Towards advanced polarized electron sources

MainPolarized electrons have been and continue to be important research tools in atomic physics and condensed-matter research1. The opportunities presented by polarized electron beams in high-energy and nuclear physics led to the development of polarized electron sources in the 1970s2. After a brief period of using atomic sources for polarized electrons, a remarkable breakthrough occurred in 1974, with the discovery of GaAs photocathodes, which generate polarized electrons when illuminated by circularly polarized infrared photons3,4,5.After that, numerous GaAs sources were built and successfully used in high-energy and nuclear physics experiments6,7,8,9,10,11,12,13. In addition, novel techniques were developed for GaAs-based sources to achieve electron polarization approaching 85% and high quantum efficiency14,15. Compared with other photocathodes, GaAs photocathodes exhibit extreme sensitivity to residual gas content and pressure, as well as to ion back-bombardment of the cathode surface12,16. The most important consequence of these requirements is that operational polarized GaAs sources6,7,8,9,10,11,12,13,17,18 are low-voltage (50–320 kV), low-gradient (~5 MV m−1) d.c. electrostatic electron guns. These guns provide an excellent vacuum at the level of 10−12 torr to ensure a long lifetime for the cathode’s quantum efficiency. Additionally, all of them operate at relatively low accelerating voltages to minimize dark current emission, which could otherwise compromise the system’s vacuum, increase ion back-bombardment and reduce the quantum efficiency lifetime.It is important to note that the KEK d.c. gun has been operated at voltages exceeding 500 kV (ref. 19) and was later used for sustained operation at voltages between 390 kV and 500 kV with a GaAs photocathode and quantum efficiency lifetime up to 17 h (ref. 20).It is well known that generating high-quality electron beams requires a high accelerating gradient as well as a high accelerating voltage, that is, sufficient beam energy at the gun exit21,22. In contrast with electrostatic guns, radio-frequency (RF) guns can operate with very high gradients and accelerating voltage measured in megavolts (MV). When combined with the negative affinity of GaAs photocathodes, this could enable the generation of electron beams with exceptionally low transverse emittances12. This is why the physics community has been seeking a method to operate GaAs photocathodes inside an RF gun as sources of polarized electrons for future colliders23,24,25,26,27,28. Unfortunately, all previous attempts to operate an RF gun with GaAs photocathodes failed29,30,31,32,33,34. In the best case, the cathode’s quantum efficiency survived only a few RF cycles30.There are three main challenges that cause most RF guns to operate only with very robust photocathodes, such as metal, Cs2Te or CsK2Sb.The first challenge is the relatively poor vacuum in most normal conducting (room-temperature) high-gradient RF guns, where high power loss in the cavity’s walls leads to outgassing32,33. This problem made normal conducting RF guns incompatible with GaAs photocathodes. By contrast, properly designed superconducting RF (SRF) guns, operating at cryogenic temperatures (from 2 K to 4.5 K), can provide vacuum conditions suitable for the long lifetime of GaAs photocathodes.The second challenge is to eliminate both multipacting—a resonant process in which an electron avalanche develops on the cavity surface35—and dark current, the uncontrolled cold field emission of electrons36,37,38. Both effects significantly accelerate the degradation of the photocathode’s quantum efficiency.The third challenge is the significant decline in photocathode quantum efficiency at the cryogenic temperatures typical for SRF guns39. To address this, a specialized system known as the cathode stalk is required to thermally insulate the photocathode from the gun cavity walls. This system, equipped with an RF choke, must ensure adequate electrical and thermal conductivity to maintain the photocathode at the required temperature. Failure to do so may lead to the evaporation of volatile materials, such as caesium (Cs), from the photocathode surface, which could degrade the performance of the SRF cavity.SRF electron gun at Brookhaven National LaboratoryAll these challenges were overcome in May 2024 when we successfully demonstrated the sustained operation of two GaAs photocathodes in our SRF gun (Fig. 1).Fig. 1: Layout of the CW CeC SRF gun system.The gun consists of a 113-MHz quarter-wave SRF niobium cavity (operating at 4 K and cooled by liquid helium), a photocathode storage and transfer system, and a room-temperature cathode stalk, which serves as a holder for the cathode and serves as a half-wave RF choke. RF power is supplied to the cavity through an FPC, a hollow coaxial tube that also fine-tunes the cavity frequency. Electrons are generated by illuminating the photocathode with green laser pulses (532 nm, 360 ps) delivered via an off-axis mirror in the laser cross. The electrons are accelerated (right to left) from the cathode in the cavity’s accelerating gap (~16 cm) and are focused by the gun solenoid located 0.65 m downstream of the cathode.Full size imageThe gun was constructed as part of the SRF accelerator for a coherent electron cooling (CeC) experiment at the Relativistic Heavy Ion Collider (RHIC)40,41,42,43 and was brought into operation 10 years ago. The successful operation with GaAs photocathodes was achieved after years of operation, resolving numerous challenges, implementing system improvements and developing dedicated operating