Low power commissioning of an innovative laser beam circulator for inverse Compton scattering γ -ray source

We report on the optical commissioning of the high power laser beam circulator for the high brightness Compton γ -ray source Extreme Light Infrastructure for Nuclear Physics. Tests aiming at demonstrating the optical performances of the laser beam circulator have been realized with a low-power pulsed laser-beam system and without electron beam. We show that, with the developed alignment and synchronization methods coming from the laser beam circulator design study presented in the Dupraz et al. paper [Phys. Rev. Accel. Beams 17 , 033501 (2014)], the laser beam circulator enhances the laser-beam power available at the interaction point by a factor in excess of 25. This corresponds to a potential of bringing the average laser-beam power in excess of 1 kW when the laser beam circulator is injected with the interaction point laser-beam pulse energy of 400 mJ at 100 Hz.


I. INTRODUCTION
Nowadays, several facilities around the world produce intense γ-ray with various excellent properties such as high flux, low bandwidth, energy tunablity, and high polarization, mainly for applications in nuclear physics and nondestructive analysis [1,2]. Such γ-rays sources are generated by the inverse Compton scattering, process whereby a laser photon beam scatters off a relativistic electron resulting an energy upshifted scattered photon beam. The phase space density of the electron and laser beams at the interaction are key parameters [3][4][5][6] that need particular care in order to achieve high flux along with very small bandwidth of the γ-ray beam, typically requested by nuclear physics experiments and the related societal applications [7][8][9]. An advanced γ-ray source has been built for the Nuclear Physics Pillar (ELI-NP) in the framework of the European Laser Infrastructure [10,11] by the EuroGammaS consortium [12].
The γ-rays will be generated by a high quality electron bunch train colliding with a high-power laser-pulse that recirculates 32 times at the interaction point. The gamma beam source (GBS) is expected to produce photons with tunable energy between 0.2 and 19.5 MeV with a narrow bandwidth below 0.5% and a high spectral density of 0.8-4 × 10 4 ph=s eV [13]. The GBS will consist of a lowenergy γ-ray beam line and a high-energy γ-ray beam line covering the 0.2-3.5 MeV and 3.5-19.5 MeV ranges photon energies, respectively.
Beyond the requirement at the state of the electron beam accelerator and the γ beamline, the interaction point is one of the bottleneck of the GBS. High flux and high brilliance inverse Compton scattering are based on electron ring coupled to high finesse resonant Fabry-Perot optical cavity. Stored laser power up to hundred's of kiloWatt have been demonstrated in enhanced resonant optical cavity [14], but it remains insufficient to reach the required γ-ray flux keeping, the low bandwidth and continuous energy tunability. Recent work have demonstrating an interesting option, combining the advantage of the laser power amplification of an enhanced resonant optical cavity seed in burst-mode and a room-temperature electron linac delivering a microsecond length bunch train [15,16]. But, the minimal time spacing of 15 ns between each γ-ray pulse required for the GBS ruled out this optical configuration for the interaction point. Multipass nonresonant cavity studies have been carried out for inverse Compton scattering source [17][18][19][20][21] and have shown to be adapted to a low number of bunches and tens of nanoseconds circulation period. The multipass nonresonant cavity or laser beam circulator designed for the GBS has got the unique particularity to achieve constant interaction parameter (angle, waist, polarization) at a fixed interaction point all along the passes.
In this article, the optical commissioning with a low power laser system and without the linac of the first interaction point module corresponding to the low-energy γ-ray beam line of the GBS is described. A comprehensive review of the first construction, implementation and commissioning of the new system constituted of the laser beam circulator (LBC) is given. The full GBS is quickly described offering a complete view of the machine in Sec. II. In Sec. III, the IP module elements, their individual testing and characterization are described. The details of the test conditions are given in Sec. IV followed by the presentation of the LBC pre-alignment procedure and results are given in Sec. V. The obtained performances are described in Sec. VI. Finally a summary ends the discussion in Sec. VII along with insights on the expected performances for the integration of the LBC in the GBS.

II. GAMMA BEAM SOURCE MACHINE DESCRIPTION
The GBS linear electron accelerator constitutes of a S-band photoinjector at ν rf ¼ 2856 MHz and C-band rf linac, at 5712 MHz, similar to that of SPARC_LAB at LNF-INFN, Frascati, Italy [22]. It has the capacity to deliver a high phase-space density electron-beam with energies in the 75-740 MeV range. The repetition rate of the pulsed rf is 100 Hz. In order to reach the required photon flux, the machine is foreseen to work in a multibunch mode. The 32 electron bunches spaced in time by T ¼ 46=ν rf ¼ 16.1 ns are produced in each rf pulse. The effective repetition rate of the collisions is thus 3.2 kHz. The main electron beam parameters of the ELI-NP-GBS, along with the description of the complete rf system have been reported in Refs. [11,23]. The preservation of the electron beam quality is ensured by an hybrid scheme consisting in a S-band high brightness photo-injector followed by a C-band rf linac. The reason lies in having a bunch long enough, σ z ≈ 1 mm, in the rf gun to limit the emittance degradation due to the space charge contribution, but taking advantage of the higher accelerating gradients provided by the C-band accelerating sections in the rest of the linac to keep it compact. To avoid the energy spread dilution due to rf curvature degradation effects, a bunch as short as σ z ≈ 280 μm, is injected in the C-band linac [24]. This is done by means of a velocity bunching scheme implemented in the S-band injector. The S-band injector is composed of a 1.6 cell S-band rf gun equipped with a copper photocathode and an emittance compensation solenoid, followed by two traveling wave SLAC type S-band sections. The rf gun accelerating field is E acc ∼ 120 MV=m, while the two S-band structures can operate at a maximum of 23.5 MV=m giving a maximum electron beam energy at the injector exit of 146 MeV [25]. The downstream C-band rf linac operates at 5.712 GHz, with the accelerating structures designed and developed at LNF [26], where the accelerating gradient can be set up to a maximum of 33 MV=m, providing enough margin for the off crest minimization of the energy spread in the 75-740 MeV energy range. The C-band linac is divided in two main sections (see Fig. 1): the low energy one, composed of 4 accelerating sections, carries the electron beam up to an energy of E e ¼ 320 MeV. A dogleg transport line downstream its exit, delivers the beam at the low energy interaction point (LE IP), avoiding in this way the bremsstrahlung radiation contribution. The downstream high energy linac is composed of 8 accelerating sections and brings the electron beam up to the maximum energy E e ¼ 740 MeV, then the electron beam, passing through a dogleg beam line, reaches the high energy interaction point (HE IP). In each of the two interaction point regions a quadrupole triplet provides a flexible final focusing for matching the electron beam and the counterpropagating laser pulse spot sizes.
The GBS relies on three laser-beam systems: the photocathode laser, and two interaction point laser systems. The photocathode laser-beam system is based upon a Ti:Sa chirped pulse amplification system [27]. It is composed of a standard 100 Hz front-end delivering a 25 mJ chirped pulse. The front-end is directly seeded by the optical master oscillator of the synchronization system of the GBS to reduce phase jitter. The laser pulse is then sent to a multipulse cavity generator providing 32 pulses synchronized to the rf. The pulse train is amplified in a multipass Ti:Sa cryogenically-cooled crystal pumped by 12 pump laser-beams that are time delayed. After the optical compressor and third harmonic generator, 32 UV pulses with an energy of a few hundred of microJoules each are transported 15 meters away to the copper photocathode of the GBS.
An industrial Yb:YAG high power picosecond laserbeam systems [28] that employ frequency doubling to produce laser pulses with energies in excess of 0.2J at 515 nm central wavelength and with 3.5 ps (RMS) pulse duration at a repetition rate of 100 Hz. The Yb:KGW femtosecond oscillator is synchronized on the GBS reference clock, preamplified and chirped before injecting in a water-cooled thin disk Yb:YAG regenerative amplifier. The output, a 30 mJ chirped pulse is amplified in a lowvibration cryogenically-cooled thick disk mutli-pass amplifier up to 350 mJ. After the optical compressor, a 515 nm, 5 ps FWHM duration, 200 mJ pulse is generated in a Triborate lithium (LiB 3 O 5 ) second harmonic generator.
In order to reach the foreseen world-class performance of the GBS, a new and original interaction point (IP) module has been designed. It is based on a passive optical system able to recirculate and focus 32 times a single laser-pulse at the IP. This LBC has to ensure not to spoil the excellent and unprecedented small bandwidth of the γ-ray beam. To achieve this tight requirement, a constant polar crossing angle ϕ between laser and electron beams must be maintained during the whole circulation process. In order to preserve a high number of scattered photons, the excellent overlap of the laser and electron beam distributions must also be preserved while the laser beam is circulated, thus tightly constraining the geometrical superposition of the successive laser-beam focus spots at the IP. The design of the LBC has been reported in Ref. [29]. An effective gain on the laser power, due to the circulation process, in excess of 20 is required to reach the γ-ray flux performances of the ELI-NP-GBS.
After each interaction point, at a distance of ∼10 m, the γ-ray beam collimation and characterization line delivering the γ-ray beam to user's experimental area is composed of collimation system [30], a 1m thick concrete shielding and a complete characterization system [31]. The collimation system consists of 14 tungsten slits of independently adjustable aperture placed inside a high vacuum chamber. The whole collimation system is placed on a micrometer 6 degree of freedom radiation hard motorization positioning system. The collimation system produces the required collimation in the 70-700 μrad to achieve the ≤0.5% bandwidth with the associated constraints on its alignment with respect to the γ-ray beam. The characterization beam line embeds a Compton spectrometer, a nuclear resonant scattering system, a γ-ray beam profile imager and γ sampling calorimeter.

