1 Surface Science Analysis of GaAs Photocathodes Following Sustained Electron Beam Delivery in Jefferson Lab Accelerators

Degradation of the photocathode materials employed in photoinjectors represents a challenge for sustained operation of nuclear physics accelerators and high power Free Electron Lasers (FEL). Photocathode quantum efficiency (QE) degradation is due to residual gasses in the electron source vacuum system being ionized and accelerated back to the photocathode. These investigations are a first attempt to characterize the nature of the photocathode degradation, and employ multiple surface and bulk analysis techniques to investigate damage mechanisms including sputtering of the Cs-oxidant surface monolayer, other surface chemistry effects, and ion implantation. Surface and bulk analysis studies were conducted on two GaAs photocathodes , which were removed from the JLab FEL DC photoemission gun after delivering electron beam, and two control samples. The analysis techniques include Helium Ion Microscopy (HIM), Rutherford Backscattering Spectrometry (RBS), Atomic Force Microscopy (AFM) and Secondary Ion Mass Spectrometry (SIMS). In addition, two high-polarization strained superlattice GaAs photocathode samples, one removed from the Continuous Electron Beam Accelerator Facility (CEBAF) photoinjector and one unused, were also analyzed using Transmission Electron Microscopy (TEM) and SIMS. It was found that heat cleaning the FEL GaAs wafer introduces surface roughness, which seems to be reduced by prolonged use. The bulk GaAs samples retained a fairly well organized crystalline structure after delivering beam but shows evidence of Cs depletion on the surface. Within the precision of the SIMS and RBS measurements the data showed no indication of hydrogen implantation or lattice damage from ion back bombardment in the bulk GaAs wafers. In contrast, SIMS and TEM measurements of the strained superlattice photocathode show clear crystal damage in the wafer from ion back bombardment.


Introduction
Scientific productivity at user-based electron accelerators depends on robust photocathode operation to generate high current continuous wave (CW) electron beam current, as is the case with high power Free Electron Lasers (FELs), or highly polarized CW electron beams for nuclear physics research. Photocathode QE, defined as the ratio of generated charge (or current) to incident laser power, degrades during beam delivery, and when this yield falls below usable levels, beam delivery must be interrupted to replenish or replace the photocathode.
The goal of this work is to analyze the surface morphology, composition, and crystalline quality of photocathodes that produced hundreds or thousands of Coulombs of charge in DC photoemission guns, in order to better understand the degradation mechanisms. Increasing operational lifetime through understanding photocathode damage mechanisms can help focus future improvements for the electron sources and could improve accelerator availability in both machines at Jefferson Lab (JLab) and elsewhere.
In this work, two types of GaAs-based photocathodes, used in the two accelerators at Jefferson Lab (JLab), have been analyzed; the IR-VUV FEL and the Continuous Electron Beam Accelerator Facility (CEBAF). The IR-VUV FEL operates with up to 10 mA CW of unpolarized electron beam current obtained from a 350 kV DC photoemission electron gun that uses "bulk" (epi-ready) GaAs photocathodes [1]. CEBAF uses a 100kV DC photoemission electron gun with strained superlattice GaAs photocathodes to achieve electron beam polarization over 85% at beam currents up to 200 µA CW.
There are some difficulties associated with using GaAs in practice. Primarily this type of photocathode is sensitive to poor vacuum conditions, which result in degradation of the yield or QE. However, if ultrahigh vacuum conditions exist, activated cathodes can survive for thousands of hours when not illuminated with laser light [2]. The QE lifetime is also hundreds of hours under DC high voltage with no illumination. It is when the photocathode is illuminated and an electron beam is emitted, that the QE decays exponentially with time, supporting the ion back bombardment damage hypotheses. Photocathode QE degradation, of the type seen in the JLab FEL and CEBAF DC photoemission guns, has long been associated with ionization of residual gasses in the vacuum chamber and subsequent acceleration into the photocathode material [3].
