAD model of the tripletmirror system . . . 41
3.12 Ree tivity of S Si multilayer . . . 42
3.13 CAD model of the omplete HHG setup . . . 43
3.14 Transmission of the used metal lters . . . 44
3.15 CAD model of the diagnosti bran h. . . 46
3.16 Simulationof dierentspe trometer ongurations . . . 47
3.17 Simulationof dierentslits infrontof the spe trometer . . . 48
4.1 Measured spe trumand pulse length of the driving laser. . . 52
4.2 Measured fareld beam proleof the driving laser . . . 52
4.3 Beamsize and energy of the driving laser . . . 54
4.4 Absolute alibrationof the
21
st
harmoni . . . 574.5 Spe tra for two dierent spe tral positions of the harmoni omb. . . 58
4.6 Calibrated spe trum . . . 59
4.7 H
21
near eld beam prole (f
NIR
= 1500
mm). . . 604.8 H
21
near eld beam prole (f
NIR
= 3000
mm). . . 614.9 Fo us s an of H
21
. . . 624.10 H
21
yielddependent on the position of the gas target.. . . 654.12 H
21
yielddependent on the driving laser intensity . . . 674.13 Yieldof dierent targetgeometries . . . 68
4.14 Measured targetdensity of the gas valve . . . 69
4.15 Imaged damageof the rst tripletmirror.. . . 70
4.16 HHG signal dete ted with the HHmeter. . . 73
4.17 Correlation between argon and helium for the HHmeter signal . . . 75
4.18 Singleshot HHG spe trum . . . 76
4.19 HHG spe tra generated with a basi gas nozzle. . . 77
4.20 Spe tra for dierent pulse lengthsof the driving laser . . . 77
4.21 Optimized HHG spe trumfor
f
NIR
= 1500
mmandf
NIR
= 3000
mm . . . . 784.22 S hemati representation of QPM. . . 80
4.23 Ee t of hydrogenon quasiphase mat hed HHG. . . 81
4.24 S hemati illustrationof two olor HHG . . . 81
4.25 Two olor HHG spe tra. . . 82
4.26 HHG tunnel spe trum . . . 83
C.1 Javadoo sdata display . . . VI
2.1 Harmoni orders . . . 9
3.1 Performan e of FLASH . . . 27
3.2 Partial ross se tion of argon. . . 45
4.1 Divergen e of H
21
. . . 634.2 Energy of H
21
. . . 765.1 Published spe i ations of HHG in argon . . . 86
A.1 Abbreviations . . . I
A.2 Physi al onstants . . . II
C.1 Te hni al data Basler slA1390-17fm . . . VII
C.2 Te hni al data Beta BariumbOrate . . . VIII
C.3 Te hni al data multi- hannelplate . . . VIII
C.4 Te hni al data GrenouilleFROG . . . VIII
C.5 Te hni al data HHmeter . . . VIII
C.6 Te hni al data XUV CCD . . . IX
C.7 Te hni al data wavefront sensor . . . IX
C.8 Te hni al data XUV dira tion grating . . . X
The free-ele tron laser in Hamburg (FLASH) is a user fa ility providing highly brilliant
ultrashort eXtreme UltraViolet (XUV) radiation to users. The established operating
mode SASE (Self-Amplied Spontaneous Emission) shows shot-to-shot u tuations in
the spe trum, the pulse prole, and the arrival time of the radiation. In ontrast, high
harmoni radiation is omparably stable, but with pulse energies orders of magnitude
lower than SASE. Seeding FLASH (sFLASH) with a high harmoni of an external laser
is a method to ombine the advantages of high harmoni generation (HHG) and FELs
in order to improve the performan e of the fa ility. In a rst stage of sFLASH, the
21
st
harmoni of a titanium-sapphire (Ti:Sa) laser is used for seeding. Between 2009 and 2010 the user operation of FLASH was stopped to implement the new 50 m longsFLASH se tion. Here,the ele tronbun hes are overlappedwith theXUVhighharmoni
radiation in order to initiate the ampli ation pro ess. The amplied radiation enables
thenthepossibilityfortime-dependentexperimentswithoutthela koferrati u tuations
in the time and spe tral domainand without arrival timeu tuations between the XUV
radiation and the external laser. This hapter introdu es the main features of HHG and
the ommon Free-Ele tron Laser (FEL) operation mode SASE. It further des ribes the
ideahowto ombinetheadvantagesofea hte hnique. Finally,thestru ture ofthis thesis
is presented.
1.1 High harmoni generation
HHG is a method to generate XUV and x-ray radiation. Therefore, a femtose ond laser
pulse is fo used into a noble gas target. The orresponding laser eld bends the atomi
potentialof thenoblegasatoms. Thisdeformationleadstoanin reasedpossibilityofthe
valen e ele trons to tunnel through the atomi barrier. The subsequent motion is then
determined by the laser eld. The ele trons are a elerated by the laser eld and have a
small possibility to re ombine with their parent ion. The gained energy is then emitted
in terms of XUV orx-ray photons [Winterfeldt et al., 2008; Changand Corkum, 2010℄.
HHG is a oherent pro ess where a fra tion of the driving laser is onverted into odd
multiples of the fundamental frequen y. The onversion e ien y (CE) of this pro ess
limitsthepulseenergyofthe harmoni stothenanojouleorsub-mir ojouleregime. Inthe
timedomainthehighharmoni radiation onsistsofaburstofattose ondpulses
[Papado-giannisetal.,1999;Dres heretal.,2001;Hents heletal.,2001;Agostini,2004℄andenables
real-time observation of atomi -s ale ele tron dynami s [Dres her et al., 2002; Corkum
and Krausz, 2007℄. The driving laser and the XUV radiation are syn hronized
peak br
illianc
e (phot
ons/(s mr
ad mm 0.1% b
w))
2
2
energy (eV)
Figure1.1: Peak brillian eof dierent lightsour es. Courtesyof S. S hreiber, DESY.
limitedbytheirpulseduration. Duetothespe tralandtemporalstabilityhighharmoni s
arethetoolof hoi eformonitoringatomi -s aledynami s[T.Pfeifer,2006℄. Pump-probe
experiments [Uphues et al., 2008; Allison, 2010℄ and dira tive imaging [Ravasio et al.,
2009;Sandberg etal.,2007℄for instan e are some appli ationsof HHG.
1.2 Self-amplied spontaneous emission
The most ommon operation mode of an FEL is Self-Amplied Spontaneous Emission
(SASE), produ ing highly-brilliant radiation down to the XUV and x-ray regime. In
ontrast tosyn hrotrons the light pulsesof anFEL have a mu hhigher brillian e(gure
1.1) and are oherent. The pulse duration is in the femtose ond regime and is mu h
shorter than the pi ose ond pulses delivered by syn hrotrons. The pulse energy ex eeds
those of HHG by up toveorders of magnitude.
The radiation is suited for oherent x-ray dira tion imaging [Chapman et al., 2006;
Boganetal.,2008℄,inlineholography[Rosenhahn etal.,2009℄,Fourier-transform-
holog-raphy [Boutet etal., 2008; Man uso etal., 2010; Mar hesini et al.,2008℄, oherent x-ray
experi-ments [Krikunova etal., 2011;Chapman et al.,2007; Barty et al.,2008℄.