III. INTERACTION POINT MODULE
The ELI-NP-GBS is constituted of two identical IP modules. Each is composed of the following elements.
Laser beam circulator (LBC): It is the core part of the IP module composed of two on-axis parabolic mirrors set in confocal arrangement, 31 mirror pair systems (MPS) each mounted on a piezo-electric rotation stage, a fixed pair of injection mirrors and a fixed pair of ejection mirrors.
A piezo-electric motorized inserter placed at the interaction point allows to set various tools on the LBC axis for diagnostics and alignment purposes only. The LBC as a whole is enclosed in an ultra high vacuum (UHV) vessel.
Synchronization and alignment laser tool (SALT): It is embedded in the IP module and can be remotely controlled to align and synchronize the LBC and monitor its performance.
Injection box (IB): It is a high vacuum vessel enclosed optical breadboard with various optics and laser beam diagnostics. It distributes and stabilizes the various laser beams delivered by the SALT and the main high power laser beam for their injection in the LBC. Near and far field laser beam profilers allow to set accurate injection references for the laser injection in the LBC. IP imaging optical system (IIOS): It is a high resolution optical imaging system coupled to two detectors, one giving a large field of view at the IP and one giving a time resolution down to 2.7 ns.
Cavity beam position monitors (CBPM)s: They are installed on the mechanical support fixed to the base Invar™ plate of the LBC. They provide precise transverse position measurements of the electron bunches at the entrance and exit of the IP module.
Laser-beam diagnostics (LD): They are embedded in the IP module to monitor the injected high power laser beam parameters relevant for the Compton scattering γ-ray generation and laser beam circulation. Laser-beam power, pulse length, wavefront and polarisation are monitored by the LD.
A schematic top view of the interaction point module is given in Fig. 2. Another 3D view with more details on the LBC elements and electron and laser beam directions is given in Fig. 3. It was taken downstream of the interaction point. The various instruments can be distinguished on the rear side, the large LBC vacuum chamber is on the front. The downstream CBPM support is also visible. Details of the various components of the interaction point module are presented in the remaining of this section.

A. Interaction point parameters
A complete simulation code has been developed [29] to determine the specifications and tolerances allowed on the manufacturing of the optics, the motorization and the opto-mechanical elements. It is based on two different approaches: (i) a fast computing time ray tracing and 4 × 4 ABCD matrices for geometrical purposes, and in particular alignment [32]; (ii) a slow computing time beam synthesis propagation provided by Code V [33] for optical performances studies. These two different calculations are managed by a master code that hold the geometry parameters to ensure a consistency between the two codes. The figure of merit for the optimization of the design and the tuning of the optical element is the time average spectral density (TASD) of the γ-ray beam. This quantity represents the number of γ photon per second per unit of bandwidth. The essential design parameters and associated tolerances to achieve the required performances of the GBS are summarized in Table I.

B. Interaction point module mechanical frame
The whole IP module is set on a granite structure providing a thermal inertia. The granite structure is mounted on six, 3-axis adjustable, feet for initial alignment in the mechanical reference network. A mechanical frame independent of the granite block holds the LBC and the IB vacuum vessels to decouple the vacuum chambers and the inner optical breadboards. The LBC optical table is supported by 5 feet, screwed on an Invar™ [34] plate that is set on the granite block. This intermediate optical table can be moved from air to adjust the whole inner-vacuum optical breadboard in order to extend the final alignment range of the laser focal spot on the electron beam. The SALT and LD can be removed with the help of a mobile crane and placed back into position thanks to a set of precise localization pins.

C. Laser beam circulator
The LBC is made of three groups of optics: the two parabolic mirrors M1 and M2, an injection mirror-pair M0-M0 0 and an ejection mirror-pair M33-M33 0 and the 31 MPS (Fig. 2). The parabolic mirror M2 is mounted on a five degree of freedom motorized mount suitable for UHV operations. Each MPS is mounted on a piezo-electric rotation stage.

Parabolic mirrors
Parabolic mirrors specifications have been drawn based on simulations using a model of the system [35]. The mirrors have been polished and coated by REOS-SAFRAN [36] with the characteristics gathered in Table II cosmetics are MIL-PRF-13830B [37] (S/D 20-10) corresponding to high-power laser-beam grade optics. The useful part of the parabolic mirrors is labeled the useful corona delimited by the inner circle of a radius r min ¼ R C − D M and the outer circle of radius r max ¼ R C þ D M on which refers the surface figure error specification. Careful attention must be paid to the midspatial frequencies of the power spectrum density (PSD) of surface defects, particularly to the PSD gradient in the useful corona, which is leading to a spread of the focii points of the various laserbeam passes as they are reflected at different azimuthal position by the useful corona of the parabolic mirrors. This is one of the most stringent specification of the parabolic mirrors. The targeted surface characteristics have been achieved by the polisher and the measured PSD is shown on Fig. 4 for the two parabolas along with the design specification.
The parabolic mirror M1 is mounted on a manually adjustable five-axis mount thanks to a series of high turn per inch screws to get enough accuracy for the tilt setting. The typical resolution obtained with a 1°screw rotation is of 20 μrad. This requirement is driven by the fact that the alignment of M1 with respect to the global GBS mechanical reference frame defines the final orientation of the LBC axis. The corresponding error budget for the M1 alignment in the GBS mechanical reference frame is of AE50 μm, AE100 μrad.
The five-axis motorized mount of M2 has been realized with the standard of accuracy and precision of synchrotron beam line optomechanics [39]. A 3D view of it is given in Fig. 5. It is based on two large TX and TZ axes tables. The vertical TY-axis is set on a shelf bracket on top of the TX moving part. The TY moving part holds a two rotations flex frame for the RX and RY motions. The characteristics of the five-axis motorized M2 mirror mount are collected in the Table III. Each axis is equipped with two radiation hard, UHV compatible, high accuracy compact electrical limit-switch [39], a high resolution optical encoder and a reference mark. The axes are driven by a multiaxis programmable controller [40].

