Electron beam characterization of a combined diode rf electron gun

Experimental and simulation results of an electron gun test facility, based on pulsed diode acceleration followed by a two-cell rf cavity at 1.5 GHz, are presented here. The main features of this diode-rf combination are: a high peak gradient in the diode (up to 100 MV=m) obtained without breakdown conditioning, a cathode shape providing an electrostatic focusing, and an in-vacuum pulsed solenoid to focus the electron beam between the diode and the rf cavity. Although the test stand was initially developed for testing field emitter arrays cathodes, it became also interesting to explore the limits of this electron gun with metallic photocathodes illuminated by laser pulses. The ultimate goal of this test facility is to fulfill the requirements of the SwissFEL project of Paul Scherrer Institute [B.D. Patterson et al., New J. Phys. 12, 035012 (2010)]; a projected normalized emittance below 0:4 m for a charge of 200 pC and a bunch length of less than 10 ps (rms). A normalized projected emittance of 0:23 m with 13 pC has been measured at 5 MeV using a Gaussian laser longitudinal intensity distribution on the photocathode. Good agreements with simulations have been obtained for different electron bunch charge and diode geometries. Emittance measurements at a bunch charge below 1 pC were performed for different laser spot sizes in agreement with intrinsic emittance theory [e.g. 0:54 m=mm of laser spot size (rms) for Cu at 274 nm]. Finally, a projected emittance of 1:25þ = 0:2 m was measured with 200 pC and 100 MV=m diode gradient.


I. INTRODUCTION
In the context of the SwissFEL project [1][2][3], an electron gun test stand has been developed and built to generate, transport, and characterize low emittance electron beams.This diode-rf gun was initially [4] designed in order to use field emitter arrays (FEA) [5,6] as an electron source for an FEL.While the development of FEA is still going on at Paul Scherrer Institute (PSI), the diode-rf gun has been characterized with a metallic photocathode and a laser as a trigger source.This paper summarizes the performance of this diode-rf gun when used with metallic photocathodes illuminated by vacuum ultraviolet laser pulses.
In an electron gun the beam emittance results from three main contributions: (1) the intrinsic emittance due to the electron emission process, (2) the contribution due to space charge forces, and (3) the contribution due to external fields [radio frequency (rf); electrostatic, . ..].To reduce the intrinsic emittance, one can either change the laser wavelength to match the work function of the emitting material or reduce the diameter of the emitting area.The latter results in stronger space charge forces which can again increase the emittance.Space charge emittance degradation can be reduced by using a stronger accelerating gradient.The gradient is limited however by vacuum breakdown.Finally, nonlinear external fields, such as rf field curvature in rf photoinjector cavities or the nonlinear radial fields near the anode iris of dc guns, also lead to emittance growth.As in other diode-rf gun combinations [7][8][9][10][11], the goal is to reduce all three of these contributions.
In our dc gun the electric field gradient is much lower than in other pulsed dc-rf guns (1 GV=m) [9,10,12] and comparable to rf photoinjector accelerating gradient (100 MV=m) [13,14].In our design 100 MV=m-500 kV is reached in the diode without conditioning and no dark current is detected downstream of the anode.In addition, the cathode geometry is shaped such that it produces nonlinear focusing forces which compensate the nonlinear defocusing forces due to space charge and radial field at anode iris.Another feature of the presented diode-rf gun design, is the presence of an in-vacuum pulsed solenoid which has been installed between the diode and the rf cavity in order to match the beam to the rf acceleration and preserve emittance.Finally, thanks to a wavelength tunable laser, we have demonstrated the possibility of decreasing the intrinsic emittance [15].
A detailed description of the test facility is presented in Sec.II and simulations of the beam dynamics in the gun are presented in Sec.III.Experimental results and discussion are presented in Sec.IV.The ultimate goal of this test stand is to provide electron bunches of 10 and 200 pC charge at 5 MeV with projected normalized emittances below 0.25 and 0:4 m, respectively, in order to fulfill the SwissFEL parameter requirements [3].