Unfortunately,thegenerationofthelightpulsesisbasedonastatisti alpro ess,leading
to errati u tuations of the pulse energy and the spe tral and temporal domain
[A k-ermann et al.,2007℄. Moreover, the ele tron a eleration pro ess introdu es arrivaltime
u tuations of the ele tron bun h at the undulator entran e, whi h leads to a temporal
jitterwithrespe t toexternallaser pulses. Thus, theresolutionof time-dependent
pump-probe experiments is redu ed [Dres her et al., 2010; Maltezopoulos et al., 2008; Azima
et al., 2007; Rad lie etal., 2007℄.
1.3 HHG vs. SASE
Both,HHG andSASE arete hniques togenerate XUV radiationfortime-resolved
exper-iments. HHG is able to provide oherent radiation in the nanojoule regime. In ontrast
the pulse energy of FLASH is up to
300 µ
J [S hreiber, 2011b℄. Due to the sto hasti natureof SASEthe radiationdiers initsspe trumand intensity onashot-to-shotbasis.These jittersour es are absent inHHG.
In HHG the radiation onsists of odd multiples of the driving laser frequen y. The
lowestprodu edwavelengthdependsonthefundamentalwavelengthandthetargetwhi h
usually isanoblegas. Itisfurtherpossibletoshiftand/orbroaden thespe trumslightly
at the expense of the pulse energy. In an FEL the wavelength depends on the ele tron
energy and onthe
K
-parameterof the used undulators and an be hanged ontinuously. FLASH for instan e an produ e radiation with4.1
nm< λ < 47
nm [S hreiber et al., 2011;Tiedtke etal.,2009℄.Thepulse lengthofthe FEL radiationdepends ontheele tron bun h anditispossible
to hange the pulse length of FLASH between
10
fs< ∆t < 200
fs (rms) [S hreiber, 2011b℄. In ontrast, the pulse length of the HHG related radiationis determined by thepulse lengthof the driving laser. Furthermore, the driving laser pulse length has a large
impa t on the CE. Thus, the pulse length of the harmoni s is more or less xed. On
the other hand, the arrival time jitter of the radiation is only determined by the jitter
of the driving laser. In ontrast to an FEL, where the a eleration pro ess introdu es
arrival time u tuations in the order of
FWHM
≈ 80
fs (FullWidth at Half Maximum) [S hreiber,2011b℄. Thus, thetemporaljitterofanFELismu hlarger omparedtoHHG.1.4 Seeded FELs
Seeding anFELis analternativete hnique toSASE, wherethelasing pro ess isinitiated
by an external laser and not from spontaneous emission. The external laser introdu es
a density modulation of the ele tron bun h leading to oherent radiation. In this ase,
only a fra tion of the ele tron bun h is being modulated, thus the pulse length of the
orresponding radiationdepends on the seed pulse length. In addition,the ampli ation
pro esstakespla eonlywithinthepulselength. Hen e,thearrivaltimeu tuationsofthe
jitter issmaller than the width of the ele tron bun h.
The wavelength of the radiationis determined by the seed, thus, itis an odd multiple
of the HHG driving laser. The pulse energy of a seeded FEL should be omparable to
SASE radiationinthe mi rojouleregime [Milt hevetal., 2008℄.
In summary, a seeded FEL ombines the advantages of the stability of HHG and the
highpowerofSASE.Therefore,oneisabletoimprovethetimeresolutionof orresponding
experimentsand to enhan e the energy resolution of wavelength-dependent experiments.
So far FLASH is operated in the SASE mode. This thesis des ribes the joint venture
between the University of Hamburg and DESY, where the FLASH fa ility is upgraded
to a seeded FEL (sFLASH). It is meant to get further insights of the pe uliarities of a
seeded operationmode and is arst step towards a seeded user fa ility(FLASH II).
1.5 Thesis stru ture
This thesis is separated into ve hapters. The following hapter des ribes the
funda-mental physi al on epts of HHG and FEL theory ne essary to understand this thesis.
In hapter 3 the experimental environment, i.e. FLASH, sFLASH, and the onstru ted
harmoni sour e are des ribed. Chapter 4 is the main part of this thesis and des ribes
the hara terization of the harmoni sour e and the a hievements of the seeding
experi-ment at FLASH. Chapter 5 on ludes and presents an outlook of sFLASH. Finally, the
This hapter is addressed to the fundamentals of this thesis. Primarily the theoreti al
on ept of HHG and FEL theory willbe presented.
2.1 Des ription of laser pulses
The behavior of laser pulses is based on opti s and has been investigated thoroughly in
the past. A omprehensive des ription an be found in [Saleh and Tei h, 1991; He ht
and Zaja , 1997; Frank L. Pedrotti, 1987; L.Bergmann et al., 2004℄. This hapter only
a ounts for the fundamentals and summarizes the ne essary issues to understand the
following hapters.
2.1.1 Geometri al opti s
The simplest way to des ribe the behavior of lightis interms of geometri al opti s. The
theory is based on a few postulates as des ribed for instan e in [Saleh and Tei h, 1991℄.
It is assumed that light travels in form of rays in a medium whi h is des ribed by the
refra tive index. The path of the rays is determined by Fermat's prin iple: "Light rays
travel along the path of least time". Based on these three prin iples the propagation of
light rays is determined.
Forrayshavingasmall angle
ϑ
tothe opti alaxisofanopti alelement,their behavior is des ribed by the paraxial approximation: sin(ϑ)
≈ ϑ
. One onsequen e is the image equation for athin lens ora spheri al mirror:1
b
+
1
g
=
1
f
,
(2.1)where
g
isthe distan e fromthe obje ttothe lens,b
is the distan e fromthe lens tothe image, andf
is the fo al length of the lens. The equation an be employed to al ulate the position of an image depending on the distan e of the obje t to the fo using opti .Theparaxialapproximationla ksonthefa tthataberrationsasforinstan e astigmatism
annot be des ribed. Astigmatismis indu ed, whenever the lightrays have a large angle
ϑ
tothe opti alaxis of the fo usingelement. In this ase, two linefo iappear instead of one pointfo us. Thelinefo iinthesagittaland meridionalplaneare separated inspa e:f
m
= f
·
sin(ϑ)
f
s
= f
·
1
sin
(ϑ)
where
f
m
, andf
s
are the ee tive fo al lengths of the meridional and sagittal plane, respe tively.In general, geometri al opti s is an idealized model and implies that
λ → 0
. It is not possible to des ribe ee ts like interferen e or dira tion. To a ount for the wavenature of light the ele tromagneti eld theory has to be used. Some major results will
bepresented withinthe next se tion.