modes. After finding a solution of how to overcome multipacting problems35 and developing methods of conditioning out dark current44, the gun started demonstrating reliable months-long operation with CsK2Sb photocathodes45 at a 1.25-MV accelerating voltage and generating world-brightest continuous-wave (CW) electron beams22.The months-long lifetime of the CsK2Sb photocathodes, which are sensitive to all three effects mentioned above, was a very encouraging indicator that the CeC SRF gun could sustain operation with the more demanding GaAs counterparts. However, several remaining challenges still posed a risk to a successful test.First, the ultrahigh-vacuum system used for cathode transfers was designed for operating with CsK2Sb cathodes, which is typically insufficient for the more sensitive GaAs. Furthermore, unlike typical d.c. polarized guns, where GaAs photocathodes are prepared in a system connected to the gun, our SRF gun is located in the RHIC tunnel—a radiation-protected area with restricted access—available only for a few hours once every 2 weeks. Our photocathodes are prepared at the laboratory located few kilometres away from the gun (Extended Data Figs. 1–4). They are then transferred in a portable vacuum chamber, which is sealed using a vacuum valve. The sealed chamber is transported to the RHIC tunnel, where it is connected to the load-lock chamber of the SRF gun’s cathode transfer system. After baking the load-lock chamber, the valve can be opened, allowing the photocathode to be transferred into the cathode stalk using three vacuum manipulator arms—one vertical and two horizontal—with a total transfer length of ~2.5 m. Before the start of the experiment, we performed a test (Extended Data Fig. 5a) to ensure that GaAs survives such transfer.These operations demand an extremely high vacuum. Achieving the required vacuum level involved multiple upgrades to both cathode deposition system and portable vacuum chamber. Additionally, the entire cathode transfer system’s vacuum setup had to be overhauled, including the installation of two 1,500 l s−1 non-evaporable getter pumps. We reached an acceptable—though not ideal—level of vacuum for GaAs cathodes.Second, our SRF gun was designed to operate with CsK2Sb photocathodes, which are deposited on the polished front surface of the molybdenum (Mo) puck with two side groves for the puck transfer in the system and the inner stub that is used to conduct the electric current and the heat induced in the cathode to the end effector and to the water-cooled stalk (Fig. 2).Fig. 2: SRF gun’s cathode system.a, The cathode puck is transferred into the operational position at the end of the cathode stalk, which is maintained at room temperature by circulating water. The stalk is separated from the body of the SRF gun’s niobium cavity operated at a temperature of 4 K. The stalk served as a half-wave RF choke transferring both electric current and heat induced in the cathode puck by the RF electromagnetic field. The cathode puck is maintained by the end effector attached to the long horizontal transfer arm. b, Cut-off of the end effector before connection to the cathode puck. It shows the inner RF fingers that serve as electric and thermal connections between the parts. The thin yellow lines show the centre line of the end effector and location of three spring plunger balls to clutch the cathode puck: only one ball is seen in the cut-off. c, Photograph of the cathode puck attached to the end effector during the transfer. The external RF fingers serve as electric and thermal connections between the end effector and cathode stalk.Full size imageExcellent heat and electrical current transfer to the stalk are critically important for the proper operation of the SRF gun and the photocathodes. Overheating of the cathode puck will result in the evaporation of the volatile photocathode material, which would result in an increase in the dark current or even arcing and complete failure of the gun. In addition, the deposition of the photocathode material inside the SRF gun will result in multipacting, which would destroy the quantum efficiency of the photocathode. Furthermore, poor current transfer could result in arcing, vacuum bursts and destruction of the photocathode.In contrast with CsK2Sb photocathodes, the GaAs photocathode uses a cesiated GaAs crystal substrate, which should be connected to the cathode puck both electrically and thermally. We modified two Mo pucks (Fig. 3a) to house thin (100–300 µm) GaAs crystals. The crystals were connected to the body of the puck using indium and were held in place by Mo cups that were crewed onto the pucks (Fig. 3b). The cups have openings with a 5-mm diameter used to activate the GaAs.Fig. 3: Modifications of the cathode puck for GaAs crystals.a, Parts of the GaAs cathode assembly comprising a polished titanium cap, and the modified body of a Mo cathode puck with a central cut to house the GaAs substrate. b, Assembled GaAs cathode placed in the holder for activation in the cathode preparation system.Full size imageThird, we took every possible measure to ensure good thermal contact between the GaAs substrate and the Mo puck. We also conducted simulations to assess the heat generated in the substrate and its resulting temperature rise. The simulation predicted 5 W of RF losses in the GaAs substrate when operating the gun at its nominal voltage of 1.25 MV. Details are provided in the Methods and Extended Data Fig. 6.Our standard turn-on routine of the SRF gun, used with CsK2Sb photocathodes, involves a rapid initial voltage jump to approximately 200 kV (to prevent vacuum excursions caused by