Você está na página 1de 7

Emittance Measurements of a Nanopatterned Cathode on the DC beam line

Enrique Mendez
University of Hawaii at M
anoa
(Dated: August 28, 2015)
In the generation of photoemitted electron bunches for the use in accelerators, metals are limited
by their low QE. Previous work [P. Musumeci et al., Phys. Rev. Lett. 110(2013)] has shown that
charge yield from infrared light is increased by a factor of 120 by etching nanopatterns into the
cathode. However, utility of the electron bunch is also based on its thermal emittance, or angular
spread, which is conserved during transit along the beam line. In that same work, a measure of
(1.4 0.1) mrad was found in the emittance for a copper nanopatterned cathode held in an electric
field strength of 75 MV m1 . It is proposed that the presence of the nanopattern will alter the
electric field and likely the emittance. Measurements were taken for the same cathode in an electric
field strength of (1.6 0.2) mrad. Hence no apparent effect of the electric field strength on the
emittance was found.
I.

qBz
where q, me are the charge and mass of
where k = 2m
ev
the electron, v is the velocity of the entire beam bunch,
= 1 2 , and z is the length of the magnetic field
1(v/c)

traversed. We denote this matrix M (k, z).


Since the magnetic field is approximately constant
in small regions along the beam beam line, we can
write the transformation matrix T for the beam line as
87 cm
Solenoid 2
41.6 cm

Solenoid 1

12

.5

cm

Virtual Instrument, a program written in LabVIEW.

Fitting Emittance

In all accelerators, the dynamics can often be described


by linear transformations. For a constant magnetic field
aligned along the axis of the beam line, a particle with
transverse position x and angle x0 relative to the beam
line axis transforms to[3]
  
 
xf
x
cos(kz) k1 sin(kz)
=
(1)
x0f
x0
k sin(kz) cos(kz)

Our goal was to set up emittance measurements on the


DC beam line at UCLA which runs at a field strength
5 MV m1 and a total voltage of 30 kV . A schematic of
the setup is given in Fig. 1. A VI1 was adapted to take
a series of photos at varying currents for a solenoid scan.
A MATLAB script was written to analyze these photos,
extract dimensions of the beam and calculate emittance
measurements.

A.

THEORY

Cathode

The beam brightness is a fundamental quantity determined by charge and emittance. High brightness means
high charge and low emittance. Both semiconductor
and metal photocathodes are used in the pursuit of high
brightness beams. However, whereas as semiconductors
can be tailored to have high quantum efficiency (QE),
they are easily destroyed by exposure to air. Metals are
more robust and durable but have QE several orders of
magnitude lower. Nonlinear optical processes cause the
current density that is emitted to be proportional to I n
where I is the intensity of the incident light. It is also
proportional to (1 R)n , where R is the reflectivity of
the material to the incident light[1]. Since metals are
highly reflective in the infrared (IR), which is the wavelength that is used in the UCLAs particle beam physics
laboratory, the effect is not as efficient as it could be.
Investigations were made into etching nanopatterns into
the cathode to decrease reflectivity and increase charge
yield. Previous work has shown a factor of 120 increase in
charge yield[2], as well given thermal emittance measurements on this cathode of values (1.4 0.1) mrad. What
is speculated however is that the emittance is a monotonically increasing function of the electric field applied,
and that the emittance will decrease as the electric field
decreases.

II.

INTRODUCTION

56 cm

10.4 cm

102 cm

FIG. 1. A (not to scale) schematic of the DC beam line.


The upward pointing arrows represent the dipole steering (or
kicker) magnets. The DRZ is a scintillating screen mounted at
a 45 degree angle relative to the beam line axis and camera.
The MCP (microchannel plate) is parallel to the cathode.
Data in this paper was measured at the DRZ.

Since we can describe the dynamics of one particle in the


beam, we can describe the dynamics of non-interacting
ensemble. If the matrix transformation of the beam line
is written as


B C
T =
(3)
B0 C 0
then we can write the observed spot sizes as linear combinations of products of B and C and the initial beam
parameters[4]. If we fit the beam parameters to the data
then we will be able to calculate the geometric emittance
g which is related to the normalized rms emittance by
 = g [3] where = v/c. To ensure that the emittance
measured isnt due to space charge (the self repulsion of
the beam), measurements are taken at low charges as
determined by simulations.
B.