Mirror pair system
The MPS is one of the key-elements of the LBC since it allows the circulation of the laser-beam pulses 32 times at the IP. Each MPS is meant to be an optical invariant so that the laser-beam exits the system parallel to the input. Tight mechanical requirements are imposed on this system in order to avoid vignetting of the laser-beam and marring the alignement of the LBC [29]. The MPS mirrors parallelism must stand mechanical shocks that can be generated during the installation in the LBC, pressure variation from atmospheric to ultra high vacuum pressure, and have elastic behavior to a thermal stress of dT ¼ 1 K around its working temperature.
Two technological solutions were studied. The first one is based on a flexor to set the angles between the two mirrors maintained within flanges by a lateral-side fixationretaining leaf-spring. In the second one the mirrors are glued on their fixation holder with an angular tuning control on one of the mirrors while the reticulation process takes place. The first solution has the main advantage of an angular tuning but, in the meantime, presents a worse mechanical stability due to the presence of the leaf spring and the flexor differential screws. The second solution has very good mechanical stability but implies a complicated process to control the parallelism during the gluing process. A hybrid solution has been integrated, in which the mirrors are glued in the mirror flange. The realization of this last solution is shown in Fig. 6. In detail, each MPS is composed of two flat high grade mirrors, two mechanical mirror mounts, a two-blades flexor, a piezo-electric motorized rotation stage and a positioning base plate. The mechanical mirror mount is made of stainless steel. The mirror mount is sharpened at the equator to avoid beam vignetting. The mirror sits on three pins on the bottom of the mirror mount. The mirror position can be secured by a TABLE I. Summary of the optimized parameters to reach ELI-NP-GBS specifications, using design electron beam parameters [10]. The reader is referred to Ref. [29] for a detailed description of the meaning of the parameters. 0.25 mm leaf spring bent on the top of the mirrors by two screws. The two-blades stainless-steel flexor is bent thanks to a differential screw made of brass. The inner screw is screwed in a swiveling nut.
The motorization of the MPS allows a rotation of the mirrors pairs around θ 0 within a range of AE2.5°to allow for the synchronization of each pass of the LBC with the electron bunch train. It is a modified version of a piezoelectric rotator for ultra high vacuum [41]. It has a torque of 2N/cm with a bidirectional repeatability of 17.6 μrad and a sensor resolution of 0.2 μrad. A step of δθ on the rotation gives a time delay on the circulation period of Difference between S and P polarization reflectivities For pulses 515 nm, 3.5 ps, 100 Hz, c Average of the measurement made on 3 points along the radius in the useful corona, polishing technology and fine and accurate hand polishing to randomize and reduce the typical micro-structures of MRF polishing, thus limiting any systematic surface error. Their characteristics are shown in Table IV. They are made of fused silica with high-grade cosmetic aspect (S/D 10-5). It has been shown that systematic macroscopic surface error on the MPS mirror significantly limits the LBC performances [35]. In the pursue of reducing this defect, Institut Fresnel [44] developed a process [45] to compensate for the coating stress and also to randomize the focus surface errors corresponding to the Zernike coefficient Z 0 2 in the Zernike decomposition of the mirror surface [46]. Each substrate has been controlled with NewView Zygo™ 7300 [47]. The back-coating compensation is then calculated after calibration studies have been made [45]. The mirrors have been coated with a Leyblod™ Optics Helios [48,49] at Institut Fresnel. The multilayer coatings are constituted of Hf 2 O 5 high index and SiO 2 low index layers. The multilayer stacking has been optimized to get a highreflectivity and small difference between the s-polarized R s and p-polarized R p laser-beam reflectivities. The mirror's active surfaces have been controlled again after the coating took place. The measured distribution of peak-to-valley (PV) surface errors and Z 0 2 focus Zernike amplitude are shown in Fig. 7. It has been decided to polish and coat 80 mirrors in order to have the possibility to select the best combination of 62 mirrors. In order to reduce the cumulated focus while also preserving a low PV wavefront error within the Maréchal criterion for all passes [50], an algorithmic ordering procedure of the MPS mirrors has been applied [51]. Using the developed simulation code several configurations have been tested. The real surface maps of the coated parabolic and MPS mirrors have been implemented for these simulations. The injection mirrors are considered perfect, meaning that the residual focusing error in the wavefront is compensated in the laser system and laser transport beam line up to the first pass and that the parabolic mirrors are confocal and perfectly aligned. For each simulation, the orientation of the surface maps to be attached to each mirror is randomly selected between 0 and 2π and the maximum parallelism defect of the MPS mirrors is fixed to Δϵ ¼ 7.5 μrad. From one simulation to another for a given configuration, both the mirrors and their order are identical. Various selection rules configuration have been tested. The results are displayed in Fig. 8. The configuration 1 corresponds to a random selection of the mirrors without ordering. The configuration 2 consists in selecting the 62 mirrors by placing the mirror with the best focus and PV in first position and sequentially minimizing the absolute value of the accumulated focus for the following mirrors. For the configuration 4, the 80 mirrors   [46]. c LIDT: laser induced damage threshold measured by LIDARIS for 3.5 ps length 515 nm central wavelength laser under vacuum at 23°AOI at f rep ¼ 50 kHz for 10 7 -on-1 are classified by the ordering algorithm in which the absolute value of the accumulated focus is constrained not to exceed f thresh ¼ 2 nm at each pass. When the ordering algorithm does not find a mirror respecting this criteria, the mirror with the best PV mirror is selected. The configuration 4, f thresh ¼ 1.5 nm and when the algorithm does not find a mirror candidate to keep the cumulated value of the focus below f thresh , the threshold value is incremented to f thresh þ 0.5 nm. When a mirror candidate is found, the f thresh value is reset to its initial value of 1.5 nm for the next mirror. It is observed in Fig. 8 that the relative TASD improves by 0.06 in configuration 2. The configuration 4 is the best, the relative TASD improving by ∼0.09 and is the chosen classification scheme.
The inter-mirror distance D MPS is first set with a PEEK [52] spacer and a rough parallelism steering is then realized. The mirrors are then glued with an epoxy two-parts Varian TorrSeal™ [53] low outgassing UHV-compatible glue with high radiation hardness. Several tests have been done to get the most stable and simple gluing process. Four points of glue are applied, as shown on Fig. 9, with a calibrated needle to minimize the amount of glue and to get a reproducible application of its quantity. They are placed at the equatorial position of the mirrors in the front and back part of the mirror mechanical mount. Immediately after application the MPS are put in an evacuated chamber with a pressure below 1.5 × 10 −1 mbar for at least 24 hours while the reticulation process takes place. The angular beam deviation induced by the MPS parallelism instability is measured with an electronic autocollimator, UltraSpec™ TA US 300-57, with a focal length f ¼ 300 mm [54]. It has been modified by a stainless steel support to improve its vertical stability.
The stability of the MPS system has been investigated in air and vacuum over a long time period. A double selfreferencing measurement is performed by use of a reference mirror and the reflection on the back-surface of the first mirror of the MPS. The measurement is made with the MPS placed such that the laser-beam of the autocollimator impinges the MPS close to normal incidence. The accuracy and precision of the autocollimator are AE1.  AE0.02 μrad, respectively. A long term measurement is shown on Fig. 10. It is observed that the horizontal angle beam deviation Δϵ x varies slowly with time while the vertical angle beam deviation Δϵ y is fluctuating faster and follows a daytime period roughly anticorrelated with the air temperature variation. We can notice a little variation in the x and y in the parallelism at each transition air-vacuum and vacuum-air. The amplitude of the variation is small Δϵ vac-air x ∼ −1.5 μrad and Δϵ vac-air The mechanical stability of one of the MPS has been checked by repeating the procedure of installing it inside the LBC several times on localization pins. This operation is delicate, particularly since the insertion of the base support of the MPS can generate a mechanical shock of a few g. The variation of the angular beam deviation is measured by the autocollimator to be within the specifi- μrad for a single MPS. The parallelism is affected by the thermal dilatation/ contraction of the different materials of the MPS optomechanics. The thermal sensitivity of the MPS parallelism has been studied at various working temperatures between 20°C and 25°C. The foreseen temperature in the accelerator tunnel of ELI-NP-GBS is 22 AE 0.5°C. Figure 11 shows the angular beam deviation on both axis as function of the temperature. The thermal sensitivity is found different for the two axes due to the shape of the mechanical design. The thermal sensitivity of the vertical and horizontal parallelism is a y ¼ −17 AE 5 μrad=K and a x ¼ −1.9 AE 1 μrad=K, respectively. The deviation to perfect parallelism is significantly larger than specified. It appears that the vertical axis is most sensitive to temperature variation. The horizontal thermal angular drift is very low and compliant with the requirement. As a result, a thermal stabilization of the MPS installed in the LBC is required. A UHV compatible ceramic platinum thermoresistance (PT100, class 1=3 C) cabled with four Kapton insulated wires has been implemented to monitor the temperature of the mount. The probe is fixed with silver UHV conducting glue on a small piece of stainless steel and then fixed on the body of the MPS flexor thanks to one of the screws used to fix it on the rotation stage. Four MPS have been equipped with such thermometers. A picoLog™ PT104 four-channel temperature data logger is used. A high-performance 24-bit ADC is employed to achieve a 0.001°K resolution and an accuracy of 0.015°K. These thermometers provide precise and under vacuum monitoring of the MPS temperature that is critical to monitor the LBC performance.
An important discrepancy has showed up during the tests between the differential screws inducing a coupling between the x and y axes.This defect is one of the main mechanical limitations of the present design that complicates greatly the parallelism tuning. Consequently, those deficient MPS differential screws had to be reset in another screwing range or replaced. For some MPS a mechanical relaxation of the ensemble of the flexure blade and the differential screw has been observed as long as a few hours after the setting. For some MPS a mechanical relaxation of the ensemble of the flexure blade and the differential screw has been observed as long as a few hours after the setting.

