October 2002
SLAC-PUB-9529
Techniques for Electro-Optic Bunch Length
Measurement at the Femtosecond Level†
P. Bolton, D. Dowell, P. Krejcik*, J. Rifkin
Stanford Linear Accelerator Center
2575 Sand Hill Rd, Menlo Park CA 94025
Contributed paper at the 10th Beam Instrumentation Workshop 2002,
Brookhaven, NY, May 5-9, 2002.
*
corresponding author, pkr@slac.stanford.edu
†
This work is supported under the DOE contract DE-AC03-76SF00515.
Techniques for Electro-Optic Bunch Length
Measurement at the Femtosecond Level†
P. Bolton, D. Dowell, P. Krejcik*, J. Rifkin
Stanford Linear Accelerator Center
2575 Sand Hill Rd, Menlo Park CA 94025
Abstract. Electro optic methods to modulate ultra-short laser pulses using the electric field of a
relativistic electron bunch have been demonstrated by several groups to obtain information about the
electron bunch length charge distribution. We discuss the merits of different approaches of transforming
the temporal coordinate of the electron bunch into either the spatial or frequency domains. The
requirements for achieving femtosecond resolution with this technique are discussed. These techniques
are being applied to the Linac Coherent Light Source (LCLS) and the Sub-Picosecond Photon Source
(SPPS) currently under construction at SLAC.
INTRODUCTION
New generations of accelerator-based x-ray laser sources will utilize extremely
short electron bunches. The bunches are considerably shorter than can be measured
with existing streak camera technology and will require new, innovative techniques to
diagnose and tune them. The Linac Coherent Light Source (LCLS) [1] to be built at
SLAC utilizes electron bunches as short as 80 femtoseconds rms to generate selfamplified stimulated emission (SASE) X-ray radiation in a FEL. The new SubPicosecond Photon Source (SPPS)[2] at SLAC offers a near term opportunity to test
and compare these different diagnostic techniques with bunches as short as 30 fs rms,
far shorter than anything so far produced in a high energy electron accelerator.
Conventional laser technology has succeeded in producing and characterizing
visible light pulses with the time structure that is well matched to these electron beam
requirements. The bridge between the two systems exists in the form of electro optic
(EO) crystals whose birefringence properties are modulated by the electric field of the
electron bunch to be measured. The transmission of polarized light through an EO
crystal is in turn modulated and the problem of measuring an electron bunch length is
thereby transformed into one of measuring the duration of a light pulse.
This paper discusses first the choices of configurations between laser and electron
beam geometry. For a given geometry the sensitivity and bandwidth requirements are
then presented. The light pulse detection techniques discussed show the evolution
from interferometric techniques to nonlinear autocorrelation methods and the
advantages they offer. Finally, the limits to resolution are discussed and the latest
developments towards pushing this limit.
*
corresponding author, pkr@slac.stanford.edu
†
This work is supported under the DOE contract DE-AC03-76SF00515.
ELECTRO OPTIC PROBE GEOMETRY
Two basic configurations are possible. One is where the probe laser pulse is
transverse to the electron beam and the second where the two beams are parallel and
co-propagating. Each of these geometries has implications for: laser power in the
crystal; whether the pulse sampling should be performed in the spatial or frequency
domain; and on the ultimate resolution that could be achieved. An illustration of the
transverse geometry is shown in figure 1, similar to that proposed in reference [3].
In the transverse scheme a cylindrical lens brings the polarized laser beam to a
ribbon focus at the EO crystal where it overlaps the electric field of the electron
bunch. The length of the ribbon focus determines the duration of the timing gate in
which the electron bunch must be coincident. A cross polarizer normally extinguishes
the light reaching the CCD detector. However, the light will be modulated in
proportion to the change in birefringence induced by the electric field of the bunch.
The image recorded on the detector is therefore a convolution of the electron bunch
charge distribution and the temporal profile of the laser pulse. In order to achieve good
resolution a laser pulse length much shorter than the electron bunch length must be
used. For example, a Ti:sapphire laser with 20 fs would resolve sub-picosecond
electron bunches.
A 100 fs electron bunch is 30 µm in length so some magnifying optics must be used
in front of the detector to overcome the pixel resolution limit of approximately 9 µm
in a CCD camera. Furthermore, the retardation in the crystal is wavelength dependant
and since a short pulse laser has by nature a very large bandwidth, care must be given
to calibrating the system. Wavelength calibration is difficult in the transverse
geometry where the extent of the bunch is to be measured in the spatial domain.
