You are on page 1of 8

PHYSICAL REVIEW A 77, 063830 共2008兲

Virtual dielectric waveguide mode description of a high-gain free-electron laser. I. Theory

Erik Hemsing,1 Avraham Gover,2 and James Rosenzweig1


1
Particle Beam Physics Laboratory, Department of Physics and Astronomy, University of California Los Angeles, Los Angeles,
California 90095, USA
2
Faculty of Engineering, Department of Physical Electronics, Tel-Aviv University, Ramat-Aviv 69978, Tel-Aviv, Israel
共Received 15 February 2008; published 20 June 2008兲
A set of mode-coupled excitation equations for the slowly growing amplitudes of dielectric waveguide
eigenmodes is derived as a description of the electromagnetic signal field of a high-gain free-electron laser
共FEL兲, including the effects of longitudinal space charge. This approach of describing the field basis set has
notable advantages for FEL analysis in providing an efficient characterization of eigenmodes, and in allowing
a clear connection to free-space propagation of the input 共seeding兲 and output radiation. The formulation
describes the entire evolution of the radiation wave through the linear gain regime, prior to the onset of
saturation, with arbitrary initial conditions. By virtue of the flexibility in the expansion basis, this technique can
be used to find the direct coupling and amplification of a particular mode. A simple transformation converts the
derived coupled differential excitation equations into a set of coupled algebraic equations and yields a matrix
determinant equation for the FEL eigenmodes. A quadratic index medium is used as a model dielectric
waveguide to obtain an expression for the predicted spot size of the dominant system eigenmode, in the
approximation that it is a single Gaussian mode.

DOI: 10.1103/PhysRevA.77.063830 PACS number共s兲: 42.25.Dd, 41.60.Cr, 42.60.Jf, 42.50.Tx

I. INTRODUCTION In general, the choice of analytical model and expansion


The optical guiding of light in free-electron lasers 共FELs兲 basis depends on the FEL geometry and operational charac-
is a well-known phenomena that results during amplification teristics. Since the fields inside the e-beam are optically
when the coherent interaction between the source electron guided during high-gain, a guided expansion basis is a natu-
beam 共e-beam兲 and the electromagnetic 共EM兲 field intro- ral choice, but may be plagued by inefficiency in describing
duces an inward curvature in the phase front of the light, the EM field structure if the guiding characteristics of the
refracting it back toward the lasing core of the e-beam 关1–3兴. e-beam are markedly different from those of the virtual
During the gain process the e-beam can behave similar to a waveguide that yields the particular basis set. 共The wave-
guiding structure that suppresses diffraction, reducing trans- guide is referred to as “virtual” because no such waveguide
is assumed to be present in the physical system.兲 In Ref. 关5兴
verse power losses, and enhancing the EM field amplifica-
an analysis is presented for calculating the gain-guided FEL
tion. For a sufficiently long interaction length, the guided
supermode by means of a field expansion using eigenmodes
EM field eventually dynamically settles into a propagating, of a hollow waveguide with conducting walls. This method
self-similar eigenmode of the FEL system, which, although can be useful at short wavelengths 共e.g., x-ray FEL兲, even
describing a system in which the field amplitude grows ex- when no such physical boundary exists and the optical beam
ponentially as a function of longitudinal coordinate z, has a is guided only by the e-beam. Clearly, in this example analy-
fixed form for the transverse profile and stationary spot size sis and in general, this approach is legitimate if the presence
关4,5兴. of the artificial boundary does not affect the physical result.
Several different approaches currently exist for describing The virtual waveguide dimensions must be taken wide
the FEL linear gain process. Analytic derivations of guided enough so that the fields in the actual problem vanish at the
FEL eigenmodes have been performed previously by direct virtual boundaries, but not so wide that many modes are
derivation of the eigenmode equations from the coupled required to describe the radiation field such that the calcula-
Maxwell-Vlasov equations 关4,6,7兴; through solutions to the tion is inefficient or does not converge. As an alternate ap-
Maxwell equations using Green’s functions 关8兴; or through proach described here, a general method that uses eigen-
expansion of the FEL signal field in terms of eigenmodes of modes of a smooth, transversely anisotropic dielectric
a hollow, conducting-boundary waveguide 关5兴, eigenmodes waveguide can be used to model the field while avoiding
of a step-index fiber 关3兴, and free-space paraxial waves 关9兴. both the undesirable influence of artificial boundary condi-
Such analytic descriptions of the three-dimensional FEL tions and precise a priori knowledge of the characteristic
equations, particularly those that exploit radiation mode ex- transverse dimensions of the waveguide features of the
pansions to find the FEL eigenmodes, have certain utility for e-beam. This approach also provides flexibility in the form
providing physical insight into the character of the radiation. of the expansion basis, which is determined by the refractive
Since, in a high-gain FEL, the e-beam that propagates index distribution. Of particular interest in any expansion
through the periodic undulator operates simultaneously as an mode approach is the coupling to the naturally occurring
optical source and as a waveguide structure, an expansion Hermite-Gaussian 共HG兲 or Laguerre-Gaussian 共LG兲 modes
mode description of the FEL light permits investigation of that describe free-space wave propagation in the paraxial
the amplification, guiding and coupling efficiency of the in- limit. This connection is useful for relating the FEL modes to
dividual basis modes to the e-beam. free-space propagating modes, which are present both before

