Polarized Photocathode Development at SLAC
Jym Clendenin
Stanford Linear Accelerator Center Stanford, CA 95309
Sinclair Symposium for Photoemission Guns and Applications
Thomas Jefferson National Accelerator Facility October 26, 2001
1
Abbreviated History of Polarized Electron Beams at SLAC 1974 First acceleration of polarized electrons in a high energy
accelerator (using the YALE-SLAC Li atomic beam source).
1978 First accelerator-based experiment with polarized electrons
using a III-V semiconductor (GaAs) photocathode source. 1991 First demonstration of III-V photocathode (strained
InGaAs-GaAs) with Pe>>50%. 1991 Discovery of surface charge limit (SCL). 1992-1998 SLC operates continuously with polarized electrons using
newly designed source and (beginning in 1993) strained GaAs photocathodes.
1995 Observation of anisotropic QE effects associated with
strained GaAs photocathodes. 2001 Demonstration of highly polarized beam with peak and
average charge meeting NLC specifications and showing no SCL effects.
2
Early studies at SLAC (<1991)
Search for Pe>50%: 1. Circular polarization of recombination light studied when uniaxial stress applied.
Clearly difficult to get more than a few meV splitting of the hh and lh VB without breaking the crystal.
2. Superlattice structures of GaAs-AlGaAs (matched lattice constants) type.
Small enhancement – transport in CB needed improvement. 3. Ternary chalcopyrites. No VB degeneracy.
Polarization from pseudo direct band gap ZnSiAs2 not better than GaAs. Direct band gap types (ZnGeAs2) no NEA. More ideal variations (CdSiAs2, CdGeP2) extremely difficult to grow.
4. Strained GaAs.
3
For thick (bulk) GaAs, Pe=50% is the theoretical maximum. However, typically Pe<<50%. One route to the theoretical maximum value is to use thin (�d<<1) GaAs.
50
40
30
20
10
0550 650 750 850
PO
LAR
IZA
TIO
N
(P
erce
nt)
0.2 mµ0.425 mµ0.7 mµ0.9 mµ
λ (nm)11-886181A1
Electron spin polarization as a function of excitation photon wavelength for four different thicknesses of GaAs samples. All the measurements were made within 5 h after cathode activation. (MARUYAMA 89) Note: Polarimeter for these measurements may have given values too high by as much as 10%.
4
Higher Polarization Grow thin lattice-mismatched layer on a substrate. Mismatch accommodated by elastic strain in epilayer. Pseudomorphic growth takes place. Stored elastic strain is relieved, producting misfit dislocations when a characteristic critical thickness is exceeded. 70% polarization from bi-axial compressive strained In0.13Ga0.87As-GaAs achieved in early 1991. Thin (0.1 �m) and thick (1.14 �m) samples MBE-grown at UC Berkeley.
1.2
10
QU
AN
TU
M E
FF
ICIE
NC
Y
hν (eV)2-91 6818A4
1.4 1.6 1.8 2.0
-5
10-4
10-3
10-2
Cathode quantum efficiency as a function of excitation photon energy. The solid curve is for the 0.1 �m-thick sample, and the dashed curve for the 1.14 �m-thick sample. The band gap energies of GaAs (solid arrow) and relaxed InGaAs (dashed arrow) are indicated. (MARUYAMA 91)
1.2 1.6 2.00
20
40
60
80
hν (eV)
PO
LAR
IZA
TIO
N
(%
)
1.4 1.8 2.2
2-91 6818A3
(a) (b)
Electron spin-polarization as a function of excitation photon energy for the 0.1 �m-thick sample (a) and for the 1.14 �m-thick sample (b). (MARUYAMA 91)
5
Better performance expected from strained GaAs-GaAs1-xPx:
Larger band gap implies higher QE, Surface preparation of GaAs better understood.
SLAC studies have included:
Varying the epilayer thickness and P content. Samples MOCVD-grown by Spire Corp.
4
3
2
1
0
Qua
ntum
Effi
cien
cy
(x1
0– 3)
1.45 1.50 1.55
hν (eV)1-92 7084A8
Sample12345
Pol
ariz
atio
n (
perc
ent)
1.451.40 1.50 1.55hν (eV) 7084A21–92
40
60
80
100
Sample12345
(MARUYAMA 92)
(MARUYAMA 92)
No effect on Pe :
Lowering cathode temperature to 77 K, Heat cleaning cathode at up to 570 ºC.