Understanding the relative contributions of surface damage via sputtering of the Csoxidant surface monolayer versus the damage due to ion implantation and ion-induced dislocations in the crystal and heat-induced dopant density fluctuations could provide direction for future efforts to improve photocathode lifetime and yield, particularly for the higher currents demanded by future projects. Although the cathode electrode in the FEL gun does not provide focusing, there is clear evidence of localized QE degradation, particularly in the geometric center of the photocathode as shown in Figure 1, which corresponds to the electrostatic center of the anode-cathode geometry. The CEBAF gun electrodes are based on Pierce geometry to provide electron beam focusing, which does not strongly focus ions. However, as the laser beam is moved to many spots around the photocathode active area, the accumulation of damage at the electrostatic center from delivering beam from multiple spots leads to significant and nearly irreversible QE degradation at the photocathode electrostatic center, and reversible damage at the illuminated area and on a band between the two locations [3].

Photocathode Procedures:
A direct evolution of the original JLab IR-Demo FEL 350kV DC photoemission gun, the IR-VUV FEL Upgrade version has been in operation since 2003 [4,5]. In 1992 when the JLab IR FEL was proposed, GaAs photocathodes were chosen for electron beam characteristics reasons and because of the experience at CEBAF with polarized photoinjectors. The FEL does not require polarized electrons, but the fact that GaAs can be switched on and off at mega-Hertz repetition rate and generate ~50 ps-long electron pulses when illuminated with 532 nm wavelength laser beam, made GaAs the photocathode of choice to meet the pulse train structure and 5 mA CW current required by the IR FEL. Each photocathode installed in the photoemission gun is a 3.2 cm diameter, bulk GaAs wafer with crystal orientation (100) ± 5º, 600 µm thick and Zn-doped at ~1x10 18 cm -3 . The wafers arrive in hermetically sealed packages from the manufacturer.
For installation, a GaAs sample is attached to the end of a 4-ft. long stalk in a clean room by means of indium foil (which melts on heating providing good thermal contact, and acts as glue on cooling) and a tantalum ring that prevents the wafer falling from the stalk during heating.
The stalk is then mounted in a hydrogen cleaning chamber. The wafer was heated to 550 °C for one hour then hydrogen cleaned at 300 °C for 15 minutes and allowed to cool. The stalk with the clean GaAs wafer is then inserted into the photoemission gun for nominal operation. The complete gun assembly is then baked for one week at 250 ºC to achieve vacuum conditions below 5x10 -11 Torr. The photocathode wafer is subsequently heat cleaned at 550 °C inside the gun for three hours, then cooled to room temperature. Finally, the wafer is activated in-situ, into a negative electron affinity (NEA) photocathode by a combination of Cs-atom deposition in a low-pressure background of NF 3 (1x10 -10 Torr) until the photocurrent saturates. In the FEL gun, typical QE values resulting from this activation process are between 5 and 7% at 532 nm for bulk GaAs, although higher QE has been reported elsewhere [6].
The photocathode is nominally illuminated with a frequency-doubled (527 nm), modelocked Nd:YLF drive laser to generate 135 pC per bunch. Each laser pulse is 50 ps FWHM with a transverse top-hat profile 8 mm in diameter. At a repetition rate of 75 MHz the DC photoemission gun can deliver up to 10 mA CW [7]. The QE decays exponentially when electron beam current is extracted from the photocathode during accelerator operation. To compensate for this the drive laser power is increased. When the drive laser can deliver no more power, the photocathode is rejuvenated with a fresh layer of Cs, recovering about 96% of the previous QE.
Approximately every 6 re-cesiations, excessive field emission, observed by a downstream electron beam screen, necessitates a "heat clean" and reactivation of the wafer into a NEA photocathode. The reactivation process recovers the original 5-7% QE and eliminates the field emission, which might be due to accumulation of Cs on the wafer surface. Excessive field emission is problematic for high current FEL operations, as it results in electron beam halo that scrapes the beam pipe chamber walls generating undesired radiation and deteriorating vacuum conditions.
With progressive use of the photocathode the beam halo increases. Halo is simply electrons that are emitted from unwanted regions on the photocathode. Eventually, the halo becomes too problematic in downstream regions of the accelerator and the photocathode must be replaced.