multipacting at low voltages) followed by a fast voltage ramp to 1 MV. However, previous failures with GaAs cathodes in RF guns introduced enough uncertainty that we chose not to attempt immediate operation at high voltages. Instead, we developed operation modes at lower voltages, starting as low as 52 kV, maintaining excellent vacuum conditions. Additional processing of the system was performed (Methods and Extended Data Fig. 7a) to establish multipacting-free operations below 200 kV. We began our GaAs photocathode test at 52 kV, where estimated heat losses in the crystal were below 1 mW, and then gradually increased the operating voltage.Fourth, one of the important advantages of the SRF gun is that it provides an excellent vacuum. Its body operated at a temperature of 4 K, serving as a very powerful cryogenic vacuum pump that freezes all gases including hydrogen. The frozen gasses attach to the surface and can be released only when the temperature rises to the evaporation level. The situation is similar for cold-to-room-temperature transition in the front and back of the SRF gun cavity. The transitions are cooled by helium gas evaporated by the SRF cavity. The helium gas propagates through channels, attaches to the transitions and gradually warms up to room temperature. The temperature gradient results is the distribution of frozen residual gas species, originating in a warm section of the accelerator, along the surface of the transition according to their temperatures of freezing. Variations in the helium pressure would result in a variation in its flow, changes in the temperature profile and corresponding vacuum pressure bursts (Extended Data Fig. 7b). Such bursts were observed in our SRF gun system and were acceptable for operation with CsK2Sb photocathodes but would be intolerable for GaAs. This was the reason that our cryogenic group developed the mode of operating the cryogenic system that prevented variations in the helium gas flow and eliminated the vacuum bursts.Operation of SRF gun with GaAs photocathodesWith the completed upgrades and a pause in the CeC schedule, we demonstrated sustained SRF gun operation with a GaAs photocathode: it was limited to 20 days, after which the facility resumed the regular CeC program.Two GaAs photocathodes (referred to as #1 and #2 hereafter) with an initial quantum efficiency of 4.8% in green (532 nm) light—the laser wavelength used for all of the reported quantum efficiency measurements—were produced in March 2024. The vacuum suit with these cathodes was moved to the RHIC tunnel and connected to the SRF gun’s cathode transfer system on 2 April 2024. After finishing the bake-out of the load-lock system on 8 April 2024, we performed our planned ‘survival test’ by transferring cathode #1 out of the vacuum suit to the vertical transfer arm, lowered to the tip of the long horizontal transfer arm and back to the suit. The cathode survived but its quantum efficiency dropped to its estimated value of ~1% (Extended Data Fig. 5a).The operating modes of the SRF gun were first verified with a blank Mo puck. Cathode #1 was transferred into the gun on 15 May 2024 (Fig. 4), and a sustained operation up to 1 MV was demonstrated. To confirm reproducibility, cathode #1 was replaced with cathode #2 on 22 May 2024. The quantum efficiency evolution and maps of both cathodes are shown in Fig. 5 and discussed below.Fig. 4: Transfer of GaAs cathode #1 into the SRF gun.a, Vacuum levels (in torr) recorded by a cathode manipulator system with three ion pumps during the 2-h transfer on 15 May 2024. The pressure is shown in orange for the vertical manipulator, sea green for the inside of the cathode stalk and purple for the outside of the cathode stalk. The integrated pressures measured at these locations were 1.35 × 10−7, 9.4×10−7 and 1.3 × 10−8 torr s. Because the cathode moved between these locations, we estimate its net exposure to vacuum spikes to be on the order of ~10−7 torr s. b, Photograph of GaAs cathode #1 inserted into the SRF gun. The shiny cathode cap is visible (20-mm diameter with a 5-mm opening). The surrounding circle is the tip of the cathode stalk, whereas the gun cavity nose is seen further in the background. c, The same image under scattered laser illumination, showing the metal cup and the 0.5-mm-diameter laser spot centred on the open GaAs photocathode area. This configuration was used to measure the quantum efficiency map of the cathode.Full size imageFig. 5: Measured quantum efficiency evolution and quantum efficiency maps of GaAs photocathodes.a, Evolution of quantum efficiency (QE) during the SRF gun operation. Red dots and the solid line represent cathode #1, whereas blue dots and the solid line represent cathode #2. The blue dotted line shows an exponential fit to the quantum efficiency data for cathode #2, yielding a decay time of 131 ± 23 h. The two arrows indicate times when significant vacuum spikes were measured in the FPC area in front of the gun; due to the nature of the SRF gun, it is impossible to measure the pressure inside the cryogenic niobium cavity. The integrated pressures were 1.6 × 10−6 and 7 × 10−7 torr s for cathodes #1 and #2, respectively. The quantum efficiency was measured at the cathode centre using a hexagonal laser spot of 3.4-mm diameter. Error bars include variations in the laser power, laser profile and non-homogeneity of the quantum efficiency. b,c, Measured quantum efficiency maps of GaAs photocathodes #1 (b) and #2 (c). The photoemission area is limited by the 5-mm circular opening of the cathode cup. The colour scale shows relative units, with a peak quantum efficiency of 0.1%. The