1Institute of Electronic Engineering and Optoelectronic Technology, Nanjing Institute of Science and Technology, Nanjing,Jiangsu 210094, China
2Department of Electronic Engineering, East China Institute of Technology, Fuzhou, Jiangxi 344000, China3Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China
Heavily doped p-type GaAs cleaved in an ultrahighvacuum and coated with monolayers of cesium andoxygen has a low enough work function to permitphotoexcited electrons from all energies, giving ahigh quantum yield and producing a high-efficiencyphotocathode [1–4].
Research on increasing the quantum efficiency ofthe photocathode has been developing for quite a longtime. There have been some attempts. One of theattempts was making the band-bending region (BBR)smaller to decrease the possibility scattering and re-combination. The width of the band-bending region(in cm) is given by [5]
W � �2�VBB�qNA�1�2, (1)
where q is the electron charge (in C), � is the dielectricconstant (in F�cm), VBB is the amount of band bend-ing (in V), and NA is the acceptor doping concentra-tion (in cm�3).
W can be made small by using a high doping den-sity, NA. Unfortunately, high doping density alsocauses a short diffusion length, which reduces thephotocathode quantum yield.
Another way is to add an electric field on the pho-tocathode to help the photoexcited electron diffuse tothe surface and escape, but it is hard to achieve be-cause the photocathode is too thin.
Since it is hard to add an electric field on thephotocathode [6], a gradient-doping structure is in-vented, in which the surface doping concentration islower than the bulk concentration. The number ofband-bending regions of the material is more thanone. Each band-bending region has an electric field.So the electric field in the gradient-doping GaAs ma-terial is greater than the uniform-doping GaAs ma-terial. It should be easier for the electron to diffuse tothe surface in the gradient-doping GaAs material.
In this paper, the characteristics between auniform-doping GaAs and a gradient-doping GaAsare compared. Two GaAs (100) reflection modephotocathodes are grown (A and B) by molecularbeam-pitaxy (MBE) for the study. Sample A is auniform-doping GaAs, with p-type beryllium doping.Sample B, which is divided into four sections, is a
gradient-doping GaAs, with p-type beryllium doping.Each section has a different doping concentration.
2. Experiment
A. Setup
A GaAs photocathode multiinformation measure-ment system is used in the experiment. The blockingof the GaAs photocathode multiinformation measure-ment system is shown in Fig. 1. The system is madeup of two parts. One part activates the GaAs photo-cathodes. The other part measures the parameters ofactivated GaAs photocathodes such as integratedsensitivity and quantum efficiency.
B. Preparation of GaAs Photocathode
Before being sent into the ultrahigh vacuum cham-ber, both of the samples were degreased by carbontetrachloride, acetone, ethanol, and hydrofluoric, andrinsed in an ultrasonic bath. Then according to the“high–low temperature” two-step activation tech-nique, monolayers of cesium and oxygen (Cs:O) weredeposited on the surface of GaAs under ultrahighvacuum conditions to create a stable negative elec-tron affinity (NEA) surface suitable for photoemis-
sion. The doping structure of sample A and sample Bare shown in Fig. 2.
3. Experiment Results
Spectral response curves and quantum yield curves ofthe activated samples are collected after the two sam-ples are activated. The curves are shown in Fig. 3.
It is easy to know from the curves that the spectralresponse sensitivity and quantum efficiency of sam-ple B, which is gradient doped, is much higher thanthat of sample A, which is uniform doped, especiallyfor illumination with wavelength below �800 nm.Also there is some increase between �800 and910 nm. Some parameters of curves are shown inTable 1.
4. Discussion
The transport of the photoexcited electrons at photonenergies near the bandgap can be described by thecontinuity equation [7]
d2n�dx2 � n�LD2 � �g�x��Dn, (2)
where n is the excess electron density, Dn is the elec-tron diffusion constant, LD is the electron diffusionlength, and g�x� is the generation rate of electrons.
g�x� � �1 � R�I0 exp���x�, (3)
where I0 is the incident light density, � is the absorp-tion coefficient of GaAs, and R is the reflection coef-ficient of GaAs.
Solving Eq. (2) gives the carrier density at dis-tance x:
n�x� � C1 exp��x
LD�� C2 exp� x
LD��
�
1 � �2LD2 g�x�,
(4)Fig. 2. Doping structure of (a) sample A and (b) sample B.