6
Cathode characterization at low voltage greatly speeded up in 1990s with construction of a dedicated
Mott Polarimeter
Load Lock
Monochromater
Test Chamber
Sample Mount
Vertical Drive
Laser
Ion Pump
Transfer ArmNEGPump
Lens
Lens
FilterLC/LP
Mirror
Cathode Test System with a load-lock
In the figure above, the residual strain, measured by X-ray diffraction, varies with change of epilayer thickness, d, and phosphorus content, x. The P fraction (not shown) varies from x=0.2 to 0.33. All samples doped at 5�1018 cm-3 except the two samples indicated by the cross that are at 5�1018 cm-3. Generally Pe increases as d decreases, to 0.1 �m level. Higher dopant density reduces Pe.
7
SLC “production” cathodes, GaAs-GaAs1-xPx: x=0.28, d=0.1 �m.
Typical polarization and QE spectra measured at low voltage.
8
Surface ChargeLimit (SCL) Discovered at SLAC in 1991 during initial checkout of the SLC PES with the high-power pulsed laser operating at the band gap threshold. The effect is particularly severe for long pulse bunches with high charge, but is mitigated by high doping.
5
4
3
2
1
0
12
10
8
6
4
2
0
2001000
0.6
0.5
0.4
0.3
0.2
0.1
FAR
C in
tens
ity (
arb.
uni
ts)
150 100500
1.6
1.2
0.8
0.4
0.04003002001000-100
time (ns)time (ns)
100 200 3000
Faraday Cup Signal
5 x 1018
1 x 1019
2 x 1019
5 x 1019
Results for a series of measurements using unstrained 100-nm thick GaAs using the 120 kV gun. Each curve in a given frame respresents a different laser energy for the laser flat top pulse.
9
Subashiev’s charge limit model based on surface photovoltage
(MULHOLLAN 2001)
dU(F)
BBR
EF
V
D
+-
-
-
-
-
Ec
EF
V
dU(F)dYY
=D
dU(J)
D~
dU(J)dYY
=D~
_
+
dU(F) = q F ( - 1)4 ( + 1)p
eee0
s
s
Bias effect on Yield "Shottky effect"
Photovoltage effect on Yield
Restoring Current
j (U) = j [exp(U / E ) - 1 ]0 0
10
So for this data it appears the QE varies linearly with dU(F). The affinity, D (elsewhere �), can be calculated from the slope of such plots.
11
Time variation of the photovoltage (U):
� � � � � �UjJRdqUjjqdtdUC ppel ����� 1�
where jel flow of electrons to surface jp flow of excess hole current to surface J excitation light intensity C surface capacitance � optical absorption coefficient d cathode thickness R surface reflectivity With some assumptions, integration yields:
� � � �� �� � �
��
�
���
�
�
���
�
0,
0,0,0
0~1
~1exp~1ln~1
p
pp
jJqtjJqjJqE
YJY �
where
� �Rdjj pp �� 1~0,0, �
0,0 pqjCE�� is the photovoltage relaxation time.
C, E0 and jp,0 can be calculated from the constants of the band bending region including the dopant concentration.
12
From yield equation, we see that the photovoltage build up time is
� �0,~1 pjJq�
���
�
and saturation yield is
� ���
�
�
��
�
��
�
0,
0
0~1ln~1
pjqJE
YJY
Em
issi
on C
urre
nt (
A/c
m )2
0 20 40 60 80 100 120 140 160
0.0
0.1
0.2
0.3
0.4
0.5
Light Intensity (W/cm )2
0 20 40 60 80 100 120 140 160
0.0
0.1
0.2
0.3
0.4
0.5
Light Intensity (W/cm )2
5 x 1018 1 x 10
19
Y = 0.6 %
Y = 0.45 %
Y = 0.9 %
Y = 0.3 %
Measured saturation current for various conditions. Solid lines are fit of
saturation yield equation to the data. SCL more pronounced for samples with lower dopant density and/or lower (due to “aging”) QE.
13
160
120
80
40
0
Sample 1aSample 1b
Sample 2aSample 2b
40
20
0150100500
Sat
urat
ion
Tim
e τ'
(ns)
Light Intensity (W/cm2)
(a)
(b)
1-20018585A5
The saturation time constant �' of the photovoltage as a function of excitation light intensity for (a) dopant density 5�1018 cm-3 and yield 1a of 0.6%, 1b 0.45% and for (b) density 5�1018 cm-3 and yield 2a of 0.9%, 2b 0.3%.