General Particle Tracer (GPT) Model

The DC beam line (for which a schematic may be found


~
in Fig. 1) is implemented using a 1D E-field
for the
2
cathode , a normalized 1D field map of Bz for the first
solenoid, and the bzsolenoid function with parameters
passed to it that best fit the 1D field map of Bz for the
second solenoid.
If we set the pulse length to 50 fs, the radius to 100 m,
the initial excess energy3 of the beam as 0.4 eV, with the
initial distribution set using the following code:
setrxydist("beam","u", radius/2, radius);
setphidist("beam", "u", 0, 2*pi);
setGBzdist( "beam", "u", GB, 0);
setGBthetadist( "beam", "u", pi/4, pi/2);
setGBphidist("beam","u",0,2*pi);
settdist("beam","g",0,pulse,3,3);
with GB () defined using the excess energy given, we
find the solenoid scan given in Fig. 2.
To test the emittance fitting implementation, it is used
to fit a simulated data set. Results are plotted in Fig 2.

As a check that the E-field is correct, we can integrate to find a


value of 30.03 kV. The model reaches a value of 30.00 kV at
9.32 mm and 30.02 kV at 10.7 mm.
Excess energy is defined to be the thermal energy, kT , of the
electron bunch.

Simulated Solenoid Scan


Variance (109 m2 )

M (kn , dz)M (kn1 , dz)...M (k1 , dz) where each ki is the


calculated k for the given field strength at its corresponding position in the beam line. This method accounts for
sections of zero magnetic fields as well, as in the limit of
no magnetic fields, the matrix reduces to the standard
drift matrix


1z
M (0, z) =
.
(2)
0 1

80
Data
Fit

60
40

1.5

1.6

1.7

1.8

Current (A)
FIG. 2. Space-charge free simulated solenoid scan done in
GPT with 50 fs pulse length, 100 m spot size radius, and
an initial excess energy of 0.4 eV. An emittance is fitted on
equally weighted data and found to be (36 4) nm, compared
to the simulated (36.3 0.2) nm. Note that the fit is accurate
when the maximum values of the variances are approximately
twice that of the variance at the waist.

Calculations were also performed to check whether


the fitting algorithm would find that emittance was conserved along the beam line. It was found to be constant
from the beginning and up to 35 cm into the beam line.

III.

PROCEDURE

The current in solenoid one is fixed. Steering magnet


and solenoid two currents are adjusted until the beam
waist appears on the screen. The current in solenoid two
is then varied across a range of currents that cause the
beam to go through the beam waist. Multiple images are
taken at each value of the second solenoid current. The
beam is then deflected off the screen and background
images are taken for removing noise.
An average background image is calculated. Each
beam image has the background subtracted to be used
for analysis. Image masks are automatically generated
using the method outlined below. The resultant cropped
images are then rotated by the calculated Larmor angle,
projected along each of two perpendicular axes and fit
using Gaussians (defined by Eq. (8)) from which standard deviations are taken. The set of fitted standard
deviations is then recorded for each image at a particular current as the beam width. At each current value,
the average beam width is calculated and an error determined from the standard deviation of the fitted beam
widths. The resulting averages and errors are then fitted
to find the emittance as outlined in the Theory section.

IV.

IMAGE ANALYSIS

In the case of low charge as is the case for taking emittance measurements, the beam ends up being dim. Fur-

3
thermore, in some datasets, there are dead pixels or stray
light sources that distract the analysis routine. To reduce
the amount of manual work required on each particular
data, a semi-automatic routine was developed to pick out
dim beams.

A.

Noise Management

Given just an image of a black screen there will be


variations in the pixels due to noise. If the peak beam
brightness is only a few factors above the deviation in the
noise, an initial guess for the beam is hard to pick out
even when using thresholding since the noise will vary
above the average background level. That is, thresholding will lead to an image with a beam and various dim
spots around the image.

When considering the root mean square xx as a


measure of the beam where
R
(x x
)2 I(x, y)dxdy
R
xx =
(4)
I(x, y)dxdy
with I(x, y) being the image intensity as a function of
pixel location and x
being the centroid of the image along
the x-axis, specifically defined as
R
xI(x, y)dxdy
x
= R
.
(5)
I(x, y)dxdy
we have that size will be greatly over estimated due to
the noise. This is since the squaring in Eq. (4) greatly
weights large distances from the centroid despite the fact
that the noise is far less dense than the signal itself.
This is a problem as the initial parameters passed
to the Gaussian fitter are the rms values. Bad inital estimates lead to the fitter converging to a local
minimum that does not reflect the signal.
So to
compensate for this effect, it was proposed that one
could utilize a modified image with a boosted signal to
determine initial rms values and then pass this initial
estimate to a fitter on the unmodified4 data. With
a boosted signal more noise can be cut and a better
initial estimate on the rms values (and the position of
the beam) can be made and passed to the Gaussian fitter.
To demonstrate the idea behind the signal boosting
method proposed, we consider an N th order moment of
a one dimensional distribution which is defined to be
R
(x x
)N f (x)dx
R
(6)
f (x)dx
with x
defined similarly as before. For large N , the tails
of the distribution are weighted heavily in determining