2.1.2 Gauss opti s
Light an be des ribed with the Maxwell equations and it is possible to dedu e simple
statements by employingsome assumptions. Firstly, the wave equation
∇
2
~E =
c
1
2
∂
2
~E
∂t
2
,
(2.3)with
~E
theele tri eld,c
thespeed oflight,andt
the timeisadire t onsequen e ofthe Maxwell equations. Foranele tromagneti wave~E(~r, t) = E(~r)T(t)
equation 2.3 an beseparated in spa e and time and an be rewritten as:
∇
2
+ k
2
E(~r)
= 0,
(2.4)d
2
dt
2
+ ω
2
T
(t)
= 0,
(2.5) withω = kc,
andk =
2π
λ
,
where
ω
istheangularfrequen y,k
isthewaveve tor,andλ
isthewavelength. Equation 2.4 is known as the Helmholtz equation. Assuming an ele tromagneti wave travelingalongthez-dire tionwithaslowlyvaryingenvelope(paraxialapproximation),onesolution
of equation 2.4 orresponds toa Gaussianbeam [Ja kson, 1999℄ with
w(z) = w
0
s
1 +
z
z
R
2
(2.6)R(z) = z
1 +
z
R
z
2
(2.7)ζ(z)
=
tan−1
z
z
R
(2.8)w
0
=
r
λz
R
π
,
(2.9)where
w(z)
isthe beamwidth depending onthe spatial oordinatez
andz
R
the Rayleigh length.R(z)
orresponds to the wavefront radius of urvature,ζ(z)
represents the phase−200
−100
0
100
200
0
20
40
60
80
100
120
spatial coordinate z (mm)
beam width w (µm)
2 z
R
w
0
4 z
R
Figure 2.1: Cal ulatedbeamwidth for aGaussianbeam with
λ = 38.1
nm,w
0
= 40 µ
m, andz
R
= 132
mm.retardation of the beam relative to a uniform plane wave, and
w
0
is the minimal beam widthatitswaistdependingonthe wavelengthλ
andontheRayleighlengthz
R
. Ingure 2.1 the al ulated beam widthw(z)
is shown for a Gaussian beam withλ = 38.1
nmand
w
0
= 40 µ
m. In this ase the beam widthw
at positionz
is dened in terms ofthe intensity: the Gaussianbeam has its peakintensity onaxis forevery position
z
. The beamradiusw
, however, is dened asthe radialdistan e tothe opti alaxisatwhi hthe intensityisredu ed to1/e
2
ofitspeakintensity. AlsotheRayleighlengthdepends onthe
beam waistand refers tothe distan e
z
atwhi h the fo alspot area istwi e asbig asat its waistw(z = 0) = w
0
,i.e. the beam radius is√
2w
0
. The intensity at that positionisI(z =
±z
R
) = 0.5
· I(z = 0).
Dependingonthe ontextdierentdenitionsofthebeamwaistareused. Asexplained
before
w
orresponds tothe radius,where the intensity de reases by a fa tor of1/e
2
. In
ontrast, the
FWHM
value of a beam is dened as the diameter of the beam at whi h the intensity dropsdown by afa tor oftwo. The rms widthσ
of abeam isdened asthe mean of the photon-distribution. In ase of a Gaussianbeam the rms-width (root meansquare) ishalf asbigasthe
w
-radius(w = 2σ
). The dierent denitionsof the Gaussian beam widthsare onne tedas follows:FWHM = 1.177w = 2.35σ
Besides, gure 2.1 indi ates that the beam width in reases linearly in the near eld.
For
z
≫ z
R
equation 2.6 an berewritten asw(z) =
w
0
z
R
· z
Thedivergen e,i.e. halfopeningangleofthe one anbeexpressedby ombiningequation 2.9 and 2.10.
ϑ =
w
0
z
R
=
λ
πw
0
,
(2.11) with tan(ϑ)
≈ ϑ.
The phase retardation
ζ(z)
of a Gaussian beam (equation 2.8) hanges its sign at the position of the waist [Gouy, 1890℄. This ee t is known as the Guoy phase-shift∆ζ = π
between the near eld on ea hside ofthe fo us.Stri tlyspeaking, the aboveequationsare onlyvalidfora GaussianTEM
0
inx- andy-dire tion, respe tively. [Siegman, 1993℄introdu ed the beam qualityfa tor
M
2
toextend
the above equations. The beam quality- or beam propagation- fa tor is dened as the
times dira tion limited (TDL) value for an arbitrary real beam ompared to a TEM
0
Gaussian beam". It is further shown that the above equation have to be modied in a
way that
λ → M
2
λ
, where
M
2
≥ 1
in order to des ribe modes deviating from TEM0
mode in x- and y- dire tion. During this work the modied beam width, as well as themodied Rayleighlength is being used. The equations 2.6 and 2.9 an berewritten as
W(z) = W
0
s
1 +
z
z
R
2
(2.12) withz
R
=
πW
2
0
M
2
λ
,
(2.13)where
W
0
is the modied beam waist. During this work laser pulses are assumed to be Gaussian in the spatial and spe tral domain, unless it is expli itlynoted to be dierent.Adetaileddedu tionofGaussianbeamproperties anbefoundin[Siegman,1986℄. Laser
related fundamentalsare further dis ussed in[Svelto, 1982;Rulliere,1998℄.
2.2 XUV opti s
The XUV regime is in the range of
10
nm≤ λ ≤ 120
nm. Unfortunately, there are no existing materialshaving a high ree tivity within this regime. Most of the radiation isbeingabsorbedwithinapenetrationdepthoflessthan
1 µ
m,leadingtotypi alree tivity ofR
≤ 10
−4
at normal in iden e [Jaeglé, 2006℄. A solution to over ome this pe uliarity
are grazing in iden e mirrors, as well as normal in iden e multilayer mirrors. Grazing
in iden e mirrors have a large bandwidth ompared to multilayer mirrors. As explained
in hapter 2.1.1a fo usingopti employed ata grazingin iden e angle leads toan
astig-matism,whi hisusuallynotwanted. Thus, grazingin iden emirrorsare ommonlyused
for simple (non-imaging)ree tions. In ontrast, multilayer mirrors an be employed at
anglesaroundzero degree. Thosemirrors onsistof aperiodi thin layerstru tureontop
of the substrate, leadingto onstru tiveinterferen e of the ree ted XUV radiation. The
a hieved bandwidth is small ompared tothe grazing in iden e mirrors, but the fo used
found in [Jaeglé, 2006; Attwood, 2000℄. [Lawren e Berkeley National Laboratory, 2011℄
providesaweb-interfa eto al ulatetheree tivityandtransmissionofvariousmaterials.
A report ondira tiongratingsinthe XUVregimewhi hareused inthis thesishas been
givenby[S hroedter,2009℄andbasi gratingtheory anbefoundin[PalmerandLoewen,
2005℄.
2.3 High Harmoni Generation
HHG is a oherent nonlinear intera tion of light and matter [Shen, 1984; Menzel, 2001℄:
if a high intensity laser is fo used into a gas target, the atoms emit light pulses of odd
multiplesofthe in identfrequen y [W.F.Drake,2006;Jaeglé,2006℄. Therstobservation
of opti al harmoni s (se ond harmoni generation) has been des ribed in 1961 [Franken
etal.,1961℄oneyear aftertheinventionofthe laser[Maiman,1960℄. Firstexperimentson
low-orderharmoni generationingashavebeenpublished in[New andWard,1967;Ward
and New, 1969℄, whereas the rst high-order harmoni s have been observed in the late
1980s[M Phersonetal.,1987;M.Ferrayetal.,1988℄. Earlypublishedresultsarereviewed
in [Huillieret al., 1991;Gavrila, 1992℄. More re entreviews an befound in [Winterfeldt
et al., 2008; Brabe and Krausz, 2000℄. In table 2.1 a sele tion of harmoni s generated
withan
800
nmTi:Salaserare shown,whereH21
willdrawanin reasedattention during the following work.Table 2.1: Sele ted harmoni orders indierent units.
harmoni order
λ
(nm)E
photon
(eV)H
01
800
1.55
H09
88.9
13.95
H11
72.7
17.05
H13
61.5
20.15
H19
42.1
29.45
H21
38.1
32.55
H23
34.8
35.65
H61
13.1
94.54
2.3.1 Three-step modelA simple empiri model has been developed to des ribe the pro ess of HHG [Corkum,
1993;Kulanderetal.,1993℄. Itiswidely knownasthesimpleman'smodel,thethree-step
model, or in the 1990s asthe two-step quasi- lassi al approa h [L'Huillier etal., 1993℄.