Injection and ejection mirrors
The optics used for the injection and ejection mirror systems, see Figs. 2 and 12, have the same characteristics as the MPS optics, that are detailed in Table IV, except that the M0 0 diameter is 50.8 mm. The injection mirrors, dubbed M0 and M0, act like folding mirrors forwarding the laserbeam coming from the IB to the fixed parabolic mirror M1. Due to space constraints a two-mirrors setup has been chosen.
A nonvignetting mirror M0 is set within the LBC circulation corona allowing the first mirror to work at 45°of incidence. Thanks to this arrangement, the laserbeam is injected in the LBC perpendicularly to the electron beam axis, thus simplifying the vacuum vessel design. A flexor-based design has been made to allow for a tuning of the position of M0 0 according to the fixed M0 position and the injected laser-beam, as can be seen on Fig. 12. The flex brackets are designed with fine, high turn per inch, threaded adjusting screws. The ejection mirrors system is similar to the injection one and allows to steer the laser beam out of the LBC vacuum chamber on a high-power air cooled beam dump and a series of laser diagnostics.

D. Embedded instruments
In addition to the LBC, the IP module is equipped with three other main instruments. They are remotely controlled and monitored through an autonomous supervision software and a data acquisition system [55], for expert diagnostics, initial alignment and synchronization of the system. It is interfaced with the main EPICS control system of ELI-NP-GBS for its regular operations.

Synchronization and alignment laser-beam tool
The SALT, schematized on Fig. 13, uses a low frequency-noise laser-oscillator with a repetition frequency ν ð46Þ ¼ ν rf =46 ¼ 62.08695 MHz with a 1030 nm central wavelength and bandwidth of ≈13 nm (FWHM), delivering an output power in excess of 100 mW. It is frequency locked on the optical master clock of ELI-NP-GBS that is distributed by a nonstabilized fiber link [56]. The laser pulse is sent in a noncritical phase-matched secondary harmonic generator (SHG) based on a 7 mm long LBO [57] crystal hold in a temperature-controlled oven that is set at 194°C. The conversion efficiency is 18% for standard operations. The measured laser pulse duration after the SHG is t p < 200 fs. The laser pulse is further split in two beam arms to generate a reference beam (REF), used as reference to synchronize the LBC on ν ð46Þ , and a circulation beam (CIRC), used for alignment and synchronization purposes, as if it were the main laser pulse that circulates in the LBC under normal operations. The repetition frequency of the CIRC beam is reduced to 700 kHz in order to inject only one laser pulse per circulation time period T circ;32 ¼ 515.4 ns where T circ;n ¼ n=ν 46 . It is done thanks to a two BBO-crystal Pockels-cell pulse-picker (PC). The measured pulse picking extinction ratio is greater than 1=500. The REF beam is sent to a piezo-electric motorized delay-line that allows to synchronize the REF beam with the first pass of the CIRC beam in the LBC by remotely tuning the delay over a range of 166 ps. As shown, in Fig. 13, both REF and CIRC beams are then expanded through telescopes to match the required laser beam size w 0 ≈ 8 mm at the LBC input.
An active-stabilization loop corrects the laser beams low-frequency pointing drifts du x ; du y down to 4 μrad and position drifts dx; dy below 90 μm by means of a software PID and motorized mounts. At the output of the SALT, the CIRC and REF beam powers are ∼170 μW at 700 kHz and 1.8 mW at 62 MHz, respectively. These values correspond to a splitting balance of 90%/10% between the REF and CIRC beams. These parameters are summed up in Table V.