From the fitted curves the relaxation time � can be calculated and independently jp,0. From this, E0=45 meV (average value), consistent with light holes being the main contributors to the restoring current. Resulting value of jp,0 about 100 times smaller than predicted for a Schottky barrier with metallic Cs layer.
14
Gradient Doping To avoid depolarization, high dopant density desired in only final few nm of epilayer at surface. Problem is how to activate surface. If using normal heat cleaning:
Does high-density dopant diffuse? Since there is evaporation of the surface when heat cleaning, must initially have extra thickness.
15
GaAs
Ga As P
Ga As Px 1-x
x = 0 0.29
0.290.71
GaAs sub.
High doping
Low doping
100 nm
2 m
2 m
Low doping = 5�1017 cm-3, high doping = 5�1017 cm-3 Sample 1: High-doping thickness = 25 nm; anodized sequentially at 5, 8.4, 1.9, 1.9 V. Sample 2: High-doping thickness = 40 nm; anodized once at 5 V.
16
0.01
0.1
1
QE
(%)
900850800750700650
wavelength (nm)
90
80
70
60
50
40
30
20
0.01
0.1
1
900850800750700650
wavelength (nm)
90
80
70
60
50
40
30
20
pola
riza
tion (
%)
0.001
0.01
0.1
1
900850800750700650
wavelength (nm)
90
80
70
60
50
40
30
20
pola
riza
tion (
%)
0.001
0.01
0.1
1
QE
(%)
900850800750700650
wavelength (nm)
90
80
70
60
50
40
30
20
a) b)
c) d)
Peak polarization (Sample 1) increases 62�83% as highly-doped thickness is decreased from a) to d).
17
Secondary Ion Mass Spectrometry (SIMS) analysis of surface dopant (Zn) concentration.
3
4
567
1018
2
3
4
567
1019
2
Zn
Dop
ing
Leve
l
80x10-3
6040200
Depth (nm)
unused heat-treated anodized at 5 V
(Note: SIMS not reliable in first few nm.) Conclusions: Anodization rate ~1.5 nm per Volt. Evaporation rate 0.5 nm per hour at full temperature. (Sample at 600 �C for total of 4.75 h�although actual temperature may have been 30-40� cooler.) Zn concentration profile unchanged by heating, although some Zn may have diffused out of surface.
18
SIMS results used to plot polarization:
90
80
70
60
50
Pol
ariz
atio
n (%
)
4003002001000
High doped layer thickness (A)
M05-5405
M05-5406
o
Polarization as function of thickness of highly-doped layer as determined by SIMS. MO5-5405 is Sample 1; MO5-5406 is Sample 2.
19
A Strained-Layer Cathode That Meets NLC Requirements NLC requires 190 micro-bunches, spaced 1.4 ns apart (total of 266 ns), each micro-bunch with at least 1.4�1010 e- in 0.5 ns (full width) at the gun, totaling 2.7�1012 e- per train. This is 3 times the saturation level of the SLC type cathodes. NLC peak micro-bunch current = 9 A. Used a modified version of SLC cathode MOCVD-grown by Bandwidth Semiconductor (formerly Spire):
90 nm
2.5 m GaAs P0.340.66
GaAs Py1-yy=0 -> 0.34
GaAs substrate
2.5 m
GaAs P0.95 0.05
GaAs10 nm 5 x 10 cm19 -3
5 x 10 cm17 -3
The energy barrier at the surface due to the high dopant density is eliminated by adding small amount P in epilayer except high dopant density volume near surface.
20
80
70
60
50
40
30
20
Pol
ariz
atio
n(%
)
850800750700650
Wavelength (nm)
0.01
0.1
1
QE
(%)
Polarization and QE spectra measured at low voltage. Peak polarization ~80%, QE at polarization peak >0.1%. Note that polarization peak is 40-nm blue shifted compared to standard SLC cathodes.
21
2.0
1.5
1.0
0.5
0.0
FA
RC
outp
ut(a
rb.u
nits
)
5004003002001000
Time (ns)
Maximum laser energy for flat top pulse is 80 �J at cathode. Laser fills cathode area (diameter 20 mm).