high-resolution quantum efficiency image of photocathode #2 is shown in Extended Data Fig. 5c.Full size imageThe quantum efficiency of cathode #1 was ~0.5% before transfer into the gun. The cathode was first moved vertically by a manual system, then placed on the end effector of a long horizontal arm and slowly inserted into its final position with a computer-controlled system, a procedure necessary to minimize vacuum spikes from friction. Initial low-charge measurements showed the quantum efficiency dropped to 0.04% during transfer. Since quantum efficiency degradation in GaAs depends on both pressure and gas composition, this 12.5-fold reduction is consistent with exposure to ~1.5 × 10−7 torr s of reactive gases, in line with prior results46. This demonstrates that the present transfer system—designed for more robust photocathodes—does not meet the GaAs requirements; future polarized SRF guns will need a more reliable, possibly simpler, transfer mechanism with local activation capability.Energizing the gun to 52 kV produced no vacuum spikes, and illumination with the green laser generated the first electrons (~100 pC per bunch). The voltage, measured using the method described in ref. 47 (Extended Data Fig. 8a), was then gradually raised to 1 MV and monitoring vacuum, dark current and helium consumption. At 500 kV, quantum efficiency mapping revealed a twofold spatial variation with a local maximum of ~0.1% (Fig. 5b).In particular, during this period, the quantum efficiency increased rather than degraded. The quantum efficiency of cathode #1 initially increased from 0.04% to 0.093% during the first 2 days of SRF gun operation, a behaviour also observed with CsK2Sb photocathodes. This rise is attributed to the removal of contamination accumulated during transfer, possibly by X-rays or extreme-ultraviolet radiation generated in the gun. After 50 h, a vacuum excursion of 1.6 × 10−6 torr s in the fundamental power coupler (FPC) caused the quantum efficiency to drop by a factor of two, followed by a gradual decay with an 80-h e-fold time until the cathode exchange (Fig. 5a).The quantum efficiency of cathode #2 could not be measured before transfer. During transfer, the manipulator recorded an integrated pressure of 8.5 × 10−7 torr s, and the initial quantum efficiency was 0.12%. This cathode was used to test higher-voltage operation and monitor the quantum efficiency, quantum efficiency maps and beam parameters. It operated successfully up to 1.375 MV (Fig. 6a,b) without any quantum efficiency degradation attributable to voltage. With sufficient laser power, the maximum extracted charge showed a nearly linear dependence on gun voltage, with a slope of 1.2 nC MV−1 (Extended Data Fig. 5b); at voltages above 1 MV, bunch charges routinely exceeded 1 nC (Fig. 6c).Fig. 6: Parameters of the SRF gun and generated electron beam during the experiment.a, Measured SRF gun voltage during tests with two GaAs photocathodes: the first cathode was used from 15 to 21 May 2024, and the second cathode was operated from 22 May to 4 June 2024. After the 1.4-MV quench event, the operation was resumed at the standard voltage of 1.25 MV. b, Typical measurements of the electron-beam kinetic energy as a function of the laser phase, with the SRF gun operating at 1.35 MV. The error bars indicate the uncertainty in beam energy measurement at each of the 16 RF phase settings for the laser pulse arrival. The solid line is Vcos(φ) fit of the data. Details of the energy measurements are provided in the Methods, with further explanation available in ref. 47. c, Measured charge per bunch using GaAs photocathode #2 on 25 May, and operating the SRF gun between 0.9 and 1.07 MV.Full size imageThe quantum efficiency evolution of cathode #2 was generally more stable than cathode #1, showing a mostly gradual decline (Fig. 5a). Decay was not uniform: the quantum efficiency remained nearly constant between 35 and 70 h of operation and partially recovered after a 40% drop caused by a vacuum spike, followed by ~40 h of stable quantum efficiency at gun voltages above 1 MV.The experiment concluded on 1 June 2024. When a significant a.c. power dip caused multiple equipment failures, resulting in a vacuum excursion of ~4 × 10−5 torr s and a 17-fold drop in GaAs quantum efficiency a few days later, GaAs cathode #2 was replaced with a standard CsK2Sb photocathode.Possible future stepsAlthough we successfully demonstrated more than a threefold increase in accelerating voltage and achieved a reasonable, proof-of-principle quantum efficiency lifetime for two GaAs photocathodes, several challenges remain: Cathode transfer vacuum limitations: the vacuum level in the cathode transfer system was insufficient to preserve the GaAs quantum efficiency. Each transfer caused substantial quantum efficiency reduction and introduced significant non-uniformity. Laser system constraints: the laser system’s performance was inadequate for a full evaluation of the generated electron-beam quality (Methods and Extended Data Fig. 8b,c). The CeC SRF gun currently operates using liquid helium supplied by the RHIC cryogenic system. RHIC is scheduled to cease operations at the end of 2025 to allow the construction of the Electron-Ion Collider. We are presently midway through a 6-month run focused on generating electron beams for the CeC experiment. Although the laser system with 360-ps flat-top pulses is now fully operational, the likelihood of conducting additional SRF gun tests with GaAs photocathodes before RHIC shutdown is low.We are considering relocating the SRF gun to Stony Brook University to establish an advanced SRF gun research laboratory. This