The CEBAF photoinjector delivers up to 200 µA of CW polarized electron beam to three experimental halls simultaneously, at the CEBAF accelerator frequency of 1497 MHz. The photocathode (12.8 mm dia.) is illuminated with three frequency-doubled, gain-switched, fibercoupled seed lasers with ErYb-doped amplifiers, each interleaved at 499 MHz, to provide independent current control for the three experimental halls. The laser beams (~0.5 mm diameter FWHM) are moved to new locations across the photocathode when photocathode QE falls to an unacceptable level. The photocathode can support operation from approximately 6 photocathode locations. After exhausting the QE from the entire photocathode, the sample must be heated and reactivated. The CEBAF photocathode is not re-cesiated between heat/reactivation cycles.
The CEBAF wafers are delivered from the manufacturer with an arsenic cap, and diced into 1.5 cm squares. The analyzed photocathode was used in a gun similar to, but smaller than, the FEL electron source, with the photocathode mounted on a 26" long stalk that is inserted into the gun vacuum chamber and baked at 250 °C for 30 hours. As with one of the FEL wafers, this CEBAF wafer was anodized in a mild acid solution to define the active area of the photocathode prior to installation, thus reducing beam halo from unfocused emission near the edges of the photocathode. Because the high polarization strained superlattice photocathodes are more delicate and require better vacuum, the horizontal CEBAF gun was pumped with a large array of NEG pumps surrounding the cathode/anode gap to provide maximal pumping in the region and pressure was measured around 5x10 -12 Torr in the system. A heat cycle of ~500 °C was found sufficient to activate the photocathode to a QE of over 1% with 780 nm light. Higher temperatures were avoided to minimize dopant migration from the highly doped surface layer.

FEL photocathode samples:
In 2003, a GaAs wafer from vendor Matek delivered over 1000 Coulombs in one year of operation and up to 9 mA CW for short periods of time. Due to excessive electron beam halo at 5-8 mA CW, in 2004 it was replaced with a new bulk GaAs wafer grown by AXT Inc. The extracted 1/e lifetime was 550 Coulombs or 50 operational hours at 5 mA CW. The total extracted charge from that wafer was 7000 Coulombs in 900 operational hours with a beam current between 1 and 8.5 mA CW. A fresh layer of Cs was applied to replenish the quantum yield typically once per week. The Matek wafer underwent 12 heat cleans before removal in 12 months of operation, while the AXT wafer underwent 9 heat clean cycles in 36 months of operation [7]. Both the Matek and the AXT wafers suffered localized surface damage (~1 mm 2 to the lower right of the electrostatic center, see Figure 1) while delivering beam in excess of 8.5 mA CW for FEL operations. Although the damage mechanism is unclear, gathered data indicates that the vacuum in the electron gun chamber increased from 1x10 -11 to ~1x10 -7 Torr and the electron beam delivery was suddenly stopped by the machine protection system closing the drive laser shutter. Since the gun was still at 350 kV it is believed that the sudden deterioration in vacuum conditions triggered field emission from a localized, 1 mm 2 area in the photocathode, indicated by a current spike of at least 3 mA. It is worth noting that after such an event, the QE was obliterated, but a nominal heat clean cycle followed by NEA activation restored the photocathode to the original 5% QE, albeit with some field emission detected from the damage spot.

Polarized photocathode sample:
CEBAF uses strained superlattice photocathodes to produce highly polarized electron emission by illumination with circularly polarized light just over the band gap energy. Strain, induced by growing GaAs on a lattice-mismatched substrate [8], breaks the degeneracy between two states in the valance band. If illuminated with a light source that excites electrons exclusively from the upper state, 100% polarized emission is theoretically possible. To increase electron yield and maintain high polarization, photocathodes are now grown in a strained superlattice structure using 14 paired layers of GaAs/GaAsP and a final highly doped GaAs surface layer to reduce surface charge limit effects [9,10]. Typical polarization from these photocathodes, illuminated with circularly polarized light at 780 nm, has been measured at over 85%. The sample analyzed delivered over 600 Coulombs of polarized electrons over a period of approximately 6 months and was heated and reactivated 6 times, with beam delivered to the halls for over 2600 hours during that time.