The model des ribeshowthe bound ele tronsof the gasatoms tunnelthrough itsatomi
barrier. In the se ond step the ele trons move lassi ally and may return tothe nu leus.
Thoseele tronsthatre ombinetothegroundstateemitharmoni sasillustratedingure
(a)
(b)
(c)
Figure 2.2: S hemati representation of the three-step model: (a) represents the tunnel
ionization, (b) the a eleration inthe ontinuum, and ( ) the re ombination.
First step: ionization
Wheneveralaserisfo useddown tointensitiesinthe orderof
10
15
W/ m
2
the amplitude
ofthe orrespondingele tri eldapproa hes
10
9
V/ m. Theseeldstrengthsex eedthat
of anatomi Coulombeld, leadingto adeformationof itsatomi potential. Hen e, this
deformation in reases the possibility for a bound ele tron to tunnel through its atomi
barrier. Tunnel ionization has rstly been des ribed for hydrogen [Keldysh, 1665℄ and
was extended toa generalizedtheory forarbitrary atoms [Ammosov et al., 1986℄.
Insummary,thetheoryintrodu estheKeldyshparameter[MiyazakiandTakada,1995℄:
γ =
s
I
p
2U
p
,
where
I
p
is the ionization potential of the gas andU
p
is the ponderomotive potential. The ponderomotive potential orresponds tothe averagedenergy of the ele tron's quivermotion in an ele tromagneti eld and isdire tly proportional to the wavelength square
of the driving laser. It is dened as
U
p
=
e
2
2m
e
ǫ
0
cω
2
I
∝ Iλ
2
,
where
e
is the elementary parti le harge,m
e
the ele tron mass,ǫ
0
the va uum permit-tivity,c
the speed of light,ω
the angular frequen y,I
the laser intensity, andλ
is the wavelength of the laser. However, forγ > 1
the ionizationtakes pla e interms of multi-photon ionization [Lin, 2006℄ and forγ < 1
through tunneling. Multi-photon ionization is a simultaneous absorption of several photons, leading to the ionization of the atom[Mainfray and Manus, 1991℄.
The numberof ele trons
N
e
an be dedu ed asin [Miyazakiand Takada, 1995℄ and is des ribed by the rate equationwhere
N(t)
isthetimedependentneutralatomdensity andW(t, E)
the tunnelionization rate. The ionizationrate depends onthe eldstrengthE
and leadstoanele tron density [Ammosov etal.,1986℄N
e
(t) = N
0
1 −
exp−
Z
t
−∞
W(t, E)dt
,
where
N
0
is the neutral atom density att = 0
.Besides, ionizationof the gaseous medium hanges the dira tive index and may lead
to self-fo usingof the driving laser whi hitself has inuen e on the HHG pro ess. HHG
in a rapid ionizing mediumhas been des ribed thoroughly in [Raeet al., 1994; Raeand
Burnett, 1993a,b℄. However, laser elds in the order of the atomi Coulomb eld of
10
8
− 10
9
V/ m, ne essary to drive the HHG pro ess, are known as strong elds where
the lassi alperturbation theory is not valid[Brabe , 2008℄.
Se ond step: propagation
Assoonasthe ele tronisionizeditistreatedasafree-ele tronand itssubsequenttra
je -toryisdeterminedbythe lasereld. Dependingonthephase ofthelasereld atthetime
of the ionizationthe ele trons gain dierent energies before the ele tri eld hanges its
algebrai sign. The sign hange auses the ele tron to de elerate and then to a elerate
ba k tothe parention. The averaged kineti energyanele tron angain inanos illatory
laser eld is [Gallagher,1988; Krauseet al.,1992℄:
1
2
m
e
v
2
= U
p
1 + 2
os2
(ωt
0
) ,
where
v
is the velo ity andt
0
is the ionization time. This equation indi ates that the energy an ele tron an gain is betweenU
p
and3U
p
. Classi al simulations of the prop-agation have been published in [Corkum, 1993℄ and reveal a slight modi ation of thepreviously dedu ed result. It is shown that the maximal energy an ele tron an gain is
3.17U
p
. Itisalsoshownthatanele tron,whi hgainedtheenergy of3.17U
p
hastunneled atωt
0
= 17
◦
or197
◦
of the driving laser. Thus, the maximal energy anele tron an gain
o urs twi e a laser y le. Here it is perspi uous that for an ellipti al polarization of
the driving laser the ele tron never returns to the parent ion. Thus, the CE of HHG is
strongly dependent onthe ellipti ityof the driving laser polarization[Budil et al., 1993;
Dietri h etal., 1994;S hulze etal., 1998℄.
Third step: re ombination
Some of the free-ele trons may re ombine with their parent ionand emit aphoton with
a maximal energy of
E
max
= 3.17U
p
+ I
p
,
(2.14)where
I
p
is the ionization potential of the orresponding gas atoms. Equation 2.14 is known as the uto law and des ribes the uto behavior of the harmoni spe trum[L'Huillier et al., 1993℄. The smallest possible wavelength in the harmoni spe trum
depends on
•
the wavelength of the driving laser,•
the intensity of the driving laser,•
and the ionizationpotential of the gaseousmedium.The larger the wavelength, intensity, and ionizationpotential,the smallergets the
wave-length of the highest harmoni order.
In summary,the three-step model is a quasi- lassi al model whi h des ribes the HHG
pro esswithinthreedistin tsteps. Aself- ontainedfullyquantumme hani aldes ription
of the HHG pro ess is given in [Lewenstein et al., 1994℄ and will be des ribed in the
followingse tion.
2.3.2 Quantum me hani al des ription
The quantum me hani al des ription of the HHG pro ess is based on the single a tive
ele tron (SAE) approximation[Kulanderand Res igno,1991℄ andis known asthe strong
eld approximation (SFA) or as the Lewenstein model [Lewenstein et al., 1994℄. In the
SAE approximation it is assumed that the bound state wave fun tion evolves into a
ontinuum wave pa ket. One part of the wave pa ket willnever return, whereat another
part will return to the ioni ore. Here dierent possibilities a probable. Either the
ele tron s atters o, or hanges the dire tion to gain energy, or overlaps with the bound
state, whi h leads to a time dependent dipole moment and photo emission. The system
an bedes ribed with the orresponding time-dependent S hrödinger(TDSE) equation
i
h
∂
∂t
|Ψ(
~r, t)
i =
−
h
2
2m
e
^
∇
2
+ ^
V(
~r) − ~E(t)^x
|Ψ(
~r, t)
i,
(2.15)where
V(
^
~r)
is the atomi potential operator,~E(t)
is the time-dependent laser eld, and^
x
is the dipole operator. However, the solution of equation 2.15 is a superposition of bound,os illating,and outgoing wave pa kets. In orderto solve the TDSE the followingapproximationshave tobe made:
•
the system evolution depends only onthe ground state|0
i
. All other states donot ontribute and an be negle ted.•
depletion of the ground state an be negle ted.•
the ele tron in the ontinuum moves asa free-ele tron in an ele tri eld, thus the atomi potentialhas noee t and an be negle ted.The Fourier transform of the indu ed dipolemoment
ex(t) =
hΨ(x, t)|ex|Ψ(x, t)i
(2.16)leads dire tly to the harmoni spe trum.