would require constructing a 4-K cryogenic system, but it would also enable the development of a GaAs cathode preparation and transfer system with an ultrahigh vacuum. If realized, this project could, within 3–4 years, open a new chapter in SRF gun studies with GaAs photocathodes, including the generation of polarized electron beams.Unfortunately, the operational Rosendorff SRF gun48 and its improved copy at Humboldt-Universität zu Berlin have experienced significant issues even with robust photocathodes, making sustained operation with GaAs unlikely. By contrast, a new SRF gun under construction at Michigan State University for the LCLS-II CW X-ray FEL49 could potentially serve as the next platform for advancing GaAs photocathode technology.MethodsGaAs photocathodes preparationGaAs puck design and activation procedureThe GaAs puck was designed to maintain full compatibility with the existing infrastructure used for growing and transporting alkali antimonide photocathodes to the SRF gun. To enable the use of GaAs-based photocathodes and shielding the sharp edges of the sample from the gun’s internal electric fields, the puck consists of two Mo components.The main body of the puck includes one of the two grooves needed for sample manipulation and features a pocket on its top surface to hold a small 7 × 7 mm2 GaAs sample. The second groove is provided by a cap that screws on top of the body. This cap, mechanically polished to a mirror finish, has a 5-mm-diameter hole in its centre, allowing the GaAs crystal surface to be exposed to Cs, O and Te during activation, as well as to laser illumination for electron extraction (Fig. 3).Before final assembly, both body and cap are high-pressure rinsed to remove particulates and then sealed in a dedicated transporter in a cleanroom. The GaAs crystals are chemically etched to remove surface oxides, followed by rinsing with deionized water. The assembly of the GaAs sample into the puck also takes place in the cleanroom. A thin indium foil (30 µm thick) is inserted between the GaAs crystal and the puck body. During the heat cleaning process, this indium foil melts, improving thermal conductivity between the GaAs crystal and the puck.Preliminary experiments on GaAs activation and lifetimePreliminary experiments were conducted to estimate the photocathode lifetime within our growth system and to identify the dominant mechanisms contributing to quantum efficiency degradation under the vacuum conditions of our activation chamber. These conditions range from a steady-state vacuum of ~1 × 10−11 torr to ~1 × 10−10 torr during activation.A GaAs sample with p-type doping of 1 × 1019 cm−3 was chemically etched with 35% HCl, rinsed with deionized water and then transferred to the vacuum system. The sample was heat cleaned at 580 °C for 2 h to remove surface oxides and then twice activated to negative electron affinity using Cs and oxygen. Quantum efficiency evolution was measured using a green laser (532 nm) at both 100% and 1% duty cycles.These experiments aimed to determine whether the observed quantum efficiency lifetime was primarily affected by chemical poisoning or by ion back-bombardment. Given that the photocathode bias was maintained at –50 V and the average extracted current was in the microampere range, ion back-bombardment could not be ruled out as a significant factor. The results (Extended Data Fig. 1) indicate that similar quantum efficiency lifetimes were observed over time, suggesting that further investigation is required to isolate the dominant degradation mechanism.When the same quantum efficiency data are plotted as a function of the extracted charge in Extended Data Fig. 1 (right), it becomes evident that quantum efficiency decay occurs much more rapidly—by more than a factor of 50—when the sample is operated with a 1% duty cycle compared with continuous (100%) illumination. This strongly suggests that under our experimental conditions, chemical poisoning is the dominant mechanism responsible for quantum efficiency degradation, rather than ion back-bombardment.Preliminary experiments were also performed using a modified activation procedure involving a tellurium (Te) interlayer between two Cs–O layers. The GaAs sample was initially activated by alternating exposures to Cs and oxygen. After reaching the maximum quantum efficiency, a slight excess of Cs was deliberately deposited by continuing the Cs exposure, which resulted in a small decrease in quantum efficiency (Extended Data Fig. 2).It was expected that the excess Cs would gradually react with residual oxygen in the vacuum chamber over the following hours, leading to a further increase in quantum efficiency. This effect was indeed observed over the next 2 days, as quantum efficiency measurements taken every 24 h confirmed a gradual improvement (Extended Data Fig. 3).All measurements described so far were conducted with the photocathode inside the activation vacuum chamber. To evaluate the photocathode’s robustness against vacuum degradation—particularly from mechanical motion—we attempted to transfer the activated photocathode into the preparation chamber of the activation system. During the transfer, vacuum levels degraded rapidly to the range of 10−9 torr due to outgassing from the manipulator’s bellows. After the transfer was completed, the quantum efficiency was measured again and found to have dropped to approximately 1.5% (Extended Data Fig. 3).The quantum efficiency of the GaAs photocathode also exhibited degradation with a decay time of approximately 1 h when the magnetic manipulator was used to transfer the photocathode from the preparation chamber to the