Injection box
The injection in the LBC requires an accurate control of the beam pointing [29,35,51] of both the main laser beam and CIRC/REF beams. A high vacuum chamber hosts a decoupled optical breadboard equipped with several leakage mirror pickups for the laser-beam diagnostics, a beam dump and beam dispatching in the LBC. A schematic view of the optical setup is given in Fig. 14. After the injection in the IB vacuum chamber the pointing of the CIRC beam is stabilized by a fast piezo-electric driven 2 00 steering mirror mount (TTPM) made of a 4 piezo-electric stacks giving a high stiffness and high resonant frequencies >700 Hz able to induce a maximum tilt of 1.5 mrad [58]. A beam leakage of less than 7% is focalized onto a 4-quadrant photodiode detector (4QD2) with a size of ð10 × 10Þ mm 2 thanks to two lenses of focal lengths f 1 ¼ 400 mm and f 2 ¼ −30 mm set at distances to result in an equivalent focal length in excess of 8.9 m. The 4-quadrant gap is 90 μm [58]. A closed-loop controller continuously corrects for deviations of the laser beam from the desired position. Figure 15 shows the residual pointing instabilities when the closed-loop controller is activated, and a < 0.3 μrad (RMS) beam pointing stability is achieved (see Fig. 15 for t > 1170 s). By Fourier analyzing the residual noise, it is found that perturbations up to 380 Hz are well corrected. Without activation of this fast-stabilization loop the residual pointing instabilities measured at 10 Hz are du x ¼ 3.6 μrad (RMS) and du y ¼ 3.7 μrad (RMS) (see Fig. 15 for t < 1170 s).
The CIRC beam is further reflected by high reflectivity mirrors held in stick-slip piezo-electric motorized mounts MMM1 and MMM2. They control the laser beam injection in the LBC. Its position is controlled on the near field, visualized with CCD3, by imaging the laser in an equivalent plane of MMM2 with a lens of focal length f 3 ¼ þ300 mm. The laser beam position is set with an accuracy of ≈19 μm on the near field (L3,CDD3). Its angle is controlled on the far field, visualized with CCD4, with lenses of focal lengths f 4 ¼ þ400 mm and f 5 ¼ −50 mm giving an equivalent focal length in excess of 7 m. The laser beam angular injection is monitored on the far field (L4,L5, CDD4) with an accuracy of 0.8 AE 0.1 μrad. Both steering mirror mounts are equipped with optical encoders and operated with a closed-loop controller. The resolution of the encoder is 0.17 μrad. Slight linear drifts of du x =dt ¼ −44 nrad=min and du y =dt ¼ −31 nrad= min are noticeable when the laser beam regulation is not activated on the far-field. They have been identified to be related to a thermal drift on the optical encoder.
With all the pointing stabilization loops activated in the SALT and IB systems the residual pointing instabilities    in air are dx ¼ 14.1 μm and dy ¼ 11.9 μm RMS at the interaction point by the IP imaging optical system. These values fall down to dx ¼ 2.9 μm and dy ¼ 3.1 μm when the system is pumped down and reaches the thermal equilibrium.

Interaction point imaging optical system
The IP imaging optical system (IIOS) allows to realize the spatial alignment of the LBC passes with a micrometer spatial resolution and temporal resolution of less than 5 ns. Thus for each pass the laser-beam profile can be extracted. It is additionally used to image the interference fringes between REF and CIRC beams at the IP for the synchronization of the LBC. As shown in Fig. 16, a motorized mechanical support holds a thin film pellicle (TFP), reflecting not more than 1% of the incident light depending on the CIRC and REF beam polarization. The reflected beams are used for alignement and synchronization purposes with the CIRC and REF beams. A YAG:Ce fluorescent screen (FBS) and an optical transition target (OTR) made of a fused silica window in order to stand for the large electron beam intensity [59,60] can be inserted at the IP and are additionally employed to image the electron beam. A first objective (LG1), placed under vacuum, is fixed on the targets support at the center of the LBC. A second objective (LG2), located in air, allows to image the IP region through an additional high magnification microscope objective onto two detectors, a CMOS GIGE camera and an intensified gated camera (ICCD). All the optics sit on opto-mechanical fine-height adjustment supports.
The IIOS design is driven by spatial and environmental constraints and the requirement of a high resolution. It relies on an image relay with large numerical aperture objectives (LG1,LG2) to image the IP region outside of the vacuum vessel with a good module transfer function (MTF). In air the image is obtained on the area detectors equipped with standard high resolution microscopes. Highresolution long focal-length objectives are essential components of quantum gas experiments. There is an increasing need for high resolution imaging, below 3 μm, in order to extract fine details such as topological defects in Bose-Einstein condensates or single trapped ions [61,62]. The design of the long focal length objectives (LG1, LG2) of the IIOS is mostly inspired by recent works of L. M. Bennie et al. [63] and J. D. Pritchard [64]. The design is using that of the Pritchard ATOM objectives modified to provide at a central wavelength of 515 nm an infinity-corrected and diffraction limited performance whilst maximizing the available numerical aperture given the minimum working distance of 60 mm imposed by the space inside the LBC corona and the TFP inserter. For reasons of simplicity, all elements are constrained to be of 2 inches diameter BK7 singlets. The objectives have been iteratively designed using the ray-tracing software Zemax™ [65] to numerically optimize the lens spacing while minimizing the RMS wavefront error at the wavelength λ 0 .
It results that each objective lens has a point spread function of σ ≈ 1.1 μm. The drawback of using this image relay is the inherent sensitivity to alignment especially since the first objective LG1 stands inside the vacuum chamber and the second objective LG2 after the vacuum optical window. In the present design a magnification of M 1;2 ¼ 1 are chosen due to space constraints. The design of the relay imaging outside the vacuum chamber allows for more flexibility and the possibility to use a standard microscope configuration. The imaging line magnification is designed to achieve a resolution close to 1 μm in the object plane and also to achieve a good resolving power to be able to measure the contrast of the interference fringes for the synchronization of the LBC on ν ð46Þ , the magnification of the microscope objectives has to be large enough to avoid their blurring, due to the limited MTF of the ICCD. Effectively, the combination of microchannel plates (MCP/I), spaced by less than 6 μm, and the phosphor grain considerably degrades the MTF of the intensified time-gated detectors. One of the best combination is used and a MTF ICCD ≈ 45 line pair=mm [66] has been specified by the manufacturer and confirmed during dedicated test.
The period of the interference fringes is i ¼ λ 0 =ð2sinϕÞ ¼ 1.8 μm. The minimum microscope objective power (uOBJL3) is then driven in first approximation by the sensor resolution and a sufficient number of points per fringes giving P ICCD obj ∼ 30 for the ICCD imaging line. The same argumentation considering a pseudo MTF CMOS ≈ 91 lp=mm for the CMOS detectors with pixel of 5.5 × 5.5 μm 2 leads

IV. OPTICAL COMMISSIONING CONDITIONS
The tests have been realized in the assembly hall, of ISO8 cleanliness, of the ALSYOM/SEIV-ALCEN company in collaboration with LAL/CNRS. The IP module has been assembled under an ISO5 cleanliness mobile air ceiling. The cleanliness requirement is driven by the future high power laser beam exposure of optical elements of the LBC and the IB. The volume of the ISO5 air ceiling is ð4.5 × 4 × 5Þ m 3 . Its flow is 32000 m 3 =h at a speed of 0.45 m=s. The ambient noise generated by the air ceiling is in excess of 69 dB. The hall also hosts at least five other ISO5 air ceilings of various sizes, two dust cleaning showers and a stand for UV dust inspection. All these equipments generate a significant acoustic noise. This environment is thought to be equivalent or more noisy than that of the accelerator bay where the IP module will finally be installed and where chillers used for the cooling of accelerating structures will be running continuously. The ISO8 hall is air-conditioned with temperature and relative humidity regulations. The setting temperature is T 0 ≈ 20°C and can be varied in the range from 19°C to 22°. The temperature has been monitored in various points of the IP module. The temperature under the air ceiling T LBC was oscillating in the range [20.5°C, 22.0°C] over the test period. Under the air ceiling an optical table has been installed for the MPS setting and measurement. To reduce the effect of air turbulence during the laser-beam angular deviations measurement with the autocollimator, a removable enclosure has been installed over the optical table. The temperature on the optical table in the enclosure was The aim of the commissioning is the validation of the installation process, the validation of the prealignment strategy, the implementation of alignment and synchronization algorithms and the demonstration of the performances with a low power laser beam simulating the electron and the high power laser beams.

V. PREALIGNMENT OF THE LASER BEAM CIRCULATOR
The LBC pre-alignment entails the definition of its optical axis, the IP reference point in the mechanical reference network defined by the alignment of the M1 parabolic mirror, the alignment of M2 in confocal arrangement with respect to M1, the definition of the reference position of the IP on the IIOS and the MPS alignment and installation in the LBC.