1050-2947/2008/77共6兲/063830共8兲 063830-1 ©2008 The American Physical Society


HEMSING, GOVER, AND ROSENZWEIG PHYSICAL REVIEW A 77, 063830 共2008兲

and after the exponential gain process occurs in the undula- and e-beam propagation coordinate z. Neglecting backward
tor. With this connection one may, for example, compactly propagating waves 共which is well justified in an FEL兲, the
and clearly describe the input radiaiton needed for efficiently general field expansion is
seeding the FEL. One may also robustly describe the propa-
gation of the FEL light after saturation, allowing a clear un- Ẽគ ⬜共rគ 兲 = 兺 Cq共z兲Eគ̃ ⬜q共r⬜兲eikzqz ,
derstanding of the demands made on downstream optics, and q

on expected mode structure at FEL diagnostics or experi-


ments themselves. In the virtual dielectric waveguide expan- គ 共rគ 兲 = 兺 Cq共z兲H
H̃ គ̃ ⬜q共r⬜兲eikzqz 共1兲
sion, a form of the refractive index that varies quadratically q
in the transverse dimension yields a basis of guided LG or
HG paraxial waves. This facilitates the desired natural de- where H គ̃ ⬜q = 共1 / Zq兲êz ⫻ Eគ̃ ⬜q is the magnetic field expansion
scription of the FEL radiation, and is the motivation of the mode, kzq is the qth mode axial wave number and the imped-
present work. ance is Zq = 共k / kzq兲冑␮0 / ⑀0 for TE modes. In general, the sum
Optical guiding in an FEL occurs in the exponential gain extends over both the guided and the cutoff modes, and the
regime, when the focusing effects in the source e-beam tend modes form a complete set 关10兴. The expansion modes are
to balance the natural diffraction of the generated EM radia- orthogonal and normalized to the mode power

冕冕
tion. Obviously, an expansion set consisting of guided modes
1 ⴱ
can be used to accurately describe the field during long sec- Pq␦q,q⬘ = Re关 共Eគ̃ ⬜q共r⬜兲 ⫻ H
គ̃ ⬜q⬘共r⬜兲兲 · êzd2r⬜兴.
tions of gain-guiding, over which the otherwise free-space 2
fields have many Rayleigh lengths to diffract. Such an ex- 共2兲
pansion can also accurately describe short sections of dy-
namic profile evolution 共such as the FEL start-up period in a In free-space, the equation for the full time-harmonic fields
high-gain amplifier兲. But guided modes may not efficiently in the presence of a source is
capture 共in the sense of providing a compact description兲 the
field behavior during long sections of weak guiding when ⵜ2Ẽគ + k2Ẽគ = − i␻␮0J̃គ , 共3兲
diffraction is dominant, as during saturation or low-gain. The
where the total field is Eគ 共r , t兲 = Re关Ẽគ 共r兲e−i␻t兴 and k = ␻ / c is
regimes of validity for a description using guided modes may
be estimated by inspection of the relative values of the opti- the free-space wave number. The transverse charge density
cal Rayleigh length zR = kr20 共assuming for the moment that gradient of the e-beam in an FEL is typically small and is
the field has the same characteristic radius as the e-beam, r0兲, neglected. For a given choice of expansion basis, the field
the e-folding radiation power gain length LG, and the overall expansions in Eq. 共1兲 can be inserted into Eq. 共3兲 to obtain a
length of the FEL interaction length L. In the guided mode differential equation for the amplitude coefficents Cq共z兲 关5兴.
expansion here, we restrict our attention to high-gain 共LG Here, we explore the case where the expansion mode Ẽ⬜q is
Ⰶ L兲 FEL systems, prior to the onset of saturation, for which an eigenmode of a dielectric waveguide with refractive index
the weakly diffracting 共LG Ⰶ L ⬍ zR兲 or strong-guiding 共LG n共r⬜兲. Assuming small transverse variation ⵜn共r⬜兲2 Ⰶ k, the
⬍ zR ⬍ L兲 conditions are valid, and the guided mode expan- eigenmodes can be regarded as dominantly transverse, and
sion description can be used efficiently. the dielectric eigenmode equation is
In this work, a virtual dielectric expansion description of
high-gain FEL radiation is presented. The excitation of the ⵜ⬜
2
Eគ̃ ⬜q共r⬜兲 + 关n共r⬜兲2k2 − kzq
2
兴Eគ̃ ⬜q共r⬜兲 = 0. 共4兲
slowly-growing mode amplitude coefficients in the presence The use of dielectric eigenmodes as an expansion basis must
of a local source current is derived. The e-beam is linearized be accompanied by additional terms in Eq. 共3兲. These terms
in the cold-beam fluid approximation and a set of coupled are included to offset the virtual polarization currents that
excitation equations, modified from Ref. 关5兴, is obtained. The arise from the refractive index of the virtual waveguide,
coupled equations are then recast by a simple transformation since no such waveguide exists along the e-beam axis in the
into a matrix equation for solutions of the FEL supermode as
a function of parameters for energy detuning, longitudinal physical system. The polarization current is given by Jគ̃ p共rគ 兲
space-charge effects and mode coupling to the e-beam. The = i␻⑀0␹共r⬜兲Ẽគ 共r⬜兲 = i␻⑀0关n共r⬜兲2 − 1兴Ẽគ 共r⬜兲 where n共r⬜兲
theory presented here provides a framework for detailed nu- = 冑⑀共r⬜兲 / ⑀0. Mathematically, this process amounts to adding
merical studies performed in a follow-up paper 共paper II兲. 关n共r⬜兲2 − 1兴k2Ẽគ 共r⬜兲 to both sides of Eq. 共3兲. From the dielec-
There, results that highlight the advantages of the dielectric tric eigenmode Eq. 共4兲, the fields are assumed to be domi-
eigenmode expansion approach are presented and analyzed. nantly transverse, and Eq. 共3兲 becomes