22
8
6
4
2
0
Cha
rge
(x10
11e-
/pul
se)
806040200
Laser Energy ( J)
23
2.0
1.5
1.0
0.5
0.0
Cha
rge
( x
1012
e-
/ pul
se)
200150100500
Flash-Ti only
2.3 x 1011
e- / pulse
2.2 x 1
Flash-Ti + YAG-Ti
Flash-Ti Laser Energy ( J)
Charge output for long-pulse laser (Flash-Ti) only and for long-pulse plus short-pulse (YAG-Ti) at fixed at 20 �J/pulse. Laser spot size at cathode reduced to 14-mm diameter to enhance SCL effect. Flash-Ti 270 ns, 804 nm, not flat top, energy varied. Flash-Ti alone produces total NLC charge, but not the peak current. YAG-Ti 4 ns, 790 nm, 20 �J/pulse. Used alone the YAG-Ti produced 2.3�1011 e-/pulse, equivalent to 9.2 A. The figure shows the charge produced by YAG-Ti remains constant even at maximum Flash-Ti energy (it is also independent of position relative to Flash-Ti pulse). 14-mm spot size enhances SCL effect (if any), but pulse not yet reached the space charge limit of the gun (at the 120 kV operating bias).
24
Strained Superlattice Structures Most promising route to higher polarization while in principle the QE can be retained while the SCL can be surpressed. SL may also significantly reduce the charge anisotropy due to asymmetric development of misfit dislocations when strain relaxes. Promising candidates for strained SL: GaAs-GaAsP – Pe~90% at 760 nm and QE~0.3% reported for gradient-doped MOCVD-grown strained SL. (NISHITANI 00) GaAs-AlxInyGa1-x-yAs – zero CB offset to enhance mobility of electrons moving toward the surface. (CLENDENIN 99)
25
Conclusions A healthy history of photocathode R&D focused on the SLAC linac and future collider requirements. For low-duty factor accelerators, peak currents are high, thus surface charge limit a serious limitation. This problem now solved for NLC-type pulse trains. Polarization of ~90% now looks feasible using strained superlattice structure.
26
References (MARUYAMA 89) T. Maruyama, R. Prepost, E.L. Garwin, C.K. Sinclair, B. Dunham, S. Kalem, “Enhanced electron spin polarization in photoemission from thin GaAs,” Appl. Phys. Lett. 55, 1686 (1989). (MARUYAMA 91) T. Maruyama, E.L. Garwin, R. Prepost, G.H. Zapalac, J.S. Smith, J.D. Walker, “Observation of strain-enhanced electron-spin polarization in photoemission from InGaAs,” Phys. Rev. Lett. 66, 2376 (1991). (MARUYAMA 92) T. Maruyama, E.L. Garwin, R. Prepost, G.H. Zapalac, “Electron-spin polarization in photoemission from strained GaAs grown on GaAs1-xPx,” Phys. Rev. B 46, 4261 (1992). (WOODS 93) M. Woods, J. Clendenin, J. Frisch, A. Kulikov, P. Saez, D. Schultz, J. Turner, K. Witte, M. Zolotorev, “Observation of a charge limit for semiconductor photocathodes,” J. Appl. Phys. 73, 8531 (1993). (CLENDENIN 99) J.E. Clendenin, T. Maruyama, Yu.A. Mamaev, A.V. Subashiev, Yu.P. Yashin, “Superlattice photocathodes for accelerator-based polarized electron source applications,” Proc. of the 1999 Particle Accelerator Conference, New York, 1999, p. 1988. (NISHITANI 00) T. Nishitani, O. Watanabe, T. Nakanishi, S. Okumi, K. Togawa, C. Suzuki, F. Furuta, K. Wada, M. Yamamoto, J. Watanabe, S. Kurahashi, M. Miyamoto, H. Kobayakawa, Y. Takeda, T. Saka, K. Kato, A.K. Bakarov, A.S. Jaroshevich, H.E. Scheibler, A.I. Toropov, A.S. Terekov, “Development of spin polarized electron photocathodes: GaAs-GaAsP superlattice and GaAs-AlGaAs superlattice with DBR,” in SPIN 2000, 14th Int. Spin Physics Symposium, Osaka, Japan, 16-21 October 2000, AIP Conf. Proc. (2001), p. 1021. (MULHOLLAN 01) G.A. Mulhollan, A.V. Subashiev, J.E. Clendenin, E.L. Garwin, R.E. Kirby, T. Maruyama, R. Prepost, “Photovoltage effects in photoemission from thin GaAs layers,” Phys. Lett. A 282, 309 (2001).
27