A. Optical axis and parabolas alignment
The tolerance on the angle of the LBC with respect to the electron axis is 1.7 mrad, on the basis fixed by the required performances on the γ-ray flux and bandwidth (TASD), and limited by the positioning of the optical elements and knowledge of the positioning of the optics on its mounts. When the IP module will be integrated in the GBS lowenergy linac, the residual angle error between the electron beam and the LBC optical axis will be due to the cavity beam position monitors positioning error in the mechanical reference network and the error of the LBC optical axis orientation.
The optical axis of the LBC is defined by the collimated REF beam that is sent through a 6 mm-diameter pinhole (P1) fixed on the mount base of M1 (see Fig. 2  SALT instruments in the IP module. It makes a reference out of any incoming beam, and diagnoses any misalignment between the incoming beam and the return beam. The principle of SRA is illustrated in Fig. 17.
Once the flat mirror is aligned in autocollimation on the REF axis, a 50 mm-diameter, 100 mm focal length spherical mirror mounted on a piezo-electric motor is installed to create a source point at the IP. For this purpose the TFP is turned by 90°to collect the REF beam reflected by the spherical mirror in the IIOS. The size of the focal spot generated by the spherical mirror is optimized and it is positioned on the IP reference coordinates previously taken using the IIOS. The TFP is then turned back to its original position. A collimated beam is created by M1 and reflected back, by the flat square mirror and focuses again at the IP. Wherefore, the M1 is aligned on the defined optical LBC axis and the focusing is optimized by looking to the focal spot positions on the IIOS. At this point the M1 is fixed and becomes the reference optical element of the LBC setup.
The CIRC beam is then injected and steered to superimpose the first pass, dubbed p ¼ 0 on the IP reference coordinates of the IIOS. The CIRC beam collimation is slightly adjusted to get the best focusing by using the beam expander (BEX) in the SALT. M2 is then installed in its 5-axis motorized mount. The transverse positioning is done in the mechanical network reference thanks to the laser tracker and using the motorized translation stages TX, TY and TZ (see Fig. 5). The tilts of M2 are set by superposing the second pass, p ¼ 1, on the reference marks of the first pass in the IIOS. At this point, M2 is aligned in autocollimation but not yet in confocal geometry with M1 as the error along the z-axis is still larger than 100 μm. The error budget for the prealignment is summarized in the Table VII. To set M2 in a confocal position with respect to M1 focus point, the first MPS, MPS1, is installed in the LBC. The frequency of the laser oscillator is adjusted to correspond to the frequency ν  AE3400 AE1700 Indexing f a Error and accuracy determined by the image analysis to determine the center of the diffraction pattern and the centroid of the spot. b Angular error limited by the beam pointing stability due to air turbulence and accuracy limited by the short focal length used. c The precision on the positioning of M1 and M2 is limited by mount. The budget is fixed by the paraxial approximation and aberration produced by the tilt to compensate for the decentering of M1 and M2. d The final budget is limited by the beam pointing stability in air. e The precision on the indexation MPS angle θ 0 , the MPS mirrors distance error δD MPS , the precision of the synchronization cδt sync and the tuning of the laser frequency contribute quadratically to the total uncertainty. f A systematic shift of ∼130 μm has been measured due to the calibrated block definition error used to set this length. It was corrected after the MPS installation by applying a systematic rotation of δθ ¼ 3.97 mrad of each MPS. motorized mount. The final position in z of M2 has been monitored by the laser tracker and a shift of z ¼ þ2.1 mm is measured at the end of the procedure. This discrepancy may be explained by the poor accuracy of the knowledge of the vertex coordinates of the parabolic mirrors in their mounts, see Table II.

B. Mirror pairs system parallelism setting
After the gluing process, the MPS are mounted on their piezo-electric rotation stage and base. This ensemble is installed in the test bench where the autocollimator is set (Fig. 19). The rotation stage indexing at θ 0 is done by the locating base and by observing the angular horizontal deviation of the back side of the MPS mirror thanks to the autocollimator. The yaw of the base is then set by looking at the vertical angular deviation in the autocollimator. Each MPS parallelism is tuned up within the tolerance and stored under a temperature controlled enclosure.
The MPS are then installed in cascade in the bench. For space constraints in the bench a maximum of 11 MPS can be installed. Each time a MPS is added in the cascade bench, the angles of the newly inserted MPS are adjusted to keep the total deviation ϵ 0 ¼ P i ϵ i within the tolerance, see Table I.
The results are shown in Fig. 18. The total cumulated angular deviation is ϵ 0 ¼ 4.5 AE 1.0 μrad with cascade correction andε 0 ¼ 26.2 AE 1.0 μrad without. As the autocollimator light goes forth and back in the MPS, the parallelism error is the half of the measured angular beam deviation. The MPS parallelism setting has been done at a temperature of T MPS ¼ 21.2 AE 0.1°C.

A. Laser beam circulation and mirror pair system installation
The laser beam impact does a 4π rotation on the parabolic mirrors as shown on the Fig. 20. The beam impinges M1 thus making the first impact p ¼ 0 on the parabola, the reflected beam is focused at the IP and is collimated back by M2 towards M1 through the MPS1 and generates the impact p ¼ 1 on the M1 parabola. The processus is then repeated 31 times.
The MPSi have been installed in the optical order with i ¼ ⟦1; 31⟧, where i equals the pass number p. The mechanical index of the MPS is different on the helix made by the raiser support as one can see on the Figs. 20 and 21 and is given by its position order starting from M 1 to M 2 . One can notice that the MPS16 is in the middle of the LBC. The advantage of the installation in the mechanical order is that each time a MPS is installed there is free space on one of its side. The main disadvantage is that 17 MPS must be installed in order to achieve a first circulation of the laser beam at the IP. Given the mechanical sensitivity of the MPS to shocks, the optical order installation has been chosen so that each time a MPS is installed a new pass is FIG. 18. Measured cumulated angular beam deviation with (ϵ ; x ; ϵ ; y ) and without (ε ; x ;ε ; y ) the cascade cumulated deviation error correction. generated at the IP. An online monitoring of the spot position in the IIOS is performed during installation and thus any shock on a newly installed MPS can be immediately detected. The MPS installation is a very delicate handling task especially due to the very limited free space after the MPS16 has been installed.
The successive centroids ðx ðpÞ c ; y ðpÞ c Þ of the laser spots recorded at each pass p, is dubbed circulation trace. During the installation of the first 16 MPS a very good superposition of all the passes, i.e., a flat circulation trace, has been observed after a quick alignment optimization of the injection and the M2 rotation axes and focus. During the installation of the next 15 MPS, a total of 8 MPS have been unfortunately touched or shocked inducing a significant misalignment (ϵ > 15 μrad) due to the delicate installation operation in the tight space between successive MPS bodies. The flexure blade is indeed extremely sensitive to a mechanical shock on the differential screw which is protruding on the back side of the blade. Consequently every faulty MPS is removed from the support and reset on the autocollimator bench as close as possible to its cascade value setting, for both axes. This required procedure unfortunately multiplies the number of handlings and consequently the risk of shocks and misalignments of other MPS. In addition, a temperature increase in the LBC ISO5 clean area due to human activities constrains the MPS installation time. The parallelism cannot be preserved during the mechanical installation for all the MPS within the tolerances. The maximum misalignment observed in the installed MPS is ≈50 μrad. The average misalignment of the MPS is below 20 μrad. Consequently some of them must be steered in-situ, using the IIOS as a high resolution autocollimator with the IP coordinates as reference mark. This method has been chosen to finalize the alignment.