Fig. 1. Blocking of multiinformation measurement system.
where C1 and C2 are constants and LD � �D���1�2 is thediffusion length of electrons. The value of the inte-gration constants C1 and C2 is set by the two followingboundary conditions:
n�x��x�0 � 0, n�x��x� � 0. (5)
The density of photoexcited and thermalized elec-trons at a given x is therefore
n�x� ���1 � R�LD
2I0
�1 � �2LD2�Dn
�exp���x� � exp��x
LD��. (6)
Assuming an electron escape probability P at theGaAs surface,
Yr �
PDn
dndxX�0
I0. (7)
Subject Eq. (7) to Eq. (6), and the quantum effi-ciency can be given as
Yr �P�1 � R�
1 � 1��LD. (8)
So the quantum efficiency is chiefly related to es-cape probability P and electron diffusion length LD.Escape probability and diffusion length of the twosamples are fitted and shown in Table 2.
Both the diffusion length and escape probability ofsample B are higher than for sample A. Vergara andGomez [8] have shown that for illumination withwavelengths below �800 nm, photoemission is dom-inated by the escape probability, and for wavelengthbetween �800 and 910 nm, the dominant role isplayed by the diffusion length of electrons in the bulk.From Table 1 we know that the escape probability ofsample B is 22.2% higher than that of sample A,while the diffusion length of sample B is just 11.6%higher than that of sample A. So it can be explainedthat the spectral response sensitivity and quantumefficiency of sample B are much higher than for sam-ple A for illumination with wavelengths below �800nm, while there is just a little increase for illumina-tion between �800 and 910 nm.
A. Diffusion Length
LD could be given as
LD � �Dn��1�2, (9)
where � is the mean lifetime of electron, Dn is thediffusion constant of GaAs, and � is mainly determinedby the doping concentration. The greater the dopingconcentration, the more scattering, and the smallerthe �. It could be known from Fig. 2 that the average
Fig. 3. (a) Spectral response curves and (b) quantum responsecurves of two activated samples.
Table 1. Spectral Response Property Parameters and Sensitivity of Curves
CurveNumber
Spectral Response Property
Integrated Sensitivity(�A�lm)
Start Wavelength(nm)
Cutoff Wavelength(nm)
Peak Response(mA�W)
Peak Wavelength(nm)
1 500 935 316 500 24212 500 945 231 500 1966
Table 2. Data Fitting Results of Spectral Curves
Sample A Sample B
Diffusion Length��m 5 5.5Escape Probability 0.45 0.55
doping concentration of sample B is smaller than thatof sample A, so the � of sample B is higher.
Using the Einstein relationship to determine thediffusion constant gives
Dn � �kT�q��n, (10)
where k is Boltzmann constant, T is the temperature,q is the charge of electron, and �n is the mobilitycoefficient. When the doping concentration is highand the temperature is low, �n is given as
�n T3�2�NA, (11)
where NA is the acceptor density. We know from Fig.2 that NA of sample B is smaller than the one ofsample A, so it can be inferred that electron diffusionlength Dn of sample B is higher.
B. Escape Possibility
When the electrons reach the surface with energyhigher than the vacuum level energy, they can beeither transmitted or reflected by the surface poten-tial barrier. The band structure and surface potentialbarrier of uniform-doping NEA GaAs and gradient-doping NEA GaAs [9,10] are shown in Fig. 4.
According to the widely accepted double-dipolemodel, there are two surface potential barriers, whichare called potential barrier 1 and potential barrier 2.The electrons escape to vacuum through the barriersby the tunnel effect. It is shown in Fig. 4 that poten-tial barrier 1 is high but thin, while potential barrier2 is low but thick, so if the electrons can escape tovacuum through potential barrier 1, the escape prob-ability P could be higher because it is thinner thanpotential barrier 2. On one hand, almost all the elec-trons in uniform doping GaAs do not have enoughenergy to escape to vacuum through potential barrier1. They have no choice but to escape to vacuumthrough potential barrier 2, which is thick. On theother hand, there are electric fields in the bulk ofgradient-doping GaAs. There is a band-bending re-gion at each interfacial between two different dopingconcentration sections. Each band-bending regionhas an electric field. Electrons will be acceleratedwhen they get through the band-bending regions.
We take the band-bending region between Section1 and Section 2 as an example. We call it band-bending region D1. The height of potential barrierbetween the two sections is given as
qVD1 � EF1 � EF2 � �EV � kT lnNV
NA1�� �EV � kT ln
NV
NA2�
� kT lnNA2
NA1, (12)
where EF1 and NA1 are the Fermi level and dopingconcentration of Section 1. EF2 and NA2 are the Fermilevel and doping concentration of Section 2, NA1 �
NA2. EV is the valence band energy of GaAs. NV is thevalence band density of states.