Sample identification and surface analyses:
For the work presented here, a total of six GaAs wafer samples were analyzed. The two superlattice wafers analyzed were from the same manufacturing run, with one having delivered 600 Coulombs and the other unused. Two bulk wafers, from different manufacturers, used in the JLab FEL DC electron gun were analyzed years after being removed from the system along with two bulk GaAs control samples. The Matek wafer was installed in the FEL gun in 2003 without further treatments. Prior to installation in 2004, a new wafer from AXT Inc was first anodized with a mixture of de-ionized water and 85% phosphoric acid to form a thick oxide layer, effectively eliminating the QE outside the intended use area. In Figure 1, the wafer anodized area shows in dark blue.
The following list describes each sample, (summarized in Table 1).

Helium Ion Microscopy and Atomic Force Microscopy
Helium ion microscopy was used to image photocathode surface morphology. HIM is very similar to scanning electron microscopy (SEM) except that uses He + as a probe instead of electrons [11], and has several advantages over SEM. For example, the depth of field is higher in HIM than SEM and the interaction volume is much smaller such that very high resolution images can be obtained (current instrument resolution is 0.35 nm). Furthermore, an electron beam can be used as a charge neutralizer for insulating samples, thus carbon coating of the sample, as is often done in SEM, is not necessary. HIM images were obtained using 25 keV helium ions with 5 pA beam current at normal incidence. Secondary electrons and backscattered ions were detected using Everhart-Thornley and micro-channel plate detectors respectively.
HIM images of the four bulk GaAs FEL photocathode samples are shown in Figure 2. It is clear from these images that the surface underwent roughening, due to heating (compare nanostructures with an average size of 50 nm. It is known that high temperature heating of GaAs in vacuum induces excess evaporation of As and growth of Ga clusters on the surface [12,13].
Usually these processes are studied at temperatures above 550 °C and the material is thought to remain stoichiometric when heated below the non-congruent evaporation temperature of 625 °C [12]. Figure 2 clearly shows a systematic variation in the topography of the samples as a function of heat treatment and prolonged photoinjector use, which seems to develop smaller surface nanostructures. AFM images from these samples provide quantitative roughness data and are quite similar to the HIM results. The results shown in Table 1 suggest that the surface roughness does not increase with additional heating cycles and may be reduced by photoinjector operation. The Matek sample suffered more damaging events than the AXT sample and therefore needed more heat clean cycles. Since the damage described in the "FEL photocathode samples" section is localized for each arcing event, it seems to be unrelated to the surface roughness outside those specific sites.

Rutherford Backscattering Spectrometry
Rutherford backscattering spectrometry (RBS) has been extensively used to investigate stoichiometry, structure and thickness of thin films and bulk materials [14][15][16]. In RBS, the probe is typically a He + ion beam of energy between ~0.5 and ~2.0 MeV. In this energy regime, Coulomb scattering of the incident ion and nuclei in the solid can be treated classically and reasonably accurate numerical simulations can be performed. RBS is element specific as the energy of the backscattered He + ion is dependent on the mass of the scatterer at a particular scattering angle. Since the ion loses energy as it travels through the target material, an energy spectrum of the backscattered ions also yields chemical information about the depth at which a given backscattering event occurs. Since high-energy ion beams penetrate deeply into materials, RBS can be used to study buried interfaces and diffusion profiles. RBS spectra were obtained using 2 MeV He + ions under channeling and random geometries. The backscattered spectra were collected using a silicon surface barrier detector at a scattering angle of 150 o . The incident beam diameter was 1 mm. A detailed description of the RBS experimental setup is reported elsewhere [17].
RBS in channeling geometry probes both surface and bulk material order, such as crystallinity, and is also useful in characterizing defects or damage structures, and impurity or dopant lattice site locations in single crystal materials. In channeling geometry, the incident ion beam is aligned with the low index direction of the crystal surface and can penetrate between atomic rows such that the energy spectrum of backscattered ions exhibits a surface peak (SP) associated with backscatter from atoms in the topmost layers ( Figure 3a). Because of shadowing by the surface layer, the atoms that do not scatter from the surface penetrate deep within the crystal without scattering, such that the backscattering yield from the bulk of the crystal is very low.