A. The pulsed diode acceleration section
The main emittance growth takes place in the diode section of the gun during the initial acceleration due to PHYSICAL REVIEW SPECIAL TOPICS -ACCELERATORS AND BEAMS 13, 093502 (2010) 1098-4402=10=13(9)=093502 (12) 093502-1 Ó 2010 The American Physical Society nonlinear space charge forces and nonlinear defocusing forces near the inner surface of the anode iris.High electric gradient and electrostatic transverse focusing are required in order to compensate partially the space charge effects and the nonlinear defocusing field around the anode iris.
The high voltage applied across the diode is pulsed (250 ns FWHM) with a 10 Hz repetition rate in order to reduce the probability of breakdown [16].The cathode geometry can be exchanged between a simple elliptical shape used for electrical gradient tests and a so-called ''hollow cathode'' geometry (see Figs.The maximum electric field between the electrodes is limited by vacuum breakdown, which can damage the photoemitting surface.In the case of a coated electrode, one breakdown is sufficient to destroy the coating such that not even a small gradient ( > 20 MV=m) can be sustained.
In consequence, our electrodes are not conditioned and no dark current is observed after the anode iris (down to the detection limit corresponding to 1 pC in 250 ns).An empirical study of electric field strength for different materials and surface finish has been performed and is partially summarized in Table I.Except for Mo and DLC, the gradient values are taken from [17].For this study a simplified elliptically shaped cathode and anode were used.The high gradient test procedure consists of three steps: (1) at a gap of 1 mm, the gradient is increased to 50 MV=m, (2) the gap is opened up to 7 mm while the gradient is kept constant and the voltage is increased from 50 kV to 350 kV, and (3) gap is reduced at a constant voltage of 350 kV until it breaks down.
Each material or coating presented in Table I was tested with at least five different pairs of electrodes to check the reproducibility.The metallic surfaces were hand polished with diamond paste and spray cleaned with dry ice (CO 2 snow) before installation [17,18].The Mo coating was deposited in house by sputtering and the diamond like carbon (DLC) coating was deposited by Bekaert SA [19].DLC gave very good results: more than eight DLC electrode pairs were able to sustain a surface electric field of at least 150 MV=m (more details can be found in [6]).Consequently, DLC became our material of choice to coat the hollow cathode geometry presented in Fig. 2. The emitting area (the inserted disk), however, is not coated and sees a surface field of about the half of the peak diode accelerating field.
Simulations of the hollow cathode geometry using the code POISSON [20] (Fig. 2, right) show that the electric field on the axis of the insert surface is about half of the maximum surface field present on the DLC coated holding lips and does not change by more than 10% for different anode-cathode gaps (1.5 to 6 mm).Since this geometry provides transverse focusing, the positioning and orientation of the cathode (red arrows on Fig. 2) allows the control of the beam trajectory and shape before it enters the rf cavity for further acceleration.The cathode longitudinal, transverse, and tilt positions are remotely controlled by  stepper motors (rotation over a few mrad around the cathode axis is also possible but not used here).The insert can also move longitudinally independently of the cathode holder.