II. DIELECTRIC WAVEGUIDE EXPANSION MODES IN ⵜ2Ẽគ ⬜ + n共r⬜兲2k2Ẽគ ⬜ = − i␻␮0J̃គ ⬜ + 关n共r⬜兲2 − 1兴k2Ẽគ 共r⬜兲.
THE PRESENCE OF A LOCAL SOURCE 共5兲
The radiation fields Eគ 共r , t兲 and Hគ 共r , t兲 emitted by the cur- Plugging in the expansion fields from Eq. 共1兲, the dielectric
rents in the FEL can be expanded in terms of transverse eigenmode equation in Eq. 共4兲 is used to eliminate the trans-
radiation modes 兵Eគ̃ ⬜ , H
គ̃ ⬜其 of a guiding structure, with slowly verse Laplacian term, and the excitation equation for the
growing amplitudes that vary only as a function of the axis mode q in the presence of a local source current is given by

063830-2
VIRTUAL DIELECTRIC WAVEGUIDE MODE ... . I. THEORY PHYSICAL REVIEW A 77, 063830 共2008兲

d
dz
Cq共z兲 = −
1 −ik z
4Pq
e zq 冕冕 ⴱ
J̃គ ⬜共rគ 兲 · Eគ̃ ⬜q 共r⬜兲d2r⬜
netic fields results in an axial ponderomotive field 共ṽគ ⫻ B̃គ 兲z

= 兺qCq共z兲Ẽ pm,q共r⬜兲ei共kzq+kw兲z where Ẽ pm,q共r⬜兲 = 21 关ṽគ ⬜q ⫻ Bគ̃ ⬜w

− i 兺 ␬q,q⬘Cq⬘共z兲e−i⌬kzqq⬘z ,
d
共6兲 + ṽគ ⬜w ⫻ Bគ̃ ⬜q兴 · êz, B̃⬜w is the transverse magnetic field of the
q⬘
undulator, and ṽ⬜q共r⬜兲 is the transverse electron velocity due
to the Lorentz force of the qth mode of the signal field. The
where undulator wave number is identified as kw = 2␲ / ␭w, and the

冕冕
transverse velocity due to the magnetic undulator field is
d ␻⑀0 ⴱ
␬q,q ⬘
= 关n共r⬜兲2 − 1兴Eគ̃ ⬜q⬘共r⬜兲 · Eគ̃ ⬜q 共r⬜兲d2r⬜ . ṽគ ⬜w = 共−icK / ␥兲êz ⫻ êw, where K = e兩B̃⬜w兩 / mckw is the undu-
4Pq
lator parameter and êw = Bគ̃ ⬜w / 兩Bគ̃ ⬜w兩 is the unit vector of the
共7兲 undulator field.
The difference between the axial wave numbers of the modes By combining Eqs. 共8兲 and 共9兲 and using the definition of
is ⌬kzqq⬘ = kzq − kzq⬘. The term ␬q,q
d
characterizes the mode the modulated longitudinal current density, the density
⬘ bunching evolution can be expressed as a second order dif-
overlap in the dielectric, and physically represents the virtual
ferential equation. We obtain the result from Ref. 关5兴 with
polarization currents that are necessarily subtracted when us-
transverse fields for the density modulation evolution during
ing eigenmodes of a virtual dielectric waveguide.
the FEL interaction

III. ELECTRON-BEAM FLUID AND COUPLED


EXCITATION EQUATIONS

A linear plasma fluid model for a cold e-beam 共negligible


冋 d2
dz2
− 2i
␻ d ␻ p0 f共r⬜兲 − ␻
vz0 dz
+
2

vz2
2

ñ1共rគ 兲
0

␻2p0 f共r⬜兲 ⑀0
energy spread兲 can be used to describe the signal excitation
in an FEL interaction 关5兴. A relativistic e-beam in an FEL
=i
vz20 e
兺q 共kzq + kw兲Cq共z兲Ẽpm,q共r⬜兲ei共k zq+kw兲z ,
experiences transverse oscillations, or wiggling, driven by an
interaction with a periodic structure or undulator. This peri- 共10兲
odic motion drives an axial ponderomotive force that modu-
where the longitudinal relativistic plasma frequency on axis
lates the axial electron velocity such that, to first-order, the
is ␻2p = e2n0 / ␥␥z2⑀0m. The beam current is I0 =
axial velocity of a cold beam within a static undulator can be 0