The width of band-bending region D1 is given as
WD1 � �2�0�VD1
qNA1�1�2
. (13)
So the electric field intensity in band-bending regionD1 is given as
E1 �VD1
WD1. (14)
It can be known from Fig. 2 that there are threeelectric fields in sample B, and electrons will be ac-celerated in each band bending region when they getthrough. It could be calculated from Eqs. (12)–(14)that the energy of the electrons that reach the surfaceis about 0.059 eV higher than that of sample A, andmore electrons could escape to vacuum through po-tential barrier 1, so escape probability of sample B ishigher than sample A.
Fig. 4. Band structure and surface potential barrier of NEA GaAsphotocathodes: (a) uniform-doping NEA GaAs and (b) gradient-doping NEA GaAs. EC is the conduction band minimum, EV is thevalence band peak level, Eg is the width of the bandgap, EF is theFermi level, and �S is the height of the surface band bending.
Since the escape probability and the diffusionlength of sample B are both higher than those ofsample A, based on Eq. (8), it is easy to know thatquantum efficiency of sample B is better.
5. Conclusion
We have compared two kinds of NEA GaAs photo-cathode. We found that by changing the doping struc-ture the spectral response sensitivity and quantumefficiency increase significantly. The integrated sen-sitivity of the uniform-doping one is 1966 �A�lm, andthe integrated sensitivity of the gradient-doping oneis 2421 �A�lm. From the comparison we determinedthat, although a low surface doping concentrationincreases the width of surface band-bending region, itdoes not influence it too much. What is importantis the increase of escape probability and diffusionlength. It can be inferred that the characteristic of theNEA GaAs photocathode could be better if the dopingstructure of GaAs material were optimized.
The authors thank Xiaofeng Wang for supplyingthe r-mode GaAs samples. This work is financed bythe National Natural Science Foundation of China(grant 60678043) and by the Specialized Research forthe Doctoral Program of Higher Education of China(grant 20050288010).
References1. G. A. Mulhollan, A. V. Subashiev, J. E. Clendenin, E. L. Gar-
win, R. E. Kirby, T. Maruyama, and R. Prepost, “High perfor-
mance polarized electron photocathodes based on InGaAl�AlGaAAs superlattices,” Phys. Lett. A 282, 309–318 (2001).
2. H.-J. Drouhin, C. Hermann, and G. Lampel, “Photoemissionfrom activated gallium arsenide. 1. Very-high-resolution en-ergy distribution curves,” Phys. Rev. B 31, 3859–3871 (1985).
3. T. Maruyama, A. Brachmann, J. E. Clendenin, T. Desikan,E. L. Garwin, R. E. Kirby, D.-A. Luh, J. Turner, and R. Prepost,“A very high charge, high polarization gradient-doped strainedGaAs photocathode,” Nucl. Instrum. Methods Phys. Res. A492, 199–211 (2002).
4. Liu Lei and Chang Benkang, “Spectral matching factors be-tween Super S-25 and New S-25 photocathodes and reflectiveradiation of objects,” Appl. Opt. 43, 616–619 (2004).
5. L. W. James, G. A. Antypas, J. Edgecumbe, R. L. Moon, andR. L. Bell, “Dependence on crystalline face of the band bendingin Cs20-activated GaAs,” J. Appl. Phys. 42, 4976–4980 (1971).
6. Lihui Guo, Jinmin Li, and Hou Xun, “The quantum efficiencyof field-assisted transmission-mode GaAs photocathodes,” J.Phys. D 22, 348–353 (1989).
7. Y. Z. Liu, J. L. Moll, and W. E. Spicer, “Quantum yield of GaAssemitransparent photocathode,” Appl. Phys. Lett. 17, 60–62(1970).
8. G. Vergara, L. J. Gomez, J. Capmany, and M. T. Montojo,“Influence of the dopant concentration on the photoemissionin the NEA GaAs photocathodes,” Vacuum 48, 155–160(1997).
9. X. Q. Du and B. K. Chang, “Angle-dependent XPS study of themechanism of “high-low temperature activation of GaAs pho-tocathode,” Appl. Surf. Sci. 251, 267–272 (2005).
10. C. Y. Su, W. E. Spicer, and I. Lindau, “Photoelectron spectro-scopic determination of the structure of (Cs, O) activated GaAs(110) surface,” J. Appl. Phys. 54, 1413–1422 (1983).