B. The rf cavity and the diagnostic line
Because of mechanical constraint, the anode is separated from the rf cavity entrance plane by a drift distance of d anode-rf ¼ 166 mm.To prevent an expansion of the beam during this drift and to match the beam size to the rf accelerating cavity [21], a 25 mm long pulsed solenoid is located 51 mm downstream of the anode iris.The pulsed solenoid (Fig. 2, left) consists of a coil concentric with an inner ring of 1 cm diameter [22].Current pulses of up to 2 kA and 100 s duration are applied to the outer coil inducing a current pulse in the inner ring, leading to a maximum magnetic field of B PSL ¼ 320 mT on axis.This compact pulsed focusing technology was chosen in order to fit the available space between the diode and the rf cavity.No residual magnetic field is present on the cathode surface due to the eddy currents in the materials surrounding the pulsed solenoid.The two-cell rf cavity [23] (Fig. 1) has a frequency of 1.5 GHz and is fed through a 50 Ohm coaxial coupler with an rf forward power P rf of up to 5 MW with 5 s long pulses, yielding to a maximum accelerating gradient of 45 MV=m.The maximum accelerating gradient used in the cavity was 42 MV=m.A standard 1.5 GHz power splitter was designed to equally distribute the rf power from the klystron between the two waveguides leading to the coaxial coupler.The repetition rate of the laser, pulser, and rf is presently limited to 10 Hz. Figure 3 shows the layout of the full diagnostic beam line which includes seven fluorescent screens (YAG:Ce crystals), placed along the beam axis for measurements of the transverse beam profile.The two solenoid magnets downstream of the rf cavity focus the beam in the emittance monitor.They can provide a maximum field of B DSL ¼ 250 mT and are 0.2 m long.Eight pairs of vertical and horizontal corrector magnets (0.5 mT maximum field; 1 cm length) are distributed along the axis downstream of the rf cavity.The projected emittance is measured with a pepper-pot technique [24] at different locations along the emittance monitor [25] and cross correlated with a solenoid scan technique [15] for very low charge (1 pC or less) (Fig. 3).A spectrometer setup [26], made of a 230 mm long dipole magnet and two screens, follows the emittance monitor.The two screens are equidistant from the dipole center, one screen on the gun axis (reference screen to measure beam size) and one screen on the dispersive arm (to measure momentum and momentum spread).The dispersive arm makes a 30 angle with the gun axis.The nominal bending magnetic field required for a 4 MeV beam is B dipole ¼ 34 mT.A quadrupole doublet in front of the dipole provides horizontal magnetic focusing and limits the beam size vertically on the spectrometer arm screen.Various electron bunch charge diagnostics [wall current monitor (WCM), integrating current transformer (ICT), and Faraday cups] are distributed along the beam line (Fig. 3).

C. The laser system
Two independent laser systems were used for the results presented here: a frequency quadrupled Nd:YLF laser with regenerative amplifier providing a fixed wavelength of 262 nm and a wavelength tunable Ti:sapphire based system which provides laser pulses with wavelengths ranging from  262 to 282 nm.Details of those systems can be found elsewhere [15,27].In order to obtain a uniform transverse intensity profile, both lasers are first expanded and sent through a mask aperture which selects a small fraction of the Gaussian transverse profile.The diameter D laser of the laser spot on the cathode can be varied by changing the aperture mask diameter (0:5 mm < D laser < 1:8 mm).
Both lasers have a Gaussian longitudinal profile.The duration for the Nd:YLF system is t;laser ¼ 4:2 ps.The pulse length of the other laser system is variable The accelerating diode voltage pulse length of 250 ns is significantly longer than the laser pulse duration and appears as a DC acceleration.

D. The pepper-pot analysis tool
The emittance monitor assembly consists of independently movable pepper-pot and YAG:Ce screen from z ¼ 2:1 m to 2.6 m.The pepper-pot mask has a diameter of 5 mm, with a hole spacing of 250 and 50 m hole diameter.The tungsten mask thickness is 0.5 mm and the distance to the screen can be varied from 0 up to 0.5 m.More details on the emittance monitor system can be found in [25].The vertical (horizontal) alignment of the pepper-pot mask is done with a He:Ne alignment laser and by rotation of the camera.Pitch and yaw are not corrected so that the projected hole size might be smaller than the real hole size.The final emittance resolution of the device taking into account the camera pixel size has been measured with a laser light to be 0:03 m.
For the evaluation of pepper-pot images we follow Ref. [28].Using their notation we describe the pepper pot as a sequence of pinholes with spacing w significantly larger than the hole diameter d, separated by a distance L from the detection screen.In one projection (horizontal or vertical), the transmitted beam is split into N beamlets, numbered from m ¼ 0; . . .; N À 1.
Applying a statistical approach the rms beam emittance and the rms Twiss parameters are constructed from the beam's second-order moments: and similarly for the rms beam size, The second-order moments can be extracted from the main beamlet properties: the centroid positions x m , the centroid trajectory slopes x 0 m , the intensities I m , and the rms divergence spreads 0 m : hx 02 i ¼ with I tot ¼ P I m and all the summations running over the full set of beamlets, from m ¼ 0 to m ¼ N À 1.The beamlet intensity I m is obtained by summing up the detected intensities for pixel column (or line) j within a range associated with beamlet m: I m ¼ P j I j .Figure 4 represents actually the column intensities Ij versus the pixel column j.Both x m and x 0 m are centered, i.e., The original beamlet positions are simply given by the center to center hole spacing w (with arbitrary but irrelevant offset),