expanded as vz共rគ , t兲 = vz0 + Re关ṽz1共rគ 兲e−i␻t兴 where vz0 = ␤zc is −evz0n0兰兰f共r⬜兲d2r⬜ and the effective beam area through the
the dc component and ṽz1 is the modulation at signal fre- normalization condition 兰兰f共r⬜兲d2r⬜ = Ae. For a uniform
cross-sectional distribution of the e-beam, f共r⬜兲 = 1 for r
quency ␻. Longitudinal variations in the velocity like those
ⱕ r0 and zero otherwise. For a Gaussian distribution f共r⬜兲
found in planar undulator systems are ignored for the mo-
= exp共−r2 / r20兲.
ment. The velocity modulation ṽz1 develops a density bunch-
The transverse component of the current density that ex-
ing modulation that is similarly described in a linear model cites the signal wave is written in terms of the density modu-
as n共rគ , t兲 = n0 f共r⬜兲 + Re关ñ1共rគ 兲e−i␻t兴 where n0 is the on-axis lation as
electron density and f共r⬜兲 is the transverse density profile of
the e-beam. The lowest order ac component of the longitu- 1
dinal current density results from both the axial velocity and J̃គ ⬜共rគ 兲 = − eñ1共rគ 兲ṽគ ⬜we−ikwz . 共11兲
2
density modulations and is J̃z共rគ 兲 = −e关n0 f共r⬜兲ṽz1共rគ 兲
+ vz0ñ1共rគ 兲兴. If the transverse divergence of the current density The charge density modulation ñ1共rគ 兲 appears both in the field
mode excitation Eq. 共6兲 and in the density modulation evo-
modulation is assumed small ⵜ⬜ · J̃⬜ Ⰶ ⳵J̃z / ⳵z, the continuity
lution Eq. 共10兲. These equations are coupled through the
equation can be written as
ponderomotive field, which illustrates the relationship be-
d tween the density modulation and the excited signal field in
J̃z = − i␻eñ1共rគ 兲. 共8兲 an FEL. Both equations can be simplified by expressing the
dz
density modulation as a sum over the expansion basis func-
The relativistic force equation for the axial velocity modula- tions and slowly varying amplitudes
tion is
k⑀0
d ␻ e ñ1共rគ 兲 = 兺 Bq共z兲Ẽ⬜q共r⬜兲ei␻/vz0z . 共12兲
ṽz1共rគ 兲 − i ṽz1共rគ 兲 = − 2 关ẼSC + 共ṽគ ⫻ B̃គ 兲z兴 共9兲 e q
dz v z0 ␥␥z mvz0 z
Plugging this expression into Eqs. 共11兲, 共10兲, and 共6兲, and
where the axial space-charge field ẼzSC = J̃z / i␻⑀0 is due to the integrating over the transverse dimensions in Eq. 共10兲, the
current density modulation. It is assumed that the e-beam orthogonality of the eigenfunctions can be used to simplify
radius is large compared to the bunching wavelength in the the equations and to write both coupled equations in terms of
e-beam frame r0 ⬎ ␭␥z, so the fringing fields and the trans- the slowly growing amplitudes Cq共z兲 and Bq共z兲 of the signal
verse space-charge effects are neglected. The interaction be- field and the density modulation, respectively. This yields the
tween the transverse electron motion and the transverse mag- coupled FEL excitation equations

063830-3
HEMSING, GOVER, AND ROSENZWEIG PHYSICAL REVIEW A 77, 063830 共2008兲

d sion from Eq. 共12兲, the integration over the transverse space
Cq共z兲 = − i␰qĝⴱqBq共z兲ei␪qz − i 兺 ␬q,q⬘Cq⬘共z兲ei共␪q⬘−␪q兲z ,
d
dz is straightforward by the orthogonality of the eigenmodes.
q ⬘ Thus, a simplified expression is obtained for ĩq共z兲 in terms of
the density modulation amplitudes
d2 1
2 Bq共z兲 + ␪ p 兺 Fq,jB j共z兲 = − ⴱ 兺 Qq,q⬘Cq⬘共z兲e
2 −i␪q⬘z
,
dz j ĝ ␰
q q q⬘ ĩq共z兲 = − i␰qĝⴱqBq共z兲. 共18兲
共13兲 Substitution of this into Eq. 共13兲 allows one to solve the
where ␪q = ␻ / vz0 − 共kzq + kw兲 is the characteristic detuning pa- coupled FEL excitation equations in terms of the optical cur-
rameter for a given mode q, ␰q = Kk2 / 4␥kzq, and ␪ p rent amplitudes.
= ␻ p0 / vz0 is the longitudinal plasma wave number of a uni- In the 1D limit, the axial wave numbers are degenerate
formly distributed electron beam profile used in a 1D model. kzq = k, the e-beam profile is constant f共r⬜兲 → 1 and the over-
The factor ĝⴱq = 共êz ⫻ êw兲 · êⴱq is the polarization alignment fac- lap factor becomes Fq,q⬘ = ␦q,q⬘. The e-beam coupling coeffi-
tor and measures the relative alignment of the transverse cient is then written in 1D simplified form Q
electron motion in the undulator 共êz ⫻ êw兲 with the 共complex- = JJk共␪ pK / 2冑2␥␤z兲2 = 共2kw␳兲3, where ␳
conjugated兲 electric-field mode polarization direction 共êⴱq兲. = 共JJe2K2n0 / 32⑀0␥3mc2kw2 兲1/3 is the well-known Pierce pa-
When the wiggling motion direction matches the mode po- rameter often used in FEL theory.
larization 共which may not be the case for the input mode in a The first of the coupled equations in Eq. 共13兲 describes
seeded FEL scenario兲 ĝq = 1. the excitation of the mode amplitude Cq of a dielectric wave-
The coupling between the e-beam and the signal field is guide eigenmode due to the transverse wiggling motion of
given by the mode coupling coefficient the bunching current throughout the FEL interaction. The
second equation in Eq. 共13兲 describes the evolution of the
Qq,q⬘ = JJ
␪2p⑀0
8Pq
共kzq⬘ + kw兲 冕冕 f共r⬜兲Ẽ pm,q⬘共r⬜兲
density bunching amplitude, which is excited in the e-beam
by the EM signal field. The effect of the longitudinal space-
charge in the beam is calculated in the second term of the