docking chamber, in preparation for insertion into the portable vacuum chamber (we call it a garage). This portable chamber is used to transport the activated photocathode to the SRF gun’s loading system. However, since the transfer typically takes only a few minutes, the resulting decrease in photocurrent was minimal and considered negligible.These measurements enabled us to determine the optimal maximum speed at which the linear manipulators can be operated to minimize quantum efficiency loss due to the transient gas load released during the transfer process.Photocathodes in the SRF gunFor operation in the SRF gun, two GaAs photocathodes were prepared using the activation method involving a Te interlayer, as described above. This technique produced photocathodes with very similar final quantum efficiency values, each reaching approximately 4% following the intentional deposition of a slight excess of Cs at the end of the activation process (Extended Data Fig. 4).Both photocathodes were transferred from the activation chamber to the preparation chamber and subsequently to the garage docking station in a carefully controlled manner, ensuring that vacuum levels remained below the range of 10−10 torr throughout. This procedure successfully preserved the quantum efficiency of both cathodes, with no significant degradation observed after the transfer to the preparation system.Once in the garage chamber, quantum efficiency measurements were performed using a collecting anode that could be positively biased. However, due to the geometric configuration—where the anode is positioned to the side of the cathodes—and the large incidence angle of the laser beam, accurate quantum efficiency estimation was not possible at this stage. Nevertheless, given the short transfer time, consistently low vacuum levels (with the garage chamber maintained in the range of 10−11 torr) and controlled handling, we assumed that the quantum efficiency of both photocathodes remained largely unaffected during the transfer.The ultrahigh-vacuum garage chamber was then transported to the RHIC tunnel and connected to the load lock of the SRF gun’s cathode loading system. Immediately following the connection, photoemission from the photocathodes was confirmed, with photocurrent values comparable with those measured before transporting to the garage.The load-lock system of the SRF gun was vacuum baked out at approximately 150 °C for 1 week following connection. After the bake out, one of the two photocathodes was transferred from the garage into the SRF gun’s loading chamber, where its quantum efficiency was measured and found to have decreased to around 1%, with an estimated 1/e decay time of approximately 183 h (Extended Data Fig. 5a). The photocathode remained in the loading chamber, under a vacuum of ~1 × 10−11 torr, for about 37 days as preparations for the SRF gun’s beam operation progressed.Immediately before being loaded into the SRF gun, the quantum efficiency of the first cathode was measured again and had decreased further to approximately 0.5%, corresponding to an estimated 1/e decay time of about 1,281 h (Extended Data Fig. 5a). Unfortunately, the quantum efficiency of the second GaAs photocathode could not be measured at the time of its exchange on 22 May 2024. The quantum efficiency evolution of both cathodes during SRF gun operation is shown in Extended Data Fig. 5b. Quantum efficiency was measured at the cathode centre using a hexagonal laser spot with a 3.4-mm diameter.We had sufficient laser power to saturate the cathode and observed a nearly linear relationship between the maximum extracted charge and the gun’s accelerating voltage (Extended Data Fig. 5b). For a laser spot with a 5-mm diameter, the slope of this curve was 1.2 nC MV−1, and at voltages above 1 MV, bunch charges exceeding 1 nC were routinely generated.All quantum efficiency measurements were performed using a green laser (λ = 532 nm) with the charge per bunch kept significantly—at least a factor of two—below saturation. On the gun laser table, the laser beam was transversely collimated using a remotely controlled iris with eight blades, producing an octagonal transverse profile. The optics transferred the beam one to one onto the photocathode surface. For consistency, all quantum efficiency measurements used a 3.4-mm iris opening, with the laser spot centred within the 5-mm cathode opening (Fig. 4b).Laser power, Pl, was measured at the SRF gun’s laser table using a power meter located adjacent to the laser injection point. The charge per bunch, Q, was measured using an integrated current transformer (a standard diagnostic in our accelerator system50) and cross-calibrated against a Faraday cup to ensure accuracy. The measured Q and Pl were then used to determine the quantum efficiency of the laser-illuminated photocathode area using the relation$$\mathrm{QE}=\frac{{N}_{{\rm{e}}}}{{N}_{\mathrm{ph}}},\,{N}_{{\rm{e}}}=\frac{Q}{e},\,{N}_{\mathrm{ph}}=\frac{1}{\hslash \omega }\times \frac{{P}_{{\rm{l}}}}{{f}_{{\rm{l}}}},$$where fl is the laser repetition frequency of 78 kHz and \(\hslash \omega =2.33\,\mathrm{eV}\) is the photon energy of the 532-nm laser. The main source of error in the quantum efficiency measurements arose from laser instabilities, including significant variations in the transverse and temporal pulse profiles (‘Problems with laser pulses’ section). These fluctuations, combined with non-uniformities in the cathode quantum efficiency map, could alter the average quantum efficiency within the 3.4-mm laser spot. In particular, we also observed an increase in quantum efficiency from 0.031% to 0.04% when the RF voltage