The ansatz
|Ψ(x, t)
i = e
iI
p
t/
h
a(t)|0
i +
Z
d
3
~q b(~q, t)|~qi
,
(2.17)where
a(t)
is the ground state amplitude andb(
~q, t)
is the amplitude of the ontinuum states, leads to the solution of the TDSE. Be ause the depletion of the ground state isnegle ted, the amplitude of the ground state is set to
a(t)
≈ 0
,a(t)
≈ 0
. Equation 2.16 an be evaluated by further hanging the variables to the anoni al momentum~p = ~q + e~
A(t)/c
:ex(t) = i
Z
t
0
dτ
2m
e
hπ
ν + iτ
3/2
× d
∗
x
(
~p
s
− e~
A(t)/c)
· e
−iS(~p
s
,t,τ)
× E(t − τ)
os(ω(t − τ))d
x
(
~p
s
− e~
A(t − τ)/c) + c.c.
(2.18)with
~p
s
=
~p
s
(t, τ) =
Z
t
t−τ
dt
′
e~
A(t
′
)/cτ
andS(
~p
s
, t, τ) =
Z
t
τ
dt
′′
(
~p
s
− e~
A(t
′′
)/c
2
)
2
2m
e
+ I
p
!
,
where
d
x
(
~q) =
h~q|ex|0i
is the omponent of the dipole matrix element parallel to the polarizationfor bound-free transitions,d
∗
x
its omplex onjugate,S(
~p
s
, t, τ)
is the quasi- lassi al a tion,~p
s
the momentum at the stationary point of the quasi- lassi almomen-tum, and
τ = t − t
′
is the time the ele tron is travelling in the ontinuum. Equation
2.18 an beinterpreted interms ofthe simpleman's model: thebottomline ontains the
matrix element forthe transitionbetween the bound stateand the ontinuum,
S(
ps
, t, τ)
orresponds to the propagation operator in the ontinuum, andd
∗
x
orresponds to the transitionfrom the ontinuum ba k intoa bound state.A Fourier transformation of equation 2.18 al ulates the harmoni spe trum. It an
further be shown [Salieres et al., 2001℄ that the k
th
Fourier omponent orresponds to a
sum over the quantum orbits
ex
k
=
X
n
a
n
exp(−iS(
pn
, t
n
, t
τ
)),
i.e. the kth
harmoni depends oninterfering ontributions of dierent traje tories
(Feyn-man's path integral). Finally, it should be pointed out that this model is also able to
reprodu e the uto law of the simple man's model [Lewenstein et al., 1994℄. Equation
2.14 has tobe modied:
where the fa tor
F(U
p
/I
p
)
is a orre tion term due to quantum tunneling and quantum diusion.F(U
p
/I
p
)
is equal to1.3
forI
p
≪ U
p
and de reases to1
asI
p
grows.HHGisaperiodi pro esswhi ho urseveryhalflaser y le. Moreover,one anobserve
that two ele tron traje tories ionized at dierent phases of the laser eld will result in
identi al momentaatthe timeof re- ollision[Lewensteinet al.,1995;Bal ouetal.,1999;
Milo²evi¢ and Be ker, 2002; Liu et al., 2009℄. These two traje tories are known as the
short and the long traje tory orresponding tothe time the ele tron is traveling through
the ontinuum. Sin e the pro ess takes pla e every half period of the laser y le
T/2
the Fourier transformisa dis rete fun tionwith a separation orresponding to1/(T/2) = 2f
. That is why the spe trum onsists only of odd multiple of the frequen y of the drivinglaser [Winterfeldt et al., 2008℄. This symmetry an be avoided by adding the se ond
harmoni of the driving laser tothe HHG pro ess [Mauritsson etal.,2006℄.
2.3.3 Phase mat hing
So far only a single atom has been onsidered. However, HHG is a multi-parti le
pro- ess and the olle tive ee ts have to be taken into a ount. For e ient HHG the
phase- onditions of the atoms have tobe set that the orresponding radiationadds
on-stru tively. The wave-ve tor mismat hof the
q
th
harmoni order an bewritten as
∆k = qk(ω
f
) − k(qω
f
),
(2.20)where
k
represents the wave-ve tor,q
is the harmoni order, andω
f
is the frequen y of the driving laser. Equation 2.20 and the followingis mainlytaken from[T.Pfeifer, 2006℄.The wave-ve tor
k = k(ω)
depends on itsfrequen y. It an bewritten ask(ω) = k
vac
(ω) + k
disp
(ω) + k
plasma
(ω) + k
geom
(ω),
(2.21)where
k
vac
(ω) = 2πω/c
is the wave-ve tor in free spa e,k
disp
(ω)
is the ontribution due todispersion in a neutral medium,k
plasma
(ω)
is due to dispersion in a plasma,andk
geom
(ω)
is due to geometri dispersion. Constru tive interferen e o urs, if the phase-mismat hvanishes:∆k = 0.
The phase-mismat hin free spa e
∆k
vac
= q
ω
f
c
−
qω
f
c
= 0
(2.22)is zero, thus, only the dispersion due to neutral atoms, plasma, and geometri al ee ts
havetobe onsidered. Asthetherefra tiveindexofamediumdependsonthewavelength
the phase-mismat hdue to dispersionin aneutral medium an be expressed as
∆k
disp
(ω
f
) = (n(ω
f
) − n(qω
f
))
qω
f
c
,
(2.23)where
n
isthe refra tiveindex. This ontributiontothephase-mismat hdependsonlyon the dieren eof the refra tive indi es. In ase of the800
nm driving laser the refra tiveindexislargerthanoneandsmallerthanonefortheXUVradiation,thus,thewave-ve tor
mismat his apositivevalue.
As explainedin the previous se tion onlya fra tion of the ionized ele trons re ombine
and emit harmoni radiation. Themajority ofthe ele tronsremain ionizedand lead toa
hangeof the refra tive index:
n
plasma
(ω) =
r
1 −
ω
p
ω
2
withω
p
=
s
e
2
N
e
ǫ
0
m
e
,
where
ω
p
isthe plasmafrequen y. The orresponding plasma ontributionto the phase-mismat h an be al ulated:∆k
plasma
(ω) = qk
plasma
(ω
f
) − k
plasma
(qω
f
) =
ω
2
p
(1 − q
2
)
2qcω
f
(2.24) and is∆k
plasma
(ω) < 0
forq > 1
.Thelast ontributioninequation2.21dependsonthefo usingpropertiesofthedriving
laser. As explained in hapter 2.1.2 the fo used Gaussian beam ontains an additional
phase ontributionalongthez-dire tionwhi hisknown astheGuoyphase(equation2.8).
The wave-ve tor an be approximated for
z << z
R
tok
geom
(z) =
dζ(z)
dz
≈
1
z
R
.
Thus, the phase-mismat his:
∆k
geom
= qk
geom
(ω
f
) − k
geom
(qω
f
) =
q − 1
z
R
.
(2.25)The total phase-mismat h an be al ulated by ombiningequation 2.20 -2.25:
∆k
tot
= (n(ω
f
) − n(qω
f
))
qω
f
c
|
{z
}
>0
+
ω
2
p
(1 − q
2
)
2qcω
f
|
{z
}
<0
+
q − 1
z
R
| {z }
>0
.
(2.26)In order toa hieve onstru tiveinterferen e the phase-mismat hintrodu ed by free
ele -trons have to an el out the mismat h due to dispersion and geometri al onsiderations.