Not including binning and background subtraction.

the value of the integral. In analogy to such an N th order


moment relative to x and x
, take the N th order moment
relative to image number and the average background
noise
!1/N
n
1X
N
I,N (x, y) =
max (Ii (x, y) B(x, y), 0)
n i=1
(7)
for each pixel at (x, y). It was posited that for large N ,
deviations from the noise, i.e. signals, would be weighted
heavily since large intensities grow much faster under a
power law than smaller values. The actual effect as a
function of N was calculated for a data set and is given
later.
From this constructed image intensity I,N , one can
then make dynamic masks that follow the beam for each
image and, furthermore, calculate accurate estimates on
the beam size for passing to nonlinear fitting programs on
the unmodified image. Furthermore, binning the image
before performing the above calculation further reduces
the noise. Binning can be done safely since, at the beam
waist, the size of the beam is 200 m and pixels themselves are 20 m
The program that has been developed uses this N th
order moment method with binning. The initial thresholding sets most noisy spots to zero, eroding5 sets isolated nonzero values to zero, smoothing with Gaussian
filtering6 generates circular neighborhoods for surviving
datapoints, thresholding the smoothed image sets a radius on the neighborhoods which can then be normalized
to a mask image.

B.

Determining the spot size

Taking a mask (used for cropping) determined by some


method, the image can then be analyzed to find the beam
width. Firstly, the beam is rotated as it travels through
a solenoid by an angle equal the Larmor angle, so to
ensure that the beam image is in the correct orientation
for reading consistent beam widths, the cropped image is
rotated by an angle opposite to the Larmor angle. Next,
the rotated image is then projected onto each axis and
fit with a Gaussian
F (x; A, , x0 , y0 ) = y0 + Ae 2 (
1

xx0

(8)

where y0 , , x0 and A are fit parameters. The advantage


of this method is that constant background noise levels

Image eroding scans across the image looking at various neighborhoods. Each neighborhood is then set to the minimum value
in that neighborhood, thereby cutting out isolated noise spikes
left untouched by the thresholding.
A resource on how Gaussian filters work can be found here

SNR

SNR

2
1
1.8

1
1.7
Current 1.6
(A)

FIG. 3. Signal to Noise Ratio along a Solenoid Scan as function of N , for N defined in Eq. (7). Fits were done using
a Gaussian function with a constant offset. The SNR was
calculated by taking the ratio of the amplitude of the fitted
Gaussian with the norm of the residuals in the fit.

can be added as a fit parameter rather than thresholding


the noise.
For Gaussian beams, the standard deviation of the fitted guassian is equivalent to a root mean square measure
of the data, however when the beam is not Gaussian it
is not clear that this is the case. Note that the reason
this is a possible issue is that the emittance is defined in
terms of the second moments, i.e. the root mean square.
Fortunately, the beam is Gaussian at the beam waist due
to the nature of focusing when considered as a transformation in phase space. Thus our method is limited to
scans close to the beam waist.

C.

1.8

5
1.5 10

Testing the Image Analysis Routine

It was postulated that in Eq. (7), that for large N ,


the signal would be weighted largely against the noise
and increase the signal to noise ratio (SNR). To actually
measure this, analysis was carried out on images from
a solenoid scan at 8 fC bunch charge. The images were
manually cropped around the beam.
We can identify each I,N (as defined by Eq. 7) derived
from the original images as an image. Each I,N was then
projected onto an axis and fit using Eq. 8. The results are
found in Fig. 3. It is found that the best signal to noise
ratio (SNR) is found for the first order moment N = 1,
and the SNR actually decreases as the moment goes up.
This brings up the interesting idea that perhaps one can
increase the SNR by using fractional N . The results are
plotted in Fig. 4.

6
5
4
3
2
1
1.7
Current

1.6
(A)

1.5

3 1

1
3

1
5

1
8

FIG. 4. Signal to Noise Ratio along a Scan Using Fractional


N , for N defined in Eq. (7)

V.