⫻ṽគ ⬜w · Eគ̃ ⬜q 共r⬜兲d2r⬜ , 共14兲 second equation, and takes into account the effects of longi-
tudinal plasma oscillations 共Langmuir waves兲 in the FEL in-
where JJ = 关J0共␣兲 − J1共␣兲兴2 is now included for a strong pla- teraction. The effects of fringing fields due to a transverse
nar undulator 共JJ = 1 for a helical undulator geometry兲. J0 variation in the axial space-charge field can be examined in a
and J1 are the first and second order Bessel functions and more complete 3D scenario 关11兴, but are presently neglected
␣ = K2 / 共4 + 2K2兲. The ponderomotive field Ẽ pm,q⬘ is evaluated in the approximation that r0 ⬎ ␭␥z.
explicitly for TE modes in the Appendix, and we obtain a The initial conditions for Eqs. 共13兲 specify the operating
simplified form characteristic of the FEL. For example, when operating as a
single-pass amplifier 共seeded FEL兲 there is negligible initial
Qq,q⬘ = JJ␪2p
共kzq⬘ + kw兲2 K
8kzq ␥
冉冊 2
ĝⴱqĝq⬘Fq,q⬘ , 共15兲
density and velocity modulation 关兩Bq共0兲 , dBq共z兲 / dz兩z=0 = 0兴
and the initial seed field is nonzero 关Cq共0兲 ⫽ 0兴. For a self-
amplified spontaneous emission 共SASE兲 FEL, the amplified
where Fq,q⬘ is referred to as the beam profile overlap coeffi- shot noise can be related to the prebunching conditions
cient and quantifies the spatial overlap of the e-beam profile 关兩Bq共0兲 ⫽ 0 , dBq共z兲 / dz兩z=0 ⫽ 0兴 and the input signal field van-
with the expansion modes q , q⬘: ishes 关Cq共0兲 = 0兴.

冕冕
The coupled expressions in Eq. 共13兲 describe the evolu-

f共r⬜兲Ẽ⬜q⬘共r⬜兲Ẽ⬜q 共r⬜兲d2r⬜ tion of the e-beam modulation currents and the excitation of
共16兲 the EM signal field during the FEL interaction, inclusive of

冕冕
Fq,q⬘ = .
longitudinal space-charge effects. During high-gain opera-
兩Ẽ⬜q共r⬜兲兩 d r⬜
2 2
tions when the radiation field grows exponentially, the self-
guiding effects become dominant over the diffraction, and
This coefficient also appears in connection with the longitu- the radiation field can be accurately described by a collection
dinal plasma wave dynamics left-hand side of the second of waveguide modes. The virtual dielectric eigenmode basis
equation in Eq. 共13兲. used for the radiation field expansion can be any complete
It also may be illuminating to also define the current set that satisfies the dielectric waveguide Eq. 共4兲. In the gen-
bunching amplitude eral expansion presented here, the form for n共r⬜兲 can be

冕冕
freely chosen to yield different functional forms for the ex-
1 −i共␻/v −k 兲z ⴱ
ĩq共z兲 = − e z0 w J̃គ ⬜共rគ 兲 · Eគ̃ ⬜q 共r⬜兲d2r⬜ pansion modes, allowing flexibility to choose a basis that is
4Pq perhaps better suited to describe a given FEL system. Some
共17兲 of the most typical forms include an index with a quadratic
spatial dependence which results in Hermite-Gaussian or
which is interpreted as the slowly varying amplitude of the Laguerre-Gaussian expansion modes 关12兴 共examined in de-
transverse optical current. Replacing J̃គ ⬜ with its definition tail for LG modes in paper II兲, or a step-profile optical fiber
from Eq. 共11兲 and plugging in the density modulation expan- 关3,10兴 which yields Bessel functions. The closer that the

063830-4
VIRTUAL DIELECTRIC WAVEGUIDE MODE ... . I. THEORY PHYSICAL REVIEW A 77, 063830 共2008兲

choice of the virtual dielectric waveguide distribution is to 共˜ ␦k − ⌬kq兲bq + 兺 ␬q,q


␦k − ␪兲2关共˜ d
b 兴
⬘ q⬘
the real solution to the given FEL system, the fewer the

再 冎
q⬘
number of modes that will be required to converge the equa-
tions in Eqs. 共13兲 to the correct solution for the FEL. kzj
− ␪2p 兺 ␦k − ⌬k j兲b j + 兺 ␬dj,q⬙bq⬙兴
Fq,j关共˜
kzq
j ⬙
q