During this work only loose fo using geometries are used and therefore the geometri al
part of the phase-mismat h an be negle ted. Thus, the phase-mismat h is only due to
free-ele tronsand dispersioninthe neutralmedium,i.e. argon. NotethatHHG inhollow
waveguides introdu es an additional term, whi h is addressed to waveguide dispersion
[Durfee etal.,1999℄. This term isabsentduring this workan annot beused tooptimize
2.3.4 Properties of the harmoni s
In gure 2.3 a typi al HHG spe trum is plotted s hemati ally. The spe trum an be
divided in three se tions: the perturbative regime of the lower harmoni s whi h an
betreated with the lowest-order perturbation theory [Gavrila, 1992℄, the mid-harmoni s
whi hformaplateau,and thehighestharmoni s inthe uto. Thehigh-orderharmoni s
of the plateauandthe uto regimeare des ribed intermsof the SFA,whi hisnot valid
for lowerorders. Here the atomi potentialhas an inuen e onthe ele tron's movement,
thus the approximationsmade inthe SFA are not fullled anymore ( hapter 2.3.2).
1
3
5
7 9
11
13
15
17
19
21
23
25
27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61
plateau - SFA regime
cutoff
perturbative
regime
Figure 2.3: S hemati representation of a typi al HHG spe trum (odd orders shown on
top).
Asexplainedbefore, theharmoni s are separatedby
2ω
f
, whereω
f
is thefrequen y of the driving laser. The spe tral bandwidth of ea h harmoni depends onthe intensity ofthe driving laser [He et al., 2009℄ and is larger, if the long ele tron traje tory is favored,
thus for higherlaser intensities. Moreover, destru tive interferen e between the long and
the short traje torymay o ur at high intensities[Xu et al., 2008; Brunetti et al., 2008;
Zaïr et al., 2008℄. It is further possible to slightly shift the harmoni s towards lower
wavelengths. As explainedin hapter2.3.1 the probability toionize anatom depends on
the intensity of the driving laser. Thus, for high intensities the number of free-ele trons
in reases and the refra tiveindex hanges (se tion2.3.3). This variationof the refra tive
index leads to a spe tral blueshift of the driving laser and therefore to a blueshift of
the harmoni s [Wahlströmet al., 1993℄. Detailed al ulations an be found in [L'Huillier
et al., 1992; Raeand Burnett,1993a; Raeet al., 1994; Kan etal., 1995℄.
The harmoni sforma trainof attose ond pulsesinthe time domain[Paulet al., 2001;
Antoineetal.,1996℄. Inordertomeasuretheattose ondpulsestru tureoftheharmoni s,
the phase dieren ebetween two onse utive harmoni s is measured. This measurement
isbasedona ross- orrelationte hniquebasedonatwophotontransition. The te hnique
RABITT(Re onstru tionofAttose ondharmoni BeatingByInterferen eofTwo-photon
Transitions)uses the fa tthat the drivinglaser frequen y isexa tlyhalfof the frequen y
laser intoa gas, the spe trum onsists of peaks orrespondingto the harmoni orders. If
in additionthe driving laser overlaps in spa eand time, sidebands appear whi h depend
ontherelativephase ofthe onse utiveharmoni sandthetime-delayofthedrivinglaser.
In this way, the phase dieren e an be dedu ed. In ombinationwiththe harmoni
am-plitudethetemporalprole an bere onstru ted[Hents heletal.,2001;Pauletal.,2001;
Kienberger et al., 2002℄. Moreover, the time-dependent frequen y an be measured and
depends linearly onthe harmoni orderand on the hirp of the drivinglaser [Mauritsson
et al., 2004; Norinet al., 2002℄.
The oheren e time has to be onsidered separately for the long and for the short
traje tory. The short traje tory has a phase that does not vary mu h with intensity,
thus the emitted radiation has a long oheren e time and is well ollimated[Lewenstein
et al., 1995℄. In ontrast the phase of the long traje tory hanges rapidly with the laser
intensity [Gaarde et al., 1999℄, therefore, a shorter oheren e time and a more divergent
emission appears [Salières et al., 1995℄. The spatial oheren e has been investigated by
a series of a two-slit experiments. It has been shown that for moderate laser intensities
the harmoni s have a high degree of spatial oheren e. For high intensities the spatial
oheren e isredu ed, due tothe time-dependent index of refra tion[Ditmireet al., 1996;
Bartels etal.,2002℄.
2.4 Free Ele tron Laser
In 1971 the prin iple of an FEL has been proposed by [Madey, 1971℄. After the rst
observationofmagneti bremsstrahlungat
10.6 µ
m[Eliasetal.,1976℄the rstFEL os il-lator [Dea on etal.,1977℄went into operation. Today, almost35years later, a variety ofdierent FEL fa ilitieshavebeen onstru ted inorder to providehigh-brightness,
oher-ent, ultrashort, and short wavelengthradiationto the s ienti ommunity. For example
LCLS[Emma,2009;BoutetandWilliams,2010℄,XFELatSpring-8[Shintake,2010;Hara
etal.,2010℄andtheEuropeanXFEL[S hneidmillerandYurkov,2010℄areabletoprodu e
FEL-radiationwithawavelengthintheorderof
0.1
nm, whileotherfa ilitieslikeFLASH ( hapter3.1),SCSSatSPring-8[Shintake,2010℄,FERMIatElettra [Allariaetal.,2010℄,and the SwissFEL [Garvey, 2010; Patterson et al., 2010℄ are operating in the VUV and
the softX-ray regime.
In ontrast to the syn hrotron light sour es, the FELs generate photon pulses with a
mu hhigherdegreeof oheren e anda pulselength down tothe femtose ond range. The
HHG sour es, on the other hand, are also apable to emit ultra-short oherent
XUV-pulses, but with anenergy in the nanojoules regime,while FELs typi allydeliver
mi ro-or millijoulepulses. FLASH for instan e is able to deliver pulses up to
300 µ
J [Treus h and Feldhaus, 2010℄,whereas LCLS has observed apulse energy of3
mJ [LCLS, 2011℄.Due to the phase and amplitude jitter of the a elerating ele tri eld, the ele tron
bun hes arriveatdierenttimesatthe undulatorentran e omparedtoanexternallaser
(arrival-time jitter). In the SASE operation mode, the FEL amplies the spontaneous
radiation produ ed in the beginning of the undulator se tion. The sto hasti nature of
on a shot-to-shot basis (spe tral and intensity jitter). Furthermore, the pulse shape and
the pulselength diers from shot toshot.
This hapterisontheonehanddire tedtothefundamentalsasso iatedwiththeorigins
of shot-to-shotu tuations, and onthe other hand to alternative operationmodes of an
FEL. DetailsonFEL related pro esses an be found in[S hreiber, 2010;S hmüser etal.,
2008;Huang and Kim,2007; Saldin etal.,1999b℄.
2.4.1 Ele tron sour e and a eleration
The ele trons are generated by a laser driven photo athode and are subsequently
a - elerated in a RF avity generating ele tron bun hes with a length in the pi ose ond
regime. The te hni al solutionas well as adetailed hara terization is given in[Stephan
et al., 2010; Fraser et al., 1986℄. However, the light pulses of the driving laser have a
temporalshot-to-shotjitterrelativetothesyn hronizationsystem whi histransferred to
the ele tron bun hes. Additionally, the ele trons experien etemporalu tuations of the
RF phase leading to a further arrival time jitter of the ele tron bun hes relative to the
syn hronization system.