BEST PRACTICES IN MEASUREMENT

The largest time cost in taking data is the time spent


in finding the beam after some change in the alignment
of the optical system has caused it to be lost. It has been
found that at 3 A the beam is so defocused that glancing
over the nanopattern will show up on the MCP with no
current in any of the steering magnets. However, the
(first) steering magnets hysterisis (at least in the vertical
direction) can be strong enough to deflect the beam and
cause it to disappear.
The diffraction pattern that the laser beam makes as
it passes through the irises also significantly affects how
well fine tuning the beam position changes charge output. First, the diffraction pattern is circularly symmetric
when aligned properly. However, often what is brightest
is not the center of the diffraction pattern but the outermost fringe near the boundary of the iris (Aperture 1
in Fig. 6) opening. If the iris is opened too large, one
will have to align the nanopattern on the outer fringe
for highest yield. To fix this problem reduce the opening
radius of the intensity limiting aperture until the spot
appears uniform on the virtual cathode.

VI.

DATA

800 nm incident laser light comes in at a pulse length


of 50 fs and illuminates the nanopattern. The resulting
charge is then accelerated to 30 keV energies. The electrons then impinge upon a scintillating screen known as
a DRZ.
Emittance measurements were taken at three currents
in the first solenoid (Isol1 ): 1.96 A, 1.54 A, and 1.45 A for
various charges. Assuming uniform illumination of the
nanopattern, which has a radius of r = 100 m, the root

2.5
2
1.5
1

x /
y /
r /

0.5
0
5

10

15

20

25

Thermal Emittance ( mmmrad


)
mm rms

Thermal Emittance ( mmmrad


)
mm rms

5
5
4
3
2
x /
y /

1
0
5

10

Charge (fC)

Aperture
3

Virtual
Aperture
2

Mirror 1
Mirror 2
Aperture 1

25

FIG. 7. Second Solenoid Scan taken at Isol1 = 1.45 A.


Weighted Square Residual (107 )

To the Chamber

Cathode

20

Charge (fC)

FIG. 5. Second Solenoid Scan taken at Isol1 = 1.96 A. Simulations are done in GPT for various initial excess energies
and fit to the data. The thermal emittance as a function
of charge for a simulated beam with initial excess energy of
1.7 eV is plotted.

Splitting Mirror

15

Excess Energy Fit


3
2
1
0
1

1.5

2.5

Excess Energy (eV)

Lens

From the Laser

FIG. 6. Optical Setup. Light enters intensity modulated due


to polarizers not shown here. It is then reduced in intensity
through aperture 1, passed through a lens with 50 cm focal
length and then passes through two apertures used for alignment. They are used only in alignment and remain open after.
The virtual cathode and photocathode are equidistant from
the lens.

mean square of the beam bunch is = r/2 = 50 m,


the thermal emittances are given by 50 m where  is the
emittance. The thermal emittance graphs are given below in Fig. 5 and Fig. 7. In Fig. 5, The simulated
thermal emittance of the a beam with an excess energy
that best fits the data is plotted as well. One datapoint
was obtained at Isol1 = 1.54 A and at 8 fC bunch charge
to give thermal emittances of x / = (2.9 0.8) mrad
and y / = (2.5 0.7) mrad. Taking a weighted average
of the emittances at 1.94 A in Fig. 5 given by
P
wi i
avg = Pi
(9)
i wi
where each wi is the inverse of the corresponding error
of the ith emittance measurement i . Taking the error in
the average to be
sX
1
avg = P
(wi i )2
(10)
w
i
i
i

FIG. 8. Residuals of Initial Excess Energy Fit. Simulations


were ran in GPT to return emittances as a function of excess
energy for each of the charges in the Isol1 = 1.96 A solenoid
scan. The result is plotted here, we find that the fit is minimized for (1.7 0.1) eV

Noting that each product wi i = 1 by definition


1
avg = P
i

wi

(11)

where N is the number of emittance measurements. Taking this emittance and dividing by the rms spot size yields
a value of (1.6 0.2) mrad for the thermal emittance.

VII.