When the atoms in the surface region are displaced from their original lattice position, due to surface structural change, the SP intensity is expected to increase compared to the value corresponding to the clean, undamaged (well-ordered) surface. In addition, if there is disorder in the near surface region or in the bulk, due to ion implantation or some other means, the channeling will be disrupted and the incident ion beam will be scattered by the displaced (interstitial) or implanted atoms. As a result, the backscattering yield from the atoms in a disordered or damaged region will give a rise to a damage feature in the channeling spectrum (see Figure 3b) [14,18]. No damage feature is observed in the RBS data. In random geometry, The minimum yield ratio (χ min ) is the ratio between the back scattered ion yields in the channeling and random geometries. The (χ min ) ratio is usually measured near the minimum following the surface peak and is useful in characterizing crystalline quality. A low χ min value, on the order of 3%, indicates sample is well ordered. The results shown in Figure 5  The 1000 Coulomb Matek sample was NEA re-activated two days before being removed from the electron gun and still had a substantial QE (~2%) left, while the 7000 Coulomb AXT sample was NEA re-activated about 2 weeks before being removed from the electron gun and its QE was below 0.1% prior to removal. The correlation between Cs concentration and QE of each sample prior to removal from the gun supports a QE degradation mechanism based on Cs loss due to ion back-bombardment sputtering.
It is believed that hydrogen ion back bombardment causes the photocathode QE to degrade with use. If the loss of QE is due to bulk wafer damage induced by ion backbombardment, then we would expect to observe damage peaks in the channeling spectrum ( Figure 3). Calculations indicate that hydrogen implantation should peak at a depth of 1.5 µm which, in turn, should lead to a damage feature near channel 450. The channeling spectra ( Figure   5) do not show any evidence of damage peaks. It is possible that bulk damage caused by hydrogen back-bombardment is repaired somewhat during the heat cleaning and reactivation process, although implanted hydrogen would be expected to be retained inside the sample. We attempted SIMS depth profiling to detect implanted hydrogen. The SIMS depth profiles show no evidence of hydrogen implantation and we conclude that hydrogen concentrations in all the samples are lower than our detection limit of 1 x 10 18 atoms/cm 3 . Coulomb AXT sample. Figure 6 shows TOF SIMS images of Cs smoothly distributed on both the 1000 and 7000 Coulomb samples. The measured Cs surface concentrations are consistent with those determined by RBS.

Strained Superlattice Analysis
Transmission electron microscopy (TEM) was used to study the damage of the strained superlattice photocathode sample. Sample preparation for TEM analysis of the strained superlattice structure was performed at North Carolina State University using a focused ion beam (FIB) milling and a lift-out technique to prepare the samples without damaging the superlattice structure [20]. To remove the depth cross section for TEM analysis, the GaAs is first protected by coating with an evaporative layer of gold and palladium. This is followed by a layer of platinum sputter coated with the FIB, followed by FIB milling of the sample. The TEM images obtained from the strained superlattice photocathodes (used and unused control samples) are shown in Figure 7. TEM images clearly show the layers of the superlattice in cross section.
While lattice dislocations are evident in both images, areas of significant darkening, typical of ion implantation damage, can be seen in the CEBAF sample (Figure 7a) but not in the unused control sample (Figure 7b). The damage region is non-uniform across the surface of the material at this length scale. The laser spot is ~500 μm diameter, far larger than the distance scales observed on this TEM image. The nature of this non-uniformity is not understood at this time. representative milling site in the unused sample. SIMS data was measured at seven locations across the used sample, and most data followed one of two trends: either slight implantation near the surface and little implantation beyond (data indicated by crosses) or moderate implantation at all depths (data indicated by triangles). Representative data sets for these two conditions are shown. One location yielded high readings in all species (shown in circles), and is believed to be near the electrostatic center where ions tend to be focused when running at any beam position.