was raised from 0.6 MV to 1.1 MV, which we attribute to a rise in the GaAs substrate temperature, probably reducing the photocathode’s work function. We estimate the maximum relative errors in quantum efficiency measurements to be approximately 25% for cathode #1 and 15% for cathode #2.Although we were able to quantitatively measure the quantum efficiency over relatively large illuminated areas (on the order of several square millimetres), high-resolution quantum efficiency maps with a laser spot size of ~0.5 mm (Fig. 5) were acquired in relative units, calibrated to the maximum quantum efficiency observed within each scan. The laser spot was scanned across the photocathode using a two-dimensional motorized stage at the gun laser table. Our integrated current transformer lacked sufficient sensitivity for these measurements; therefore, the relative intensity of the generated electron beam was determined from the integrated signal on a yttrium aluminium garnet (YAG) profile monitor.In addition, we captured images of photocathode emission using the YAG profile monitor (Extended Data Fig. 5c) with a low-charge, high-repetition-rate electron beam. Although these images were not perfect, they provided additional information about the spatial quantum efficiency distribution on the GaAs photocathodes. Uniform illumination of the entire photoemitting area revealed quantum efficiency degradation in regions that had been used to generate the electron beam over several days.We also observed approximately 20 small (~100-μm diameter) randomly distributed spots exhibiting zero quantum efficiency. These localized defects could not be resolved using the standard quantum efficiency scan, which has a spatial resolution of ~0.5 mm. The YAG images, with significantly higher spatial resolution than the laser scans, allowed the visualization of these ~100-μm defects. A darker octagonal spot in the centre of the image corresponds to the 3.4-mm image of the iris used to collimate the laser beam during normal electron-beam operation. Analysis of the images showed that illuminating the cathode and generating the electron beam resulted in an additional ~17% reduction in quantum efficiency relative to the surrounding area. This additional loss may be either due to ion back-bombardment or due to a temperature rise of ~1 °C caused by laser power dissipation and heat generated by the emitted electrons.Quantum efficiency evolution was also influenced by operational mishaps, including gun voltage drops, errors in ramping the voltage and vacuum excursions. The improved quantum efficiency lifetime observed with the second photocathode reflects the SRF gun’s enhanced control systems and the avoidance of failures encountered during the initial learning curve with the first cathode.On the final day of operation, an a.c. power dip triggered multiple equipment failures, resulting in an FPC vacuum excursion with an integral of 3.8 × 10−5 torr s and a drop in quantum efficiency to the 10−4 range. With insufficient laser power to extract more than a few tens of picocoulombs, the GaAs photocathode was subsequently replaced with a standard alkali antimonide photocathode.Preparations of SRF gun for operating with GaAs photocathodesPossible overheating of GaAs photocathodes, leading to the contamination of the SRF gun’s cavity by Cs, posed a significant risk and required a cautious approach. To mitigate this, we developed a detailed procedure that began with operating the gun at the lowest possible voltage, followed by a gradual voltage increase and closely monitoring vacuum and radiation conditions in the gun area.Our SRF gun utilizes cathodes recessed by 10.5 mm from the cavity nose to provide the necessary focusing for the generated electron beams. This configuration reduces the peak electric field at the cathode surface by a factor of two compared with a non-recessed cathode (Extended Data Fig. 6a). To evaluate potential heating, we first simulated temperature increases due to dielectric losses in the GaAs crystal at the SRF gun’s operating frequency of 113 MHz, as well as resistive losses in the Mo puck (Extended Data Fig. 6b).We used p-doped GaAs crystals with thicknesses of 100 μm and 300 μm for cathodes #1 and #2, respectively. Since specific data on GaAs dielectric losses at the 113-MHz frequency of the SRF gun are not available, we extrapolated existing data to estimate \(\tan {\delta }{\approx }{10}^{{-}4}\). The resulting estimate predicted that GaAs dielectric losses of approximately 0.5 W for the 100-μm crystal and ~1 W for the 300-μm crystal dominate the overall losses.All the operating temperatures remained well below the melting points of other photocathode components; however, Cs evaporation and subsequent deposition on the SRF gun’s cavity surfaces could increase the dark current and trigger cavity quench events, as observed during the experiment.Due to uncertainty regarding potential cathode overheating, gun operation was initially started at previously unexplored voltages below 100 kV. On the basis of our prior experience35, there was a substantial risk of multipacting in the SRF gun’s FPC region at voltages below 200 kV, particularly in the presence of magnetic fields from the focusing solenoids. Such multipacting would lead to increased vacuum pressure, which, in turn, would degrade the quantum efficiency of the GaAs cathodes.To address this, we dedicated the week of 8–14 May 2025 to clean up the potential multipacting zones and suppress the associated vacuum spikes in the FPC (Extended Data Fig. 7). This was accomplished by performing periodic bipolar scans of the solenoid current, ramping