In order to produ e ultra-relativisti ele tron bun hes with a high peak urrent the
ele trons are a elerated downstream the super ondu ting RF a elerating modules and
are longitudinally ompressed using the bun h ompressors realized as four dipole
mag-neti hi anes. As in ase of the RF gun, depending on their longitudinal position the
ele tronsexperien eadierentRFphaseduringthea elerationintheRF-modules. This
has a twofold ee t - it indu es a head-tail energy variation along the bun h and it also
hangesthemeanenergy and orrespondinglythebun harrivaltime. Espe iallytherst
a elerating modulewhere the bun hes are a elerated toalmost the speed oflighthas a
dire t inuen e onthe arrivaltime relative to the syn hronization system. In the bun h
ompressor,dependingonthemomentumthe ele tronstraveldierentpathways: thetail
ele trons with a lower momentum travel a shorter and the head ele trons with a higher
momentum travel a longer way. This leads to a longitudinal ompressionof the ele tron
bun h.
Therefore, the arrival time u tuations of the ele tron bun hes are mainly introdu ed
bythe rst a elerationmodule aswellas the bun h ompressor. The timing u tuation
aused by the ele tron gun is partly ompensated by the bun h ompressor. A detailed
des ription of the a eleration pro ess an be found in [Wiedemann, 2007℄. The
orre-sponding arrival time u tuation of the ele tron bun hes aswellas te hniques to redu e
those errati u tuationsis des ribed in[Löhl,2009℄.
2.4.2 Undulator radiation
The undulator is a periodi magneti stru ture onsisting of dipole magnets with
alter-nating polarity as illustrated in gure 2.4. The following does only summarize the key
aspe ts of undulators. Details an be found in [Hofmann, 2004℄ and [Duke, 2000℄.
x
y
z
u
Figure 2.4: S hemati representationof anundulator: yellow isthe sine-liketraje tory of
the ele tron and blue isthe emitted radiation.
an be hara terizedwith the undulator parameter
K =
eB
0
λ
u
2πm
e
c
.
(2.27)An ele tron whi htravelsthrough the undulator is due tothe Lorentzfor emoving ona
sine-like traje tory and emits syn hrotron radiation. The radiation emitted at dierent
positionsinthe undulatorinterferesandis on entratedinanarrow onewithanopening
angle of
1/γ
, whereγ =
E
m
e
c
2
is the Lorentzian fa tor. The fundamental wavelength of
the undulator
λ
ph
=
λ
u
2γ
2
1 +
K
2
2
+ γ
2
θ
2
(2.28)dependsonthethe undulatorperiod
λ
u
,the energyofthe ele tron,the undulator param-eter, and the emission angleθ
with respe t to the beam axis. The on axis wavelength (θ = 0
) an be hanged by tuning the ele tron energy or the magneti eld of the un-dulator. The ele trons emit radiation independent of ea h other, thus, the radiation isin oherentforwavelengths smallerthanthe bun hlengthanditspowerisproportionalto
the number of ele trons. The polarizationof the radiation is perpendi ular to the
mag-neti eld and the spe tral width is inversely proportional to the number of undulator
2.4.3 High-gain free-ele tron laser
In the high-gain FELs, the initial urrent density in the ele tron bun h is periodi ally
modulated at the radiation wavelength, thus reating the so- alled mi robun hes. This
mi robun h stru ture leads to oherent radiation. As explained inse tion 2.4.2 the
ele -tronsmoveonasine-liketraje tory. Assumedthey aresuperimposedwithan
ele tromag-neti wavethey angainorlooseenergydependingonthephaseoftheele tron os illation
relative to the phase of the ele tromagneti eld. The energy transfer an be des ribed
with the following equation:
dW
dt
=
~v
· ~F
= −ev
x
(t)E
x
(t),
∝
os((k
l
+ k
u
)z − ω
l
t + ψ
0
) +
os((k
l
− k
u
)z − ω
l
t + ψ
0
)
(2.29)=
os(ψ) +
os(χ)
(2.30)where
dW
is the transferred energy,t
the time,~F
the Lorentzian for e,~v
the velo ity of the ele tron (see gure 2.4) andE
x
is the magnitude of the ele tri eld in x dire tion.k
l
andk
u
are the wave number of the light wave and the undulator, respe tively,z
the positioninthe undulator,ω
l
the angularfrequen y of the lightwave, andψ
in equation 2.30 is known as the ponderomotive phase. It has been shown [S hmüser et al., 2008℄that a ontinuous energytransfer fromtheele tron tothe lightwavealong the undulator
is present, if the ponderomotive phase is onstant. Furthermore, the ondition for a
sustained energytransfertothe ele tromagneti waveyieldsthe samelightwavelengthas
the onaxis wavelength of anundulator (equation 2.28 onaxis, i.e.
θ = 0
). The phase of the ele tromagneti waveisthen shiftedbyπ
everyhalf y leof theos illatorymovement of the ele trons. The energy transfer from the ele trons to the light leads to aredu tionofthe ele tronmomentum. Thesmallertheele tronmomentum,thesmalleristhe radius
of the traje tory (
p = eB
0
r
). This leads to a urrent density modulationof the ele tron bun h. [S hmüser et al., 2008℄ showed also that this pro ess is self-sustained and leadstoasubstru ture of the ele tronbun h. This substru ture are longitudinalmi robun hes
with a distan e of
λ
ph
to ea h other. The ele trons in a mi robun h emit oherent light and itspowerisproportionaltoN
2
e
. Adetaileddes riptionof theself-modulationpro ess is published in [Kondratenko and Saldin, 1980℄.However, the prin ipletogenerate oherent radiationworks only if the ele tron bun h
getsdensitymodulated,whi hisusually realizedwithanele tromagneti wave. Assumed
a seed ispresent, the phase spa e motionof the ele trons withinan ele tromagneti eld
an be des ribed by the pendulum equations, i.e the time derivative of an ele tron's
ponderomotive phase and the hange of the relative energy deviation. In addition, the
inhomogeneous wave equation for the ele tri eld of the lightwave and the evolution of
a mi robun h stru ture oupled with longitudinal spa e harge for es an be des ribed
mathemati ally. These equations an be solved analyti ally only for some simple ases
e.g. assuming a homogeneous harge density and very long ele tron bun hes as well as
negle ting the betatronos illationsand the dira tionof the EM-wave. Howeverin most
dierent tools like GENISIS [Rei he, 1999℄, GINGER [Fawley, 2004℄, and FAST [Saldin
et al., 1999a℄ are available. In gure 2.5 a GENISIS simulation of the mi robun hing
pro ess performedby Sven Rei he is shown.
0
1
2
3
4
s/d
-100
-50
0
50
100
x [
µ
m]
z / λ
tra
n
s
ve
rs
e
b
e
a
m
s
ize
(
µm)
(a)0
1
2
3
4
s/ı
-100
-50
0
50
100
x [
µ
m]
z / λ
(b)0
1
2
3
4
s/
-100
-50
0
50
100
x [
µ
m]
z / λ
( )Figure 2.5: GENISISsimulationofthe mi robun hingpro ess. The plotsshowthe
devel-opmentof the mi robun h stru ture withrespe t tothe transverse and
longi-tudinal oordinates in the bun h. The distan e of the mi robun hes is equal
tothe radiatedwavelength. 2.5(a): initialdistributionof the ele tron bun h.