R. GANTER et al.
Phys.Rev. ST Accel.Beams 13, 093502 (2010) For the determination of the beamlet centroid positions and widths, we refrain from using a simple statistical method (mean and rms) since even small levels of background can have large effects on the result.Instead we apply a least-squares fit to the beamlet profile.The fit function is chosen to be the sum of N Gaussians and, if necessary, an appropriate background function.Supplied with reasonable starting values derived from the statistical means and rms values, the fit converges quickly and returns the beamlet centroids fit m and widths fit m in pixels.These values are then converted to the trajectory slopes of the beamlet centroids, and the divergence spreads 0 m;0 (the hole diameter d is subtracted quadratically), where p denotes the spatial extension of an image column (pixel size).Neglecting the hole size leads usually to an emittance increase of about 10%.To avoid possible optical or geometrical distortions from multirow (or multicolumn) projections, we use the projection of one to a few beamlet rows (or columns) near the center of the beam for emittance analysis.Given the rms definition of beam emittance, small beamlets at the beam periphery can have a large effect on the emittance while contributing only little to the total charge.To avoid the inevitable associated uncertainties in the determination of the emittance of the total beam, we restrict the analysis to the central beamlets contributing 90% of the total charge.A user-friendly application XANAROOT [29] based on the method described above was used for all emittance measurements presented here.Figure 4 illustrates a typical histogram obtained for a 200 pC beam.Rejected beamlets which are outside the 90% of the core beam are marked with a cross.With 90% charge the resulting normalized emittance is 1:3 m and becomes 1:7 m when taking all visible beamlets (100% charge) in Fig. 4. The green curve illustrates the profile fit used to estimate the beamlet width.The estimation of the beam boundary (to define 100% charge) is the main source of error.In order to give a realistic error bar on emittance, we used the difference between the 100% charge emittance and the 90% one.This difference is then defined as the total error bar size in all presented emittance results (Á" ¼ " 100% À " 90% ).

III. BEAM DYNAMICS SIMULATIONS OF THE DIODE-RF GUN
ASTRA [30] simulations were performed in order to describe the emittance evolution along the machine axis (Fig. 5).The diode part with the hollow cathode geometry was first simulated with POISSON [20] in order to get the appropriate field map (see Fig. 2, right inset).Figure 5 (top) represents the evolution of the beam size and emittance over the 4 m long test stand.The zero position (z ¼ 0) of the machine is taken at the anode surface.The bottom graph of Fig. 5 is a zoom of the diode to rf section (120 mm long).In the diode section the emittance increases very rapidly after a few mm from its initial intrinsic value (0:2 m at z ¼ À6 mm) to $1 m.On the other hand, the beam size slightly decreases and makes a waist at the anode entrance due to the electrostatic focusing provided by the hollow cathode geometry.With an accelerating field of 50 MV=m, 6 mm gap and 200 pC bunch charge, the nonlinear space charge effects, and also the nonlinear radial defocusing field of anode iris [11], are responsible for the rapid increase of emittance.The emittance reaches $1:2 m at the exit of the anode (z ¼ 20 mm).To preserve this emittance value, it is important to adjust the pulsed solenoid and realize the beam emittance compensation [21] at the rf cavity entrance plane.Pulsed solenoid focusing of the beam is illustrated on Fig. 5 (bottom) with beam size reduction for z > 60 mm.If the matching is properly done, the acceleration in the rf cavity will decrease the projected emittance (0:5 m < z < 0:8 m).The two double solenoids in the diagnostic line are then used to minimize the bunch projected emittance at the location of the measurements.The location of the pinhole array used for emittance measurements is at z ¼ 2:1 m which is for these simulation conditions not the position of a minimum emittance.Indeed, for the conditions of Fig. 5 the projected emittance is about 2 m at z ¼ 2:1 m.