This problem is partially overcome in the second transverse geometry scheme
shown in figure 1. The laser pulse is chirped and the modulation of the electron bunch
now acts to gate a portion of the stretched laser pulse. A spectrometer detects the
modulation in the frequency domain thereby avoiding limitations due to pixel
resolution and beam size. The wavelength calibration of the optical elements is readily
done with the electron beam off and using quarter-wave plates.
Incoming
chirped laser
pulse
Grating Spectrometer
ωl
Analyzer
CCD
Detector
Analyzer
Electro-Optic
Crystal
Polarizer
Ribbon focus
laser pulse
Electro-Optic
Crystal
Polarizer
t
Detector
Ι
ω
FIGURE 1. Transverse, line-focus probe geometry can produce a spatial image of the bunch[3] (left),
or gating of a chirped laser pulse (right).
Beam pipe
Electron bunch
Co-propagating
Laser pulse
Polarizer
EO Crystal
Spectrometer
Analyzer
I
ωl
t
Initial laser chirp
ωs
t
Bunch charge
Gated spectral signal
FIGURE 2. In the longitudinal geometry the electric field from a bunch modulates the polarization in
an electro optic crystal and gates the transmission of a co-propagating, chirped laser pulse.
In order to achieve good resolution the laser must be focused to a spot at the EO
crystal much smaller than the bunch length. Spot sizes of a few microns would be
adequate for a 30 micron (100 fs) bunch but the technique now suffers from excessive
power density in the EO crystal.
Transverse probe geometries were used successfully in early experiments[4] with
picosecond beams, but for femtosecond resolution the longitudinal probe geometry
shown in figure 2 has distinct advantages. In the longitudinal geometry the laser pulse
co-propagates with the electron bunch. The chirped laser pulse is stretched to around
10 ps to ensure timing overlap with the electron bunch. The chirp provides a
correlation between the time domain and the frequency spread of the pulse which can
be measured with great precision[5] with a spectrometer. The laser pulse need not be
focused to small dimensions in the EO crystal so there is no limitation due to laser
power density in the crystal. To understand the limiting resolution of this technique we
first look at some of the properties of the EO interaction.
ELECTRO OPTIC BASICS
The electric field of the bunch alters the optical retardation of the birefringent
crystal. The polarization, P, of the probe laser in the crystal changes with the electric
field, E, in proportion to the susceptibility according to
P = ε 0 01 E + 0 2 E 2 + 03 E 3 + !
(1)
The susceptibility, 0 , has the same symmetry as the crystal so that all crystals
except those with inversion symmetry exhibit some EO properties. The first term in
the expansion in equation (1) is the linear, isotropic term. It is the second-order term
that is exploited here, namely the linear EO effect, or Pockels effect, in which the
polarization changes under the influence of the external, applied electric field of the
bunch. A third-order term also exists which drives the second-order EO effect,
referred to as the Kerr effect. Birefringence is the anisotropy of the crystal’s refractive
indices and results in differing propagation velocities along the different crystal axes.
In isotropic media the index of refraction, n, is related to the susceptibility by
2
0 = n1 − 1
(2)
It is convenient to define the change in index through the EO crystal, following the
methodology of Yariv[6], by defining an index ellipsoid with principal axes
x2 y 2 z 2
(3)
+
+
=1
nx2 n y2 nz2
The change in index with applied field is then given by the coefficients rij for a
specific crystal
3
1
(4)
∆ 2 = ∑ rij E j
n i j =1
The electric field of an ultra-relativistic bunch of Ne electrons at a distance r, from
the crystal is [7]
−
Er = 9 ×109
s2
2σ z2
2 Nee e
r
2π σ z
in m.k .s units
(5)
2 Nee 1
r
2πσ z
measured at the center of a Gaussian bunch of length σz.
The opening angle, 1/γ, of the electric field lines dictates that the crystal should be
within a couple of millimeters of the bunch for low energy bunch length
measurements at the electron gun. At GeV energies the crystal can be a centimeter
away without loss of resolution or sensitivity.
The retardation, Γ, of the light transmitted by a crystal with trigonal symmetry such
as LiTaO3 or LiNbO3 is
and
Ermax = 9 × 109
d
1/γ
r
3
L
1
Ê
2
FIGURE 3. A relativistic bunch at a distance r from the crystal and laser pulse.