ANALYSIS

The thermal emittances was found to vary by a factor


of two for different scans. The smallest value appeared
at Isol1 = 1.96 A and the larger values at Isol1 = 1.45 A
and 1.54 A. It appears that the first solenoid causes an
increase in the emittance.
Interestingly, what has been found is that at low currents the beam is highly elliptical indicating the presence
of, at minimum, a quadrupole magnetic moment. The
fact that the beam has smaller emittances and shows up
as circular around the beam waist, at higher currents,
indicates the possibility that the quadrupole moment is

Beam size along the Beam line


Beam Size (mm)

2
1.5
1
0.5
0

I1 = 1.45 A, I2 = 0.8 A,
I1 = 1.96 A, I2 = 1.7 A,

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


Position (m)

solenoid one currents in Fig. 9. Note that at low currents


1.41.5 A, the beam size is large at both the kicker and
the second solenoid. At 1.9 A the beam is about half the
size at both the kicker and the solenoid, leaving it indeterminate that if the ellipticity and increase in emittance
is due to either of the two, which has the dominant effect
or whether both play a role or not. To test whether it is
not the kicker magnet one can run a solenoid scan at 2.2
A at which the beam size at the solenoid is identical to
those at 1.45, where problems have been observed, but is
minimal at the kicker.
VIII.

CONCLUSION

Beam Width (mm)

Spotsizes vs. Current in Solenoid One


3
23 cm
42 cm

2
1
0
1

1.5

2.5

The smallest thermal emittance measured was found


to be (1.6 0.2) mrad which gives an upper bound on
the thermal emittance. This value agrees well with previous measurements[5] of (1.4 0.1) mrad done at field
strengths of 75 MV m1 . We have found then that the
electric field does not have a significant effect on emittance.

Current (A)
FIG. 9. GPT Simulations of spot size along the beam line.
Current in the second solenoid is off in second plot and there
are no space charge effects. 23 cm is the location of the steering magnet. 42 cm is the location of solenoid two. Note that
the first solenoid focuses on the second solenoid at 1.8 A as
was verified experimentally. In the first plot, the beam size
is plotted as a function of position along the beamline. The
screen is at 86 cm.

actually in the steering magnets (perhaps due to different


hysterisis effects in the dipole pairs). What is possible is
that the solenoid is strong enough to focus the beam to
be at the waist as it travels through the steering magnets
thereby causing the electron beam to sample less of the
quadrupole moment.
What has been found during experimentation is that
the first solenoid focuses on the second solenoid at 1.8 A
so it is not immediately ruled out that the first solenoid
is causing the beam to pass through a waist as it travels through the kicker. To determine whether or not the
beam is sampling nonlinear fields in the second solenoid
or the quadrupole fields in the kicker magnets. Simulated beam size is plotted as a function of the various

[1] P. Musumeci, L. Cultrera, M. Ferrario, D. Filippetto,


G. Gatti, M. S. Gutierrez, J. T. Moody, N. Moore, J. B.
Rosenzweig, C. M. Scoby, G. Travish, and C. Vicario,
Multiphoton photoemission from a copper cathode illuminated by ultrashort laser pulses in an rf photoinjector,
Phys. Rev. Lett. 104 (2010).
[2] W. L. Barnes, A. Dereux, and T. W. Ebbesen, Surface
plasmon subwavelength optics, Nature 424 (2003).

A.

Further Work

Further work needs to be done to examine the emittance growth due to the the first solenoid before more
accurate thermal emittances can be measured. Investigations into non-Gaussianity of the beams compared to
rms only fitting with reduced noise experiments need to
be carried out.

B.

Acknowledgements

The author would like to thank Jared Maxson for the


large amount of time he invested in helping with gathering the data, and fleshing out these results. The author
would like to thank both Pietro Musumeci, David Cesar, and Jared for teaching the author about the world
of accelerator physics. All obtained results would have
not been possible with the work of Evan Threlkeld and
Dorian Johnson in maintaining and modifying the DC
beam line during the course of the REU. Furthermore,
none of this research nor learning experience ever have
happened without funding from the NSF.

[3] Max Hachmann, Transverse emittance measurement at


REGAE via a solenoid scan, Masters thesis, Universit
at
Hamburg (2012).
[4] H. Wiedemann, Particle Accelerator Physics I, 2nd ed.
(Springer-Verlag).
[5] R. K. Li, H. To, G. Andonian, J. Feng, A. Polyakov, C. M.
Scooby, K. Thompson, W. Wan, H. A. Padmore, and
P. Musumeci, Surface-plasmon resonance-enhanced multiphoton emission of high-brightness electron beams from

7
a nanostructured copper cathode, Phys. Rev. Lett. 110
(2013).

Você também pode gostar