Hydrogen due to out-gassing from the stainless steel vacuum chamber is the primary residual gas in the UHV vacuum system and is expected to implant, but the damage due to hydrogen ion implantation may be limited due to its low mass. Carbon and oxygen are present in the system due largely to the chemistry in ion pumps which can show residual gas spectra with CH 4 , CO and CO 2 . Note the oxygen count rates are three orders of magnitude higher than the other signals, partially due to the much higher sensitivity of the method for oxygen. The fluorine trace is present due to residual NF 3 in the system from the activation of the photocathode. Since the size of the SIMS ion beam is much larger than the scale of the damage sites on the TEM image, determining the elemental composition of the various implantation sites or the degree of implantation per layer is very difficult.

Conclusions:
In an effort to better understand quantum efficiency degradation mechanisms in bulk and superlattice GaAs photocathodes, several samples that delivered many Coulombs in DC photoemission guns have been analyzed using HIM, AFM, RBS, TOF-SIMS, TEM and SIMS.
These results suggest that ion-back bombardment, commonly recognized as the main degradation mechanism, can cause sputtering of the Cs monolayer and/or crystal damage.
What is interesting about our findings is that the FEL photocathodes in the 350 keV gun showed evidence of Cs sputtering, but not of ion induced crystal damage, at least within the resolution of the measurements. The correlation found by the RBS measurements between Cs concentration and QE of each sample prior to removal from the gun supports the Cs sputtering by ion back bombardment as a QE degradation mechanism. The combination of SIMS and RBS measurements should have indicated crystal dislocations or hydrogen implantation due to back bombardment if present, but the data showed no evidence of this. Conversely, the TEM analysis of the CEBAF strained superlattice photocathodes showed areas of significant darkening, typical of ion implantation damage in the used photocathode. SIMS measurements of the used CEBAF photocathode found implanted hydrogen, oxygen, and carbon and fluorine atoms, at varied levels across the photocathode surface, due to backscattering in the 100 kV DC CEBAF gun. Cs sputtering was not investigated as a damage mechanism for the CEBAF samples.
Due to the small number of samples, our results are not unequivocal, but rather serve as motivation to further studies. For example, it is thought that thermal effects are responsible for repairing some of the damage induced by back-accelerated ions implanted into the photocathode material during operation. Measurements at JLab using bulk GaAs indicate that extended heating (12 hours rather than 2 hours at 525°C) is significantly more effective in rejuvenating the QE and maximum yield in both bulk and strained superlattice photocathodes, particularly in heavily damaged areas such as the electrostatic center [21].
Of more profound consequences to high brightness photocathodes are the results confirming previous observations of surface nanostructure formation induced by heating during the cleaning step in the activation of negative electron affinity photocathodes. The Cornell group has studied the effect of surface morphology on thermal emittance, of similarly treated GaAs photocathodes, showing that their analytical model for photoemission predicts their emittance measurements when the surface roughness is included in the model [22]. Unfortunately, at the moment there are no means to characterize the effect of the surface morphology changes on the electron beam emittance at the JLab FEL.
It is interesting to note that the FEL photocathodes remained crystalline after prolonged use, even under the localized damage due to high voltage arcing. In addition, surface roughness appears to be most strongly correlated to the initial heat treatment, although we cannot quantify changes to roughness due to additional heating and cleaning cycles. The Coulombs of beam delivered or time in the gun may be factors in the reduced roughness observed in used samples.
These factors indicate that the heating and cleaning cycle should be evaluated to prolong the use of a photocathode in the FEL. Since the number of available samples for analyses was limited, a further study on a separate cathode preparation chamber would be valuable in answering these questions.
Proposed next generation light sources [23] will be capable of generating spatially and temporally coherent X-ray beams with unprecedented photon brightness, on the order of 10 21 -10 25 photons/s/mm 2 /mr 2 0.1% bandwidth, at photon energies between 10 1 -10 3 eV [24]. The predicted photon brightness is ultimately limited by the electron beam normalized transverse emittance at the undulator, which in turn is limited by the thermal emittance of the photocathode [25]. Developing a process to preserve the native surface quality of the GaAs wafer could be a significant contribution to low emittance photocathodes [25,22]     Matek and 7000 Coulomb AXT samples taken within the drive laser illuminated region near the electrostatic center shows the expected surface composition, with peaks due to Ga, As and Cs.