IV. ELECTRON BEAM EXPERIMENTAL CHARACTERIZATION A. Emittance measurements
Although the increase of emittance in the diode part is important because of nonlinear space charge and anode iris effects, it is partially compensated by the nonlinear electrostatic focusing of the hollow cathode geometry [31,32].Figure 6 shows a comparison for two different cathode geometries: with (hollow cathode) and without electrostatic focusing (elliptical shape).The machine parameters have been optimized for the highest charge and were kept constant while decreasing the charge.Charge is varied by attenuation of laser intensity and/or reduction of laser spot diameter.The effect of electrostatic focusing on projected emittance is important and simulations reproduce well the experimental results.The electrostatic focusing produces a beam waist at the anode iris entrance with a local emittance minimum in the gap (see also Fig. 5).The emittance minimum obtained in the gap is almost recovered after crossing the anode.In the case of the elliptical electrode, the compensation of space charge and anode iris defocusing forces does not happen; the beam is then larger and diverges when entering the anode iris.In consequence, the outer radial part of the bunch passes at very small distance from the inner surface of the anode iris where fields are highly nonlinear [11].This leads to a stronger emittance growth when crossing the anode iris in comparison to the hollow electrode.
For charge below 20 pC, space charge effects in the diode are less important and quite small values of emittance can be reached.Figure 7 shows the typical beam shape produced by the diode/rf gun with the DLC hollow cathode and a copper insert at a charge of 13 pC.0.25 mm mrad).This emittance is also comparable to what is obtained in state of the art rf-photoguns [13,33].
Table II summarizes the best obtained normalized projected emittance in the diode-rf gun for the two modes of operation of SwissFEL: 10 pC and 200 pC.The best obtained emittance of 1:25 AE 0:2 m at 200 pC corresponds to a statistical average over more than ten measurements done for slightly different experimental conditions: diode peak gradient between 66 and 100 MV=m; different solenoid settings; Nb and Al inserts.Some of these measurements are discussed in Sec.IV C.

B. Momentum and momentum spread characterization
The spectrometer arm is used to measure the mean momentum and momentum spread of the electron beam as a function of the rf phase.In contrast to an rf photogun, the electron beam enters the rf cavity with nonzero energy (the pulser energy).Figure 8 represents a typical rf phase scan used to measure the beam momentum and to adjust the phase in order to minimize the bunch momentum spread.Indeed, the minimum emittance is obtained when the rf phase is adjusted for minimum electron energy spread.Two different laser pulse durations of t;laser ¼ 0:6 ps (rms) and t;laser ¼ 0:1 ps (rms) are presented in Fig. 8.The rf electromagnetic wave in the cavity is a standing wave and the phase ' rf is defined such that ' rf ¼ 0 is for the maximum measured mean momentum.The minimum uncorrelated energy spread is about 8 keV (rms) for t;laser ¼ 0:6 ps (rms) and about 4 keV for t;laser ¼ 0:1 ps (rms).This minimum is shifted to ' rf $ 10 degrees for t;laser ¼ 0:6 ps and to ' rf $ 15 degrees for t;laser ¼ 0:1 ps because of longitudinal space charge effects.The typical stability of the rf phase is around 1 (peak to peak) leading to energy fluctuations between bunches of 20 keV (peak to peak).Such large fluctuations are due to the modulator stability and to the fact that the klystron is running at only 20% of its maximum power (20 MW).