πL
(6)
E (ne3 r33 − no3 r13 )
λ0
and is determined by the electric field strength, E, and the thickness of the crystal
along the light path, L,
The electric field corresponding to a retardation of λ/2 (when the polarized light
changes from zero to 100% transmission) is therefore
λ
1
Eπ = 0 3
L (ne r33 − no3 r13 )
(7)
2 N ee
9
= 9 ×10
r 2πσ z
from which the required thickness of the crystal can be determined.
Γ=
AUTOCORRELATION AND DETECTION
The light pulse transmitted by the EO crystal is characterized by its amplitude A,
carrier frequency ω0, carrier phase ϕ0 and chirp ϕ (t),
E (t ) = A (t )cos (ω 0t + ϕ (t ) + ϕ 0 )
(8)
Its temporal profile, C(t), can be measured by interferometric means, shown in figure
4a, which is an autocorrelation technique
∞
C (t ) = ∫ E (τ )E (t + τ ) dτ = E (−t ) ⊗ E (t )
(9)
∞
However, by the Wiener-Khinchin theorem this is equal to the Fourier transform of
2
the magnitude squared, Eν , and consequently does not preserve phase or
asymmetry information about the bunch.
3rd harmonic
generation in
nonlinear crystal
M1
M1
N.L.
Xtal
M2
BS
time t
BS
grating
CCD
Detector
delay time t
FIGURE 4. Autocorrelation (left) with an interferometer, and a nonlinear autocorrelator (right) for
Frequency Resolved Optical Gating.
Alternatively, if the light rays cross at an angle inside a nonlinear crystal, as illustrated
in figure 4b, higher harmonics are generated and the unbalanced, third-order
correlation terms preserve pulse asymmetry information. Combining the nonlinear
autocorrelator with a grating spectrometer results in a powerful diagnostic to
Frequency Resolve the Optically Gated (FROG) pulses.
LIMITS TO RESOLUTION
The EO detection method is resolution limited by the phase slippage that occurs
from the slight difference in propagation velocity for the sub-millimeter radiation from
the bunch and the 800 nm laser radiation in the crystal arising from the difference in
refractive index, ∆n,
L
∆t = ∆n
(10)
c
This effect is minimized by choosing a crystal with small ∆n and small thickness, L.
For example, in LiTaO3 ∆n= 0.1 so that a 100 µm thick crystal has a slippage of 30 fs.
As an alternative we are investigating a new technique based on measuring the
wavelength shift that occurs at the entry and exit of the crystal where there is a timedependant change in refractive index
∆λ L dn L 1 3
=
=
ne r33 − no3 r13 )
(11)
(
c dt c 2
λ
This effect has been observed in laser-plasma interactions[8] and is proposed as a
means to surpass the resolution limits set by phase slippage. For the crystal parameters
above, equation (11) predicts a 3% wavelength shift at the edges of the crystal.
In conclusion, it is expected that improvements in EO detection techniques will go
hand-in-hand with progress in the generation of ultra-short electron bunches over the
next few years.
REFERENCES
1. “LCLS CDR”, SLAC-R-593, April 2002.
2. M. Cornacchia et al., “A Subpicosecond Photon Pulse Facility for SLAC”, SLAC-PUB-8950, LCLSTN-01-7, Aug 2001. 28pp.
3. T. Srinivasan-Rao et al., “Novel Single Shot Scheme to Measure Submillimeter Electron Bunch
Lengths Using Electro-Optic Technique”, Phys. Rev. ST Accel. Beams 5, 042801 (2002).
4. M.J. Fitch et al., “Electro-optic Measurement of the Wake Fields of a Relativistic Electron Beam”, Phys. Rev.
Letters, Vol. 87, Num. 3, 034801(2001).
5 I. Wilke et al., “Single-shot Electron-beam Bunch Length Measurements”, Physical Review. Letters 88, 124801,
2002.
6. A. Yariv, “Quantum Electronics”, Springer-Verlag (1967).
7. A.W. Chao, “Physics of Collective Beam Instabilities”, John Wiley & Sons, New York, 1993.
8. P. Bolton et al., “Propagation of Intense, Ultrashort Laser Pulses Through Metal Vapor”, J.O.S.A. B, Volume
13, Issue 2, p. 336 (Feb. 1996).
View publication stats