B. Misalignment sensitivity
To give a better understanding of the LBC alignment behavior, the dependence of the circulation trace recorded by the IIOS on the various degrees of freedom-the injection angle u x , u y , the axes of M2, namely TX; TY; TZ; RX and RY, and the 31 pair of MPS angles (ϵ x , ϵ y )-is computed with a model of the system. The transverse translations TX and TY and rotations RX, RY are degenerated, as can be deduced from Figs. 22 and 23. We consider that the accuracy in the process of the optics manufacturing, positioning in its mount and alignment allow us to make the assumption that the TX and TY are determined and frozen by the pre-alignment accuracy. A compensation for a residual transverse misalignment of M2 by a rotation of M2 does not affect the optical performances as far as the misalignment stays within the paraxial approximation with an off centering error of M1 and M2 < 0.5 mm, which is the case given the prealignment tolerance. The M2 rotation RX creates two populations in the circulation trace as shown in Fig. 22. The odd laser spot population is shifted by ∼2 × ROC 2 × RX (∼2 × ROC 2 × RY) at the IP on the y-axis (x-axis) while the even laser spots remain unchanged. A translation TZ along the z-axis of the LBC creates a figure-8 at the IP centered on the first pass p ¼ 0. The odd (even) spots rotate anticlockwise on the top (bottom) part of the trace, as illustrated on Fig. 24.
A variation of the injection angle du x (du y ) generates two populations of odd and even spot centroids, that are displaced on the y (x) axis symmetrically with respect to the position of the spot p ¼ 0 when no misalignment is made as illustrated in Fig. 25. A MPS misalignment is similar to a combination of injection misalignments du x and du y but creates 3 laser spot populations. The population corresponding to passes p < k, before the incriminated MPSk is not modified. Two populations for the subsequent passes are symmetrically created, as if an injection misalignment was present for these passes only. The simulation of the circulation patterns induced by parallelism defects of ϵ 0 x ¼ 10 μrad for MPS11 and MPS13 are displayed in Fig. 26 (Top). The simulation of the circulation trace for a parallelism misalignment of ϵ 0 x ¼ 10 μrad of MPS13 only is shown in Fig. 26 (Bottom).
The typical pattern generated by one of the element of the LBC can help to optimize the alignment process. As a conclusion, the laser beam spot centroids misalignment generated pattern and circulation trace, can be divided in 3 groups: the odd spots centroid, the even spots centroid and the first pass spot. Using the discrimination in the laser spot centroids the spatial alignment has been completed.

C. Spatial alignment
Once the MPS installation process is completed the in-situ steering of the MPS mirror is done exploiting the behavior of the LBC under misalignments. First the injection is set to get the centroid of pass p ¼ 0 on the IP reference (x 0 , y 0 ) coordinates. The MPS1 is unmounted and controlled on the autocollimator bench and reinstalled. Then RX and RY are tilted to super impose the laser spot of pass p ¼ 1 and on the pass p ¼ 0. The MPS2 is unmounted, controlled, and reinstalled. Then the parallelism of all the other MPS are sequentially tuned in-situ by  superimposing the pass p with the pass p − 2. At the end of the process the obtained circulation trace is displayed in Fig. 27. Even if the difference is kept below ∼10 μm, between the pass p and p − 2 during the in situ alignment process, which corresponds to MPS parallelism alignment error of ≤ 8 μrad, we can see that the MPS12 has potentially drifted during the retuning of the next MPS and the pass p ¼ 2 is not super imposed to the pass p ¼ 0. This drift is identified to be due to the differential screw shearing. Looking to laser spot centroid at the IP with the IIOS (Fig. 28-top), the two populations (even/odd) of laser spots may be distinguished, the RX and RY rotation are used to overlap the two populations. The first pass is then isolated. The injection is steered to get the pass p ¼ 0 at x 0 ≈ 148 μm, y 0 ¼ 223 μm, to have 2 symmetrical populations respect to the pass p ¼ 0 which correspond to the misalignment error of the MPS1 (Fig. 28-bottom). From the plot of the Fig. 29 the MPS1 misalignment error is estimated to ϵ x ∼ 24 μrad and ϵ y ∼ 38 μrad. MPS1 is retuned in-situ to overlap the two populations with the first pass p ¼ 0 as can be seen by comparing the bottom plot of Fig. 28 and the plot of Fig. 29. This is the end of the manually driven in-situ alignment of the MPS of the LBC. The M2 rotation axes and the injection angles are automatically scanned to minimize the observable introduced in Ref. [29], namely the transverse distance to the barycenter (TDB). For the pth laser recirculation pass, the distance of its centroid position with respect to the barycenter of the 32 laser spot centroid positions is given by: and the average TDB of the 32 values of TDB n yields a performance factor of the alignment. As mentioned in Ref. [29], the value of the TDB is related to the relative time average spectral density (TASD r ), which is the relevant figure of merit for γ-ray users [10]. The values of TDB n are shown on Fig. 30  For an optimized alignment the spot sizes are measured for each pass and fitted to Gaussian laser spot shape. The average waist are w x ¼ 26 AE 3 μm and w y ¼ 25 AE 3 μm. As seen on Fig. 31 the beam spot sizes remain almost constant along the laser beam circulation, a slight decrease is however visible. The expected value is w 0 ∼ M 2 λ 0 EFL=ðπw M Þ ¼ 1.1λ 0 ROC=½πw M ð1 þ cos ΦÞ ∼ 25.6 μm given the beam size w M on the parabolas, the effective focal length EFL and the M 2 of the CIRC beam at the exit of the SALT. The dotted value is the waist design value. To show the conservation of the good wavefront quality the pass p ¼ 0 and p ¼ 31 are shown in the Fig. 32. We can see the diffraction due to the fact that the CIRC beam diameter was slightly too large with respect to the maximum waist design value of the circulating beam at the entrance of the LBC.
In the final integration of the complete laser system composed by the main high power IP laser and the laser beam transport line up to the interaction point, a fine tuning of the high power laser beam size and divergence is foreseen to accurately tune the laser waist in the LBC.

D. Synchronization on the optical reference
After the LBC alignment is optimized, the laser beam circulation frequency is synchronized on the ELI-NP-GBS reference clock. The ELI-NP-GBS synchronization system has been manufactured by Menlo Systems [67] and consists in a rf reference clock ν rf ¼ 2856 MHz, a subharmonic clock distribution and a client synchronization. The reference signal is provided by an ultralow phase noise μ-wave crystal reference master oscillator (RMO) [56]. The absolute jitter is measured below 60 fs on the range 10 Hz-10 MHz. The RMO provides a stable reference for the optical master oscillator (OMO). The OMO is an Er-doped mode-locked laser oscillator, that encodes the reference timing information in the frequency repetition rate ν OMO ¼ ν rf =46 ¼ 62.086956 MHz of short optical pulses at 1560 nm. Using such a configuration, the RMO guarantees the long term stability of the OMO, and imprints its low-frequency noise to the whole GBS facility timing. The OMO pulses are amplified and distributed to all clients. One of the channels is further amplified and converted by a second harmonic generator to a 780 nm free space output, used as a seed for the photocathode laser system. Four main clients have been identified in the ELI-NP-GBS machine; two interaction point lasers and two rf extractors (one at 2856 MHz and one at 5712 MHz) that provide the rf reference to the low level rf systems and power units.
The clients will be reached by means of four dispersion compensated fiber links with active length stabilization, in order to deliver short sub-200 fs synchronous pulses to the end users. A full description of the synchronization system has been reported in [23]. The frequency of the laser-beam oscillator embedded in the SALT of each IP modules of ELI-NP-GBS is locked on the reference frequency of the OMO received at the IP location through a nonstabilized fiber link. The synchronization is done thanks to an electrooptical synchronization module [68].
During the optical commissioning of the LBC, the whole synchronization system was not available on site. Thus the laser oscillator embedded in the SALT is left free run during operations. Its frequency stability is indeed judged sufficient to allow for the whole synchronization procedure to be done.
Using the delay line (DL), see Fig. 13, the first pass p ¼ 0 of the CIRC beam at the IP and a laser pulse of the REF beam are synchronized at the IP using the first order laser field linear autocorrelation technics described in Sec. III [IIOS]. Then for each pass an automatic scan is realized to acquire the image of the interference pattern as function of the MPS rotation stage position. For each motor step, several 16-bit images are acquired with the ICCD detector, each corresponding already to hardware accumulations on the camera. They are then dynamically converted to 8-bit with respect to the actual measured maximum, blurred with a Gaussian with ð5 × 5Þ pixels kernel and dynamically threshold in order to look for lines by means of a Hough transform algorithm [69]. Statistical analysis of the orientation of the lines is used to extract the angle by which the image is then rotated so that the lines become vertical. The lines are summed up and multiplied by a sliding Blackmann-Harris [70] window of width 150 pixels. The result is stored in a two dimensional array that is further discretely Fourier transformed, the magnitude of which is pedestal subtracted. The two main peak magnitudes are averaged to estimate a quantity, proportional to the contrast of the fringes, that is maximized by the automatic scan procedure. An example of the obtained results is given in Fig. 33. The dataset data1 corresponds to the first synchronization over the full range of the MPS rotation. The second, data2, corresponds to a more refined scan taken 3 hours after the initial measurement. The synchronization has been observed to be very stable over several days. No corrections on the laser circulation beam path length have been required to keep the accuracy of the synchronization scan within an average error over the 32 passes of AE37 fs. Presently the accuracy varies as function of the circulation pass number due to the fact that some passes are less reflected by the TFP, thus degrading the contrast in the ICCD camera.