C. Comparisons between simulation and experiment
As shown previously in Fig. 6, there is good agreement between ASTRA predictions and measurements for the projected emittance.Simulations of the beam dynamics in the case of elliptical electrodes were also performed with the code OPAL [34] and compared to measurements.Figure 9 shows a comparison between the measured and calculated  beam size along the machine axis.The first measurement screen is at z $ 1:2 m, just after the in-coupling laser mirror.Several measurements with the same machine parameters, but acquired on different days, are shown and give an idea of the beam reproducibility (typically 25% peak to peak variation in beam size after one day).The agreement with simulation is rather good, given that simulations assume an ideal flat top profile for the transverse laser intensity distribution.In reality, variations as large as 50% peak to peak [27] are present in the laser transverse intensity distribution.
OPAL is a 3D PIC code [34] which provides the beam transverse profile via projection on a virtual screen (see Fig. 10).The formation of a toroidal beam shape as observed in the gun test stand is qualitatively reproduced in simulation.The hollow shape shown in Fig. 10 results from the counteracting magnetic focusing forces (linear to the first order with the radial position) and the radial repulsive space charge forces (highly nonlinear with the radial position) as sometimes observed in other electron guns [35,36].
The nominal beam (200 pC; 0:4 m) required for high charge operation at SwissFEL is not reached with the conditions of Fig. 6.Indeed, for the hollow cathode geometry and V pulser ¼ 300 kV, g ¼ 6 mm, the rms projected emittance is 1:9 mm mrad AE À0:2 for a charge of 200 pC and 5.5 MeV beam momentum (see Fig. 6).Best emittance values were obtained with the Ti:sapphire laser system and a shorter pulse duration of t;laser ¼ 3 ps (Gaussian longitudinal profile) as shown in Figs.projected emittance (at 90% charge) ranges between 1.0 and 1:6 m for a peak diode gradient going from 50 to 100 MV=m (see Fig. 11).This rather small dependence on gradient is also confirmed by simulation.Considering the error bars on Fig. 11, an average of the best projected emittance obtained with this diode-rf gun is 1:25 AE 0:2 m (Table II). Figure 12 represents the typical transverse beam profile for the nominal charge of 200 pC and the corresponding pepper-pot image.The beam asymmetry is due to the laser transverse profile.This emittance is above the maximum allowed value for the SwissFEL operation.However, the longitudinal laser profile in the case of Figs.11 and 12 was Gaussian and simulations (Fig. 13) show that, with a flattop laser time profile [11.3 ps full width (FW) and 0.7 ps rise and fall time] the projected emittance can be decreased by as much as 30%.Indeed, Fig. 13 (left) represents the simulated projected emittance (ASTRA) at the exit of the diode for a uniform longitudinal laser profile.For a gap of 5 mm and a voltage of 500 kV, the predicted emittance is 0:8 m as it was 1:4 AE 0:2 m in the measurements (1:35 m in the simulation) with the Gaussian laser profile of Fig. 11.
The simulation results of Fig. 13 also show that an increase of the diode gradient to 150 MV=m (600 kV; 4 mm gap) together with a reduction of C radius to 2 mm [C radius is the hollow cathode opening as defined in Fig.

D. Intrinsic emittance
The intrinsic emittance is the ultimate lower limit in beam emittance that can be reached for a given cathode material, surface electric field, and laser wavelength.The intrinsic emittance can be expressed as [37] " where x is the rms horizontal beam size in m, h the laser photon energy in J, È 0 the work function of the material in J, e the elementary charge, F eff the electric field on cathode surface in V=m, and " 0 the vacuum permittivity.Measurements of projected emittance at very low charge below 1 pC (where space charge effects can be neglected) and short pulse duration should approach this limit of intrinsic emittance.Figure 14 illustrates such measurements for two different photocathode insert materials: polycrystalline Cu and single crystal Nb.The agreement between the two measurement methods (pepper pot and solenoid scan) is rather good.During a solenoid scan, the electron beam size downstream of the solenoid is recorded as a function of the solenoid current.The transfer matrix of the solenoid has been calculated with the measured solenoid field map.The best fit to the measurement (beam size versus solenoid current) using the transfer matrix coefficient gives then the emittance value [15].The theoretical slopes are very close to the data points in Fig. 14 if one assumes a work function of 4.13 eV for niobium Nb and 4.3 eV for Cu.These values are compatible with the literature (È Nb ¼ 4:37 eV [38] and È Cu ¼ 4:31 eV [39]) if one takes into account surface contamination of our sample (up to 1 eV variation can come with oxidation contamination [38]) or roughness effects which can change the effective work function of our samples [40,41].Including the Schottky effect, due to the applied electric field (25 MV=m), the barrier is reduced to 3.93 eV for Nb and 4.1 for Cu.The initial kinetic energy of the electrons is then about 0.6 eV for Nb and 0.43 eV for Cu when using 274 nm photons (4.53 eV).Similar measurements were performed on other materials (Al, Mo, . ..) and the laser wavelength was then varied to minimize the initial electron kinetic energy.These results are discussed in [15].If the initial electron kinetic energy varies with the surface gradient, there is also an effect on the quantum efficiency.Indeed the quantum efficiency (QE) increases with gradient due to the Schottky effect [37,42].The QE dependence on surface gradient could give an estimation of the cathode work function by comparison with a photon limited emission theory such as described in [37].In the case of Fig. 15, the emitted charge is clearly space charge limited and no information on cathode work function can therefore be deduced.Space charge limited emission, assuming a transversely uniform laser, can be approximated with the following expression: with F eff being the electrical gradient on cathode insert.Assuming a laser spot size of D laser ¼ 1:08 mm, the simple space charge consideration [Eq.( 15)] fits well to the measured data of Fig. 15.This laser spot size is comparable to the experimental laser spot size of D laser ¼ 1:8 mm, taking into account the large nonuniformities in the laser transverse distribution of $50%.Some charges might also be lost on the anode iris.Nevertheless, Fig. 15 clearly illustrates the rapid increase of charge with surface gradient, 100 pC charge increase with only 15 MV=m surface field increase.This is a way to   economize laser energy which could be invested in an improved laser profile.Indeed a large fraction of laser energy is lost in the mask aperture used to get a uniform transverse laser intensity profile.High quantum efficiency is an indirect way of improving the laser transverse profile.