IV. SUPERMODE MATRIX SOLUTIONS + 兺 Qq,q⵮bq⵮ = 0, 共21兲


q⵮
In the high-gain regime of an FEL with finite e-beam
where ␪ = ␻ / vz0 − k − kw is the detuning parameter for a 1D
width, the radiation field tends to concentrate near the beam
and produce a power amplified radiation wave with a self- model 共kzq = k兲, and ⌬kq = kzq − k. The coupled equations in
similar transverse field distribution that propagates along the Eq. 共21兲 can be solved to yield values for the supermode
interaction length. This specific complex-valued combination coefficients bq in terms of the perturbation ˜ ␦k. It is conve-
of the expansion modes is referred to as the supermode, or nient to write Eq. 共21兲 in a simplified matrix determinant
the eigenmode of the high-gain FEL. In a long undulator, the form
fundamental supermode evolves spontaneously since it has
the highest gain. The initial conditions 共i.e., the transverse ␦k − ␪兲2 − ␪2pM
兩关Iគគ 共˜ ␦k + ␬គ d − ⌬k
គ 兴关Iគ˜ គគ 兩 = 0.
=兴 + Q 共22兲
profile and phase distribution of the injected radiation field or
density prebunching兲 affect the mode excitation composition The matrix elements of M គ are given by Mq,q⬘
and evolution. However, these initial conditions only affect = 共kzq⬘ / kzq兲Fq,q⬘ and similarly for ␬គ d = 兵␬q,qd
其, Q = 兵Qq,q⬘其 and
⬘ គគ
the supermode establishment length. Evenutally, in an unde- = = 兵⌬kq␦q,q⬘其 . The matrix Iគគ is the identity.
⌬k
pleted system, the supermode will become dominant in a The solutions to Eq. 共22兲 yield 3N solutions for˜ ␦k, where
long enough undulator, prior to the onset of saturation.
To find the optimal injection parameters that match and N is the number of expansion modes. Each ˜ ␦k can then be
expedite the establishment of the supermode, one must solve inserted in Eq. 共21兲 to find a nontrivial solution 共if one ex-
Eqs. 共13兲 with the appropriate initial conditions. To simply ists兲 for the mode amplitude coefficients of an eigenmode of
find the system supermodes, however, it is enough to find the the FEL system. From Eq. 共20兲 and the definition of kSM in
eigensolutions to Eq. 共13兲. These are the combinations of the Eq. 共19兲, it can be seen that the solution for˜ ␦k with the most
expansion mode profiles that propagate self-similarly, i.e., negative imaginary component drives the highest gain, and
with constant amplitude coefficients and with distinct com- dominates over the rest of the eigenmodes. This value is used
plex wave numbers 关8兴. In the presence of gain, each super- in solving Eq. 共21兲 and will yield the coefficients of the
mode wave number will be different from the wave number dominant supermode, with the 3D power gain length, or
of free-space and can be written with a perturbative term ˜ ␦k e-folding length, given by LG = 共2兩Im兵˜ ␦k其兩兲−1.
that is due to the FEL interaction We note that when ⌬k = , ␬គគ d = 0គគ , the matrix equation in Eq.
共22兲 reduces to a generalized matrix form of the canonical
FEL cubic equation with longitudinal space charge. In the
␦k ,
kSM = k +˜ 共19兲 additional limit of a large, effectively constant transverse
beam profile f共r⬜兲 → 1 so Fq,q⬘ = ␦q,q⬘ and Qq,q⬘ → Q␦q,q⬘, and
Eq. 共22兲 takes the form of the familiar 1D FEL cubic equa-
where Re兵˜ ␦k其 anticipates an effective modified refractive in- tion with the mode-independent beam coupling parameter Q
dex to that of free-space, and Im兵˜ ␦k其 is related to the gain. and the axial plasma wave number ␪ p.
Since the supermodes evolve after the initial startup period
and have fixed transverse profiles along z, one can substitute V. SINGLE GUIDED GAUSSIAN MODE APPROXIMATION
Cq共z兲 = bqei共kSM−kzq兲z for the mode amplitude coefficients Eq.
共1兲. The mode amplitude coefficients bq are constants, and During high-gain, the proper balance between the natural
the z dependence is contained solely in the mode- diffraction of the coherent radiation and the guided focusing
independent exponential term. The time-harmonic electric of the radiation due to the e-beam determines the eventual
field of the supermode is then spot size wSM of the EM supermode field. This can be ob-
tained with the dielectric expansion formulation in a natural
way, through the solutions to the excitation Eq. 共13兲 or the
Ẽគ SM共rគ 兲 = 关 兺 bqEគ̃ q共r⬜兲兴eikSMz . 共20兲 supermode determinant Eq. 共22兲, using N modes in the field
q expansion. Clearly, choosing a suitable expansion basis 共i.e.,
one that is close in form to that of the supermode field兲 can
greatly reduce the number of expansion modes required to
This describes a transverse field that is fixed in transverse precisely describe the FEL supermode. In some cases, it may
profile, but is growing exponentially in amplitude along z. even be sufficient to use only a single mode to model the
Inserting this transformation into Eq. 共13兲 converts the field, in order to estimate the relevant supermode character-
coupled second-order differential equations into a set of istics without solving the full equations. A single mode ap-
coupled algebraic equations proximation can also be used to streamline the computation

063830-5
HEMSING, GOVER, AND ROSENZWEIG PHYSICAL REVIEW A 77, 063830 共2008兲

of the full solutions by providing an approximate value for 1 − i冑3 1/3


the expansion basis waist size w0 which is used as a scaling ␦k = Q0,0 . 共28兲
2
parameter in the basis eigenfunctions Ẽq共r⬜ ; w0兲. The super-
mode waist size 共and thus the parameter w0兲 can sometimes In a high-gain FEL, the supermode wave number kSM in Eq.
be roughly approximated as equal to the e-beam radius 共19兲 is defined as the wave number of a plane wave k that is
wSM ⬃ r0 for an axisymmetric e-beam distribution, but the modified through the FEL interaction by the perturbative fac-
actual value may vary significantly depending on the operat- ␦k. Combining the definition of kSM in Eq. 共19兲 with ␦k
tor ˜
ing parameters of the FEL. To better quantify this estimate from Eq. 共26兲 we can write
for the general case, we analyze the guided supermode spot
size by approximating it as a single Gaussian mode 共SGM兲 1
that propagates with a fixed spot size wg. Gaussian modes, kSM = k + ␦k − 共29兲
kw2g
such as the HG and LG modes discussed previously, are
found as solutions to Eq. 共4兲 with a refractive index n共r兲2 that for a single Gaussian supermode. Since the supermode is
varies quadratically in the transverse dimension, describing assumed to propagate as a guided Gaussian mode with a
what is known as a weakly guiding quadratic index medium fixed spot size and velocity, the real 共propagating兲 part of the
or QIM. In this context, the SGM analysis also provides supermode wave number is equated to the propagating
insight into the physical origin of the characteristic guiding Gaussian mode wave number in the dielectric waveguide
of the supermode in terms of guided paraxial Gaussian
modes from conventional fiber waveguide theory. Re兵kSM其 = kz0 . 共30兲
The refractive index that yields axisymmetric guided Combining the real parts of Eqs. 共28兲–共30兲, and recognizing
Gaussian eigenmodes in Eq. 共4兲 is n0 = 1 + Q1/3 / 2k ⯝ 1 as the effective refractive index on axis