2.5(b): developing mi robun h stru ture. 2.5( ): fully developed mi robun h
stru ture atthe end of the undulator. (Courtesy of Sven Rei he, PSI)
Inordertosolvethependulumequationsanalyti allythefollowingapproximationshave
to be made: the dependen e of the bun h harge density as well as the ele tromagneti
eld onthe transverse plane isnegle ted (one-dimensionalmodel). Theele tron bun h is
assumedtobeverylong,i.e. theee tsofthetailandheadofthebun h anbenegle ted.
Further,the hargedistributionisassumedtobehomogeneousand themodulationofthe
urrentdensityisassumedtobesmall. Theseassumptionsleadtotheso- alledthird-order
equation and an besolved analyti ally.
Itisfurtherpossibletodedu eanumberofproperties,des ribingtheFELpro ess. The
typi al emitted FEL radiation as a fun tion of the undulator length is shown in gure
2.6. During the rst few meters the ele tron bun h has an initial urrent density. As
soon as the ele tron bun h be omes modulated and the mi robun h stru ture develops
the emitted FEL pulse energy in reases exponentially. The power of the light wave an
bedes ribed with the followingequation:
P(z)
∝
exp(z/L
g
)
withL
g
∝
λ
u
K
2
1/3
,
where
P(z)
isthepowerofthelightwaveandL
g
isthegainlengthwhi hin reasesforlower wavelengths [S hmüser et al., 2008℄. After approximately12
m themi robun hstru ture isfullydevelopedandsaturationsetsin. Nofurtherampli ationoftheemittedradiationis possible.
2.4.4 FEL operation modes
In the previous se tion the FEL-ampli ation of an ele tromagneti wave has been
Figure 2.6: The exponential growth of the FEL pulse energy as a fun tion of the length
traveled in the undulator. After a few gain lengths the exponential growth
andthe mi robun hingstarts. The ir les orrespondtomeasurementsofthe
TESLA Test Fa ility[Ayvazyan etal., 2002℄. The progressing mi robun hing
is indi ated s hemati ally. The graphi is taken from[S hmüser et al., 2008℄
with kindpermissionof Springer S ien e and Business Media.
FEL-operationmodes, whi hwill bedis ussed inthis hapter.
Self-amplied spontaneous emission (SASE)
The most ommonoperationmode of anFEL isSASE, whi h has been observed for the
rsttimeintheXUVregimeattheTESLAtestfa ilityinHamburgatDESY[Andruszkow
et al., 2000℄. This operation mode does not need an external laser seed, it rather uses
the spontaneous undulator radiation as a seed to initiate the ampli ation pro ess: the
ele tron bun h onsists of randomly distributed ele trons. This distribution leads to a
whitenoisespe trumwithintherstpartoftheundulator. Thespe tral omponentwhi h
is within the FEL bandwidth willa t asthe seed for the following ampli ationpro ess.
However, SASE is based ona sto hasti pro ess, whi h leads tou tuating properties of
the FEL radiationona shot-to-shotbasis. The spe trum and the pulse prole issubje t
to errati u tuations. Moreover, the pulse energy and the pulse lengthvaries fromshot
toshot. Nevertheless, itispossible todes ribe theresultingradiationinastatisti alway.
In [Saldin et al., 1998℄ a detailed study of statisti al properties of the radiation from a
SASE FEL is given.
Seeding s hemes
Dierentseedings hemeshavebeenproposed,inordertoredu etheseerrati u tuations
of the FEL radiation. The following is primarily addressed to dire t seeding with an
external laser although a brief summary of other seeding s hemes is given. A general
Dire t seeding with an external laser
Seeding an FEL with an external laser has been prosed in [Garzella et al., 2004℄. The
idea is toinitiate the ampli ation pro ess of a high-gain FEL with anexternal seed, in
order tosuppress the errati u tuations of the SASE radiation. Asexplainedin hapter
2.3 HHG is a feasible method to generate VUV toXUV radiation with a high degree of
temporaland spatial oheren e and istherefore aneligible andidate.
Dierent simulations have been published [Milt hev et al.,2008; M Neilet al., 2007a℄
showingthat seedingwithanexternallaserleadstoFELradiationwithimproved
tempo-ral oheren eandspe tralbrightness. Theshot-to-shotstabilitymay learlybeenhan ed
and the pulse width isdetermined by the pulse width of the seed. The simulationsshow
further that the ampli ation pro ess washes out the attose ond pulse-train stru ture
of the harmoni s and sele tively amplies a spe i harmoni , whi h is determined by
the FEL bandwidth. It is ne essary to ensure that the seed overlaps with the ele tron
bun hinspa e,time,wavelength, andpolarizationinordertoinitiatethe seedingpro ess
before the SASE signal develops. Therefore, the seed energy threshold depends on the
wavelength [Togashiet al., 2011℄and is assumed tobeinthe nanojoule rangefor photon
energies smaller than
100
eV. Moreover, the ampli ationpro ess takespla e within the pulse length of the seed, hen e the arrival time jitter of the ele trons aused by thea - eleration pro ess has noinuen e on the yielded FEL radiationas long as the temporal
jitter doesnot ex eed the ele tron bun h length.
So far dierent seeding experiments have been performed. The rst observation of a
seeded FEL has been reported atthe Deep Ultra-Violet FEL, where the third harmoni
from rystals
λ
seed
= 266
nm ledtoalargeampli ation[DiMauroetal.,2003℄. In Japan the SPring-8Compa t SASESour e(SCSS)test a eleratorhasbeen seeded withahigh-harmoni . Here su essful seeding has been reported for
λ
seed
= 160
nm andλ
seed
=
61.2
nm [Lambert et al.,2008; Togashi et al., 2011℄. At SPARC a seedingexperiment is ongoing,where rst results have been a hieved atλ
seed
= 260
nm [Giannessietal., 2008; T herbako et al.,2006℄and Max Labis preparing aseeding experiment as published in[uti¢ etal., 2010; Thorinet al., 2007℄.
It has been shown [Lambert et al.,2008;Togashi et al., 2011℄ that the intensity of the
seeded FEL radiationis enhan ed ompared to the SASE operation mode, although the
ratiobetweentheSASEandseedingasso iatedsignaldropsdownforsmallerwavelengths.
In additionit has been shown that the spiky stru ture of the spe trum in ase of SASE
is drasti allyredu ed by seeding the FEL.
Alternative seeding s hemes
High-gainharmoni generation(HGHG)isadierents hemeofseedinganFEL[Doyuran
et al., 2004; Yu et al., 2000; M Neil et al., 2007b℄. In this ase, a seed laser modulates
the ele tron bun h inarst undulator (modulator). Afollowingdispersivese tion
trans-fers the energy modulationintoa density modulationand ase ond undulator (radiator)
radiates higher harmoni s of the seed laser. In a se ond step the radiation passes a
a se ond modulator. The ele tron bun h travels then through another dispersive se tion
andanotherradiator,whi histunedtohigherharmoni s. In[Yuetal.,2003;Wangetal.,
2006;uti¢etal.,2011℄rstresultshavebeenpublished. Thiss heme analsobeusedto
establish a temporal overlap between the ele tron bun h and the seed laser aspublished
in [Tarkeshian et al., 2010℄.
Re ently, e ho-enabled harmoni generation has been proposed [Stupakov, 2010;
Xi-ang et al., 2010℄. Here a long wavelength laser modulates the ele tron bun h to a short
wavelength mi robun hing: the ele tron bun h ismodulatedby alaserwith awave
num-ber