V. SUMMARY
We have characterized the electron beam coming from a diode/rf gun assembly by measuring the projected emittance, the beam size evolution, the momentum and momentum spread for various conditions of charge and diode gradient and/or diode voltage.Good agreement between simulation and measurement was obtained.This allowed us to understand the effect of the cathode geometry on transverse electrostatic focusing as well as space charge induced features such as the toroidal beam.Particular care was taken for the emittance diagnostics [quality of pepperpot images and analysis tool (XANAROOT)].Stable operation with diode peak electric fields of 100 MV=m for periods of several days could be demonstrated with the DLC coating on both electrodes.The chosen hollow cathode geometry gives encouraging projected normalized emittance values as low as 0:23 m with 13 pC.Projected emittance at charges below 1 pC was measured with two different techniques (pepper pot and solenoid scan) and agrees well with thermal emittance theory.Systematic study of the thermal emittance for different cathode material, surface finishing, and/or laser wavelength is now possible.For the nominal beam charge of 200 pC required for SwissFEL, the smallest projected normalized emittance obtained was 1:25 AE 0:2 m, which is larger than the requirement.However, simulation predicts that a flattop laser time profile together with small modifications of the cathode geometry (electrostatic focusing) should enable us to reach a projected emittance as low as 0:4 m.

1 and 2 )
to provide electrostatic focusing.The design of the hollow cathode allows the testing of 14 mm diameter disk inserts [field emitter arrays (FEA) or different photocathode materials].The disk edges are covered by the hollow cathode.In addition, the lips holding the cathode insert deform the electric field lines and provide electrostatic transverse focus to the electron beam (see Fig.2 right).The on-axis electric field on the insert surface is about half of the field present on the lips of the cathode surface.

FIG. 2 .
FIG. 2. Left: Cross section of the diode assembly with the hollow cathode geometry.Right: Zoom on the hollow cathode geometry showing the equipotential lines.

FIG. 3 .
FIG.3.Schematic of the diode-rf gun and diagnostic beam line.
The right part of the figure shows the corresponding pepper-pot image obtained in the emittance monitor.From the image, an rms normalized projected emittance of " n;x ¼ 0:23 AE 0:05 m was determined.Such beam quality fulfills the emittance requirements for the low charge operation mode of the SwissFEL project (10 pC;

TABLE I .
Range of maximum electric field sustained by different electrodes in the diode (without conditioning and without laser illumination) for gaps ranging between 1.5 and 6 mm.

TABLE II .
Summary table of the best measured emittance obtained with the diode-rf gun for the two modes of operation of SwissFEL(10 and 200 pC).Emittance at 200 pC corresponds to an average over more than ten measurements.