n2共r兲 = n20 − 冉冊r


zR
2
, 共23兲
from a 1D FEL model, we obtain a relation for the super-
mode spot size in terms of the 1D coupling and the filling
factor
where zR = kw2g / 2 is the Rayleigh length and wg is the char-
acteristic r.m.s. radius of the Gaussian field profile given by 1 Q1/3
= 共1 − F0,0
1/3
兲. 共31兲
Ẽq共r⬜ ; wg兲. The axial wave number, in the paraxial approxi- kw2g 2
mation is therefore given by
Assuming that the e-beam has a Gaussian transverse profile
2 f共r⬜兲 = exp共−r2 / r20兲, the filling factor is simply F0,0
kz0 ⯝ kn0 − . 共24兲 = 共1 + w2g / 2r20兲−1. We finally obtain an expression for the su-
n0kw2g
permode spot size wg in terms of the 1D coupling parameter
A single Gaussian mode then corresponds to only the Q and the e-beam radius r0:
共q , q⬘兲 = 共0 , 0兲 matrix elements in the supermode determinant
in Eq. 共22兲. For n0 ⯝ 1, an explicit analytic expression is
obtained for the dielectric mode coupling parameter ␬0,0
−1 / kwg. Ignoring the effects of space-charge waves 共␪ pL
2
d
= 冉 1+
w2g
2r20
冊冋 1−
2
kw2gQ1/3
册 3
=1 共32兲

⬍ ␲兲, the form of Eq. 共22兲 for a single Gaussian mode is Solutions to Eq. 共32兲 for wg can be easily obtained numeri-

共˜ 冋
␦k − ␪兲2 ˜
1

␦k + 2 + Q0,0 = 0,
kwg
共25兲
cally for given FEL running parameters. The calculated SGM
spot size provides a useful prediction of the supermode spot
size wSM for a Gaussian e-beam lasing at the fundamental
where ⌬k0 ⯝ −2 / kw2g has been obtained from Eq. 共24兲. Equa- and which in turn suggests an appropriate value for the ex-
tion 共25兲 then becomes a modified 1D FEL cubic equation pansion waist w0 to facilitate computational efficiency for
where Q0,0 = Q F0,0 is the SGM gain parameter, which de- solving the full excitation equations.
scribes the modification of the 1D e-beam mode coupling It is useful to note that while kSM was defined as a modi-
parameter Q by the filling factor 关Eq. 共16兲兴 of the fundamen- fication of a plane wave in Eq. 共19兲, in the SGM approxima-
tal Gaussian EM mode F0,0. One can then define a shifted tion, one could also consider defining kSM through a modifi-
perturbation parameter cation of a free-space paraxial Gaussian mode. In other
words, to consider the effect of the FEL interaction on con-
1 verting a free-space Gaussian wave to a guided Gaussian
␦k =˜
␦k + 共26兲
mode of the FEL.
kw2g
Free-space modes evaluated at the optical beam waist w0
and then set the detuning to ␪ = −1 / kw2g. Equation 共25兲 then have an associated axial wave number
becomes a simple 1D FEL cubic equation at resonance,
2
␦k3 + Q0,0 = 0. 共27兲 2
k̂z0 = k2 − . 共33兲
w20
In this form, the solution for ␦k that corresponds to the domi-
nant high-gain mode is straightforward, and well known Expressing the supermode wave number as a modified
from FEL theory: paraxial wave of free-space, it takes the form

063830-6
VIRTUAL DIELECTRIC WAVEGUIDE MODE ... . I. THEORY PHYSICAL REVIEW A 77, 063830 共2008兲

kSM = k̂z0 + ␦k, 共34兲 ṽគ ⬜w = − i


e
ê ⫻ Bគ̃ w ,
␥mkw z
where, just as before, the supermode is a guided eigenmode
of the QIM Re兵kSM其 = kz0 but it is now also expressed as a
e
deviation from the Gaussian free-space mode k̂z0 by the FEL ṽគ ⬜q共r⬜兲 = − i 关Eគ̃ 共r 兲 + vz0êz ⫻ Bគ̃ ⬜q共r⬜兲兴
␥m共␻ − kzqvz0兲 ⬜q ⬜
interaction given by ␦k. Equating the real parts from Eqs.
共19兲 and 共34兲, we find that k + Re兵˜␦k其 = k̂z0 + Re兵␦k其. By rear- e
=−i Eគ̃ 共r 兲, 共A2兲
ranging and using k̂z0 − k ⯝ −1 / kw2g for a single free-space ␥m␻ ⬜q ⬜
mode, we recover the exact expression for ␦k in Eq. 共26兲.
Thus, the SGM approximation of an FEL with a Gaussian where in the last equality for ṽគ ⬜q it has been assumed that
transverse e-beam profile describes the conversion of plane Eគ̃ ⬜q = −共␻ / kzq兲êz ⫻ Bគ̃ ⬜q for TE modes, in keeping with the
wave into a guided Gaussian mode at a detuning of ␪ = dominantly TEM mode expansion sets of the present work.
−1 / kw2g in an equivalent way as it describes the conversion
of a free-space paraxial Gaussian mode into a guided Gauss- With the equations in Eq. 共A2兲 the expression for Ẽ pm,q be-
ian mode, operating at resonance 共␪ = 0兲. comes