it to the maximum values required for each specific SRF gun voltage (Extended Data Fig. 7a).In parallel, the RF group developed a new gun turn-on procedure featuring a flexible voltage target. As a result of these efforts, we successfully established multipacting-free operation of the SRF gun across a wide voltage range—from 52 kV to 1 MV—with FPC vacuum levels in the low range of 10−10 torr.As described in the main text, the SRF gun’s cavity transitions—from room temperature to cryogenic temperatures—are cooled by evaporated helium gas. This gas travels through a spiralling tube surrounding the transition vacuum chamber, moving from the cold end to the room-temperature end. The resulting temperature gradient causes residual gas components from the warmer regions to condense onto the vacuum chamber walls at which their respective freezing points are reached. Any reduction in the helium gas flow leads to an increase in the wall temperature and the release of previously frozen gases—a phenomenon illustrated in Extended Data Fig. 7b. Although such gas bursts were tolerable during operation with CsK2Sb photocathodes, they were unacceptable for use with GaAs. The RHIC cryogenics team refined the SRF gun’s cryo-system controls to eliminate fluctuations in helium flow, thereby preventing pressure spikes and ensuring stable vacuum conditions.With all preparations complete, we proceeded with testing GaAs photocathodes in the SRF gun. The initial test began at an SRF gun voltage of 52 kV, which ensured negligible power losses—on the order of milliwatts—and a subkelvin temperature rise in the GaAs crystals.Measuring parameters of electron beamWe used well-established methods to measure key electron-beam parameters from the SRF gun, including beam energy, bunch charge and beam emittance.To measure the electron-beam energy, we used a precise method described in ref. 47, which is based on the principle that a solenoid rotates the plane of particle motion by an angle, that is,$$\theta =\frac{e}{{2pc}}{\int }_{-\infty }^{\infty }{B}_{z}\left(z\right){\rm{d}}s{,}$$where p is the electron momentum, e is the electron charge and c is the speed of light. The rotation angle θ is proportional to the integral of the longitudinal magnetic field Bz, which is, in turn, determined by the number of turns N in the solenoid coil and the coil current I, in accordance with Stokes’ theorem:$${\int }_{-\infty }^{\infty }{B}_{z}\left(z\right){\rm{d}}s\equiv \oint \vec{B}{\rm{d}}\vec{r}=\frac{4{\rm{\pi }}}{c}NI\text{.}$$Our setup consists of a calibrated solenoid with a known number of coil turns, a pair of horizontal and vertical dipole trims placed upstream of the solenoid and a YAG profile monitor downstream. We perform four measurements by scanning the electron beam horizontally and vertically using both positive and negative solenoid currents, and tracking the beam centroid on the YAG screen (Extended Data Fig. 8a).The angles between the beam trajectories produced by horizontal and vertical scans for opposite solenoid currents, namely,$$\Delta \theta =\frac{4{\rm{\pi }}e}{p{c}^{2}}NI,$$allow us to determine the beam momentum as well as the full and kinetic energies (E and Ek, respectively) of the electrons as well as the voltage of the gun (where m is the mass of the electron):$$pc=\frac{4{\rm{\pi }}e}{c}\times \frac{NI}{\Delta \theta },\,E=\sqrt{{m}^{2}{c}^{4}+{p}^{2}{c}^{2}},\,{E}_{{\rm{k}}}=E-m{c}^{2}.$$Although it is possible to use a single-dipole trim for this measurement, combining results from both horizontal and vertical scans helps cancel out astigmatism effects inherent to real-world profile monitor systems, thereby improving the measurement accuracy. This method typically achieves a relative energy measurement accuracy of ~10−3.Problems with laser pulsesUnfortunately, in our effort to optimize all other experimental conditions for the GaAs photocathodes, we failed to identify a critical issue with the laser’s temporal profile before the start of the experiment. The system designed to combine six Gaussian sub-pulses (beamlets) with alternating polarization into a single, nearly flat-top pulse of 360-ps full-width at half-maximum malfunctioned. Instead of producing a uniform pulse, it generated a signal composed of beamlets with nearly random amplitudes.We discovered the faulty temporal structure of the laser pulses (Extended Data Fig. 8c) only after completing the GaAs photocathode tests. Subsequent measurements revealed that the relative intensity of individual beamlets varied by more than an order of magnitude. It is also likely that the pulse structure drifted during the experiment, with the pulse occasionally dominated by only one or two beamlets. Furthermore, the beamlets were misaligned transversely, causing variations in laser power density across the cathode surface.Our beam dynamics simulations indicated that these temporal and transverse laser-structure issues dramatically increased the projected (not slice) emittance, making conventional emittance compensation techniques insufficient for accurate emittance measurements. In addition, the combination of transverse laser variations and strong inhomogeneity in the cathode quantum efficiency (Fig. 5) led to measurement errors in the average quantum efficiency presented in Fig. 5a.The laser malfunction prevented the optimization of beam dynamics and the achievement of low emittance values typically observed in our SRF gun21. Simulations indicated that the combination of incorrect temporal and transverse laser profiles would significantly increase the projected emittance at the YAG monitor location used for emittance measurements.