E. Stability tests and expected performances
Once the laser beam circulation is optimized and synchronized, the performances are maintained by locking the injection of the laser (u x , u y ) on the injection references (IB.NF2, IB.FF2, see Fig. 14) defined by the alignment optimization and thermalizing the optomechanical structure of the LBC to the setting temperature at AE0.1°C. The system has been pumped down in primary vacuum for the test purpose. The stability of the laser beam circulation has been monitored for several days of operation with the low power alignment laser and without active thermalization of the interaction point module. The sensitivity of the system to the temperature has been determined and is shown on Fig. 34. The stability of the circulation has been monitored regularly over 15 hours corresponding to a duration larger than a foreseen day of operation. The laser beam injection pointing angle is continuously stabilized on references while the M2 axes have been kept untouched. The variation of the TDB, along with the temperature variation inside vacuum, are presented in Fig. 34. A clear correlation between the TDB and the measured temperature of the MPS body under vacuum is visible. To ensure the stability of the γ-ray flux, while considering a perfect electron beam, the temperature stability of the aligned and synchronized LBC must be within ΔT < 100 mK. A week of operation of the LBC has unveiled that a daily optimization is required in order to ensure that TDB < 15 μm. We experience during the low power commissioning that a starting check of the circulation and a fast optimization (less 15 minutes) in the morning was enough to maintain continuous performances for more than 10h.
The performance obtained during the optical commissioning is considered to be representative of the final operation of the IP module. In fact, for the injection of the high power laser, the only property of the IP module that remains non tested is the laser damage threshold. However, the laser beam damage threshold for all the optics integrated in the LBC and IB exceed the corresponding requirements, as validated by actual measurements made on witnesses. The M 2 -quality of the high-power laser-beam has been demonstrated [71] independently. It remains to demonstrate the quality of the overlap and synchronization of the electron and laser beams at the IP. However, tools to achieve this are implemented in the design of the IP module. The spatial overlap will be achieved with a micrometer-level precision by means of the OTR light that will be analyzed with the IIOS. The two CBPMs will be used to keep constant the electron beam axis once it is overlapped with the IP defined by the LBC alignment. The γ-ray flux will be monitored bunch by bunch thanks to a diamond-based fast detector which will allow a precise and instantaneous information about the evolution in time of the system. A feasibility demonstration for the use of such a detector in this context has been made independently [72]. Once the electron beam steered at the IP to maximize the γ-ray flux, it is entirely controlled by the steering magnets of the linac, the beam positions and pointing angles of the high-power laser-beam in the IB. The CBPMs allow to monitor precisely the electron beam axis at the IP. The input (output) positions and pointing laser-beam diagnostics located in the IB (LD) are the only available tools to monitor the high-power laser-beam, the LBC becoming a black-box. The whole optomechanical structure, including the quality of the ground floor on which the IP module is set, has to ensure that the spots at the IP are stable within a μm. It has been demonstrated that this is the case in the hall where the optical commissioning has been conducted. The IP coordinates defined as x IP ¼ P 31 p¼0 x p =32 and y IP ¼ P 31 p¼0 y p =32 have been monitored under vacuum and are displayed in Fig. 35. The IP position remains stable over one workday of operations. A small drift along the x-axis is however noticeable. It is mainly attributed to the temperature variations. Small variations are expected especially due to the thermal deformation of aluminum parts of the parabolic mirror mounts.
The model [29] is used to simulate 10000 randomly chosen positions and angles of all the optical elements in the ranges AE20 μm and AE20 μrad, respectively. The gain of the LBC in terms of useful optical power at the IP for the production of γ-rays, reads: It is roughly estimated as the product of the measured LBC laser beam transmission T LBC ¼ 0.84 AE 0.03, the number N of circulation passes and the relative TASD r . The latter, which represents the loss in spectral density of γ-rays relative to a perfect theoretical system, is shown on Fig. 36. The best optimization of the alignment experimentally yields a TDB ≈ 8 μm corresponding to a TASD r ¼ 0.96 and leading to an effective gain of g ¼ 25.8 AE 0.9. During days of the commissioning performance tests, and in absence of temperature stabilization of the LBC, the TDB is measured to vary between 8 μm and 15 μm, which corresponds to 0.85 < TASD r < 0.93, thus a relative 5% variation in the gain of the LBC.

VII. SUMMARY AND OUTLOOK
The optical commissioning at low power, of the unique optomechanical system conceived for the ELI-NP-GBS γ-ray source, is described. Performances of the procured critical optical elements are given. The setting and integration of the MPS, that are key element of the system, is detailed. The complete alignment procedure has been described and the obtained performances commented. The demonstration of a high stability of the laser beam circulation constitutes in an important achievement towards the installation of the system in the ELI-NP-GBS accelerator hall. It is estimated that the system corresponds to a gain in excess of 25 for the production of γ-rays by Compton scattering. It demonstrates that the initially foreseen performances of the γ-beam source will be reached for both interaction points. Indeed with the use of the highpower laser-beam of ELI-NP-GBS, delivering 400 mJ at 100 Hz, an equivalent power of more than 1 kW will be used for collisions with the electron beam, delivering a small bandwidth high-flux source. However, it must be underlined that the stable operation of such a highly technological system heavily relies on the good stability of the environment, in particular the ground stiffness and vibration, the cleanliness of the ambient air, the stability of the temperature and the humidity. Next, a study is on going to improve on the MPS flexor blade mechanical stability before the final integration in the ELINP-GBS tunnel, in order to facilitate the integration process.

ACKNOWLEDGMENTS
We wish to acknowledge EuroGammaS consortium for the support and all the fruitful discussion all along the development of the laser beam circulator. Especially our colleagues from INFN-LNF, INFN-Milano, Università La Sapienza, Institut Fresnel in Marseille, and Amplitude laser company. The authors would like to acknowledge the efforts of the LAL staff for the continuing support. We also thanks the ALSYOM PUPS group for their help getting the right ambient conditions during the commissioning.