e共kzq + kw兲
Ẽ pm,q共r⬜兲 = i 共êz ⫻ Bគ̃ w兲ⴱ · Eគ̃ ⬜q共r⬜兲. 共A3兲
VI. CONCLUSIONS 2␥m␻kw
The signal field of a high-gain FEL has been described in The complex phasor notation for the undulator field Bw
terms of a sum over orthogonal eigenmodes of a virtual di-
electric waveguide. A set of coupled excitation equations for = Re兵Bគ̃ we−ikwz其 allows a general description for any type of
the e-beam density modulations and the field amplitudes in undulator polarization. Defining a general transverse field
the presence of longitudinal space-charge effects has been polarization unit vector as êw = Bគ̃ w / 兩B̃w兩, for a linear undula-
derived. This approach can be effectively used to predict tor êw = êy 共with vertical polarization兲, the field is given as
relevant FEL parameters such as signal field spot size and
intensity distribution for any point along the interaction Bw = 兩B̃w兩êy cos kwz 共A4兲
length, and for any arbitrary initial conditions. It further pro-
vides a novel and practical method to analyze the coupling and in a helical undulator êw = 共êx ⫾ iêy兲 / 冑2 for right-handed
and generation of particular modes, such as those associated 共+兲 or left-handed 共−兲 field orientations and
with free-space propagation, in the FEL system. It is a gen-
eral method for calculating the optically guided FEL eigen-
modes for any transverse e-beam current and profile distri- 兩B̃w兩
Bw =
bution, based on a straightforward numerical solution to an 冑2 共êx cos kwz ⫾ êy sin kwz兲. 共A5兲
algebraic dispersion equation. A follow-up analysis of this
work will explore these concepts in detail.
The undulator parameter is defined as K = e兩B̃w兩 / 共kwmc兲.
Note that the same K value corresponds to an on-axis maxi-
ACKNOWLEDGMENTS mum field amplitude of 兩B̃w兩 in a linear undulator, and an
on-axis maximum field amplitude of 兩B̃w兩 / 冑2 in a helical
The authors would like to thank Sven Reiche and Lea undulator. In terms of K, Eq. 共A3兲 can be written as
Fredrickson for helpful discussions. This research was sup-
ported by grants from Department of Energy Basic Energy
kzq + kw K
Science Contract No. DOE DE-FG02-07ER46272 and Office Ẽ pm,q共r⬜兲 = i 共êz ⫻ êwⴱ 兲 · Eគ̃ ⬜q共r⬜兲. 共A6兲
of Naval Research Contract No. ONR N00014-06-1-0925. 2k ␥

Substitution of this expression and the mode power normal-


APPENDIX: e-BEAM COUPLING COEFFICIENT ization from Eq. 共2兲 into Qq,q⬘ results in a general expression
for the mode coupling coefficients:
A generalized, explicit expression for the mode coupling
coefficient Qq,q⬘ of any spatial expansion basis can be ob-
tained starting from the ponderomotive field Qq,q⬘ = JJ␪2p
8kzq
冉冊
共kzq⬘ + kw兲2 K

2
ĝⴱqĝq⬘Fq,q⬘ , 共A7兲

1 ⴱ
Ẽ pm,q共r⬜兲 = 关ṽគ ⬜q ⫻ Bគ̃ ⬜w ⴱ
+ ṽគ ⬜w ⫻ Bគ̃ ⬜q兴 · êz . 共A1兲 where ĝq = 共êz ⫻ êwⴱ 兲 · êq is the polarization alignment factor,
2 which measures the relative direction of transverse electron
motion in the undulator compared with the electric field po-
From the relativistic Lorentz force equation we can write the larization of the mode q. The spatial overlap factor Fq,q⬘ is
transverse velocities as 关11兴 defined as

063830-7
HEMSING, GOVER, AND ROSENZWEIG PHYSICAL REVIEW A 77, 063830 共2008兲

冕冕 ⴱ
f共r⬜兲Ẽ⬜q⬘共r⬜兲Ẽ⬜q 共r⬜兲d2r⬜
共A8兲
and in their polarization relative to the undulator. It is also
clear that in order to obtain maximal coupling between two
modes q and q⬘, they must be polarization matched to each
冕冕
Fq,q⬘ = .
兩Ẽ⬜q共r⬜兲兩 d r⬜
2 2 other 共ĝq = ĝq⬘兲 and to the direction of electron motion in the
undulator 共ĝq = 1兲, as well as have finite spatial overlap with
The general expression for the mode coupling coefficient in both the e-beam distribution. In the single mode limit 共q
Eq. 共A7兲 permits analysis of mode coupling for radiation = q⬘兲, Eq. 共A8兲 simplifies to the commonly used “filling fac-
modes that vary transversely in their complex field amplitude tor” parameter.

关1兴 N. M. Kroll, P. L. Morton, and M. N. Rosenbluth, IEEE J. 79 共2001兲.


Quantum Electron. 17, 1436 共1981兲. 关8兴 E. L. Saldin, E. A. Schneidmiller, and M. V. Yurkov, Opt.
关2兴 G. T. Moore, Nucl. Instrum. Methods Phys. Res. A 239, 19 Commun. 97, 272 共1993兲.
共1985兲. 关9兴 P. Sprangle, A. Ting, and C. M. Tang, Phys. Rev. Lett. 59, 202
关3兴 E. T. Scharlemann, A. M. Sessler, and J. S. Wurtele, Phys. Rev. 共1987兲.
Lett. 54, 1925 共1985兲. 关10兴 D. Marcuse, Theory of Dielectric Optical Waveguides 共Aca-
关4兴 M. Xie and D. A. G. Deacon, Nucl. Instrum. Methods Phys. demic Press, New York, 1974兲.
Res. A 250, 426 共1986兲. 关11兴 B. Z. Steinberg, A. Gover, and S. Ruschin, Phys. Rev. A 36,
关5兴 Y. Pinhasi and A. Gover, Phys. Rev. E 51, 2472 共1995兲. 147 共1987兲.
关6兴 M. Xie, Nucl. Instrum. Methods Phys. Res. A 445, 59 共2000兲. 关12兴 A. Yariv, Quantum Electronics, 3rd ed. 共Wiley, New York,
关7兴 J. Wu and L. H. Yu, Nucl. Instrum. Methods Phys. Res. A 475, 1989兲.

063830-8

You might also like