CERN-SPSC-2013-013 / SPSC-TDR-003 24/09/2013 rtnDie lsaWkfil Acceleration Wakefield Plasma Proton-Driven A n plctost iheeg colliders. energy high will acceleration to this wakefield applications will plasma and and experiment proton-driven The 2016 of program. of tests larger-scale experimental end future year detail. the 3–4 inform in initial at an discussed expected by are are followed equipment experiment be diagnostic the the to and protons as cell First main required plasma The infrastructure the and as interactions. area on well based experimental plasma is the and program components, beam MeV experimental 20 accelerator The of about simulations meters. from few numerical electrons a of detailed within bunch above scale a GeV fields accelerate the peak to to with harnessed plasma be the will in that wakefield be strong will bunches a proton will initiate facility long micro-bunches then These the CNGS where micro-bunches. the shorter cell in significantly plasma into m beam 10 modulated proton a into compared SPS injected length The be reduced will much accelerators. a of linear but proposed energy could to high technology of This colliders time. future proton- first to the demonstrate lead for to acceleration order wakefield in plasma formed driven been has Collaboration AWAKE The Abstract W K einReport AWAKE Design xeieta CERN at Experiment AWAKE Collaboration 1 1 GV/m 2 Contents

1 Executive Summary ...... 5 2 Introduction ...... 6 3 The Self-Modulation Instability ...... 7 3.1 Transverse Modulation of a Long Bunch ...... 7 3.2 Injection and Acceleration of Witness Electrons ...... 9 3.3 Seeding of the SMI ...... 10 4 Baseline Design ...... 11 5 Plasma Sources ...... 14 5.1 Metal Vapor Plasma Source ...... 14 5.2 Argon Discharge Plasma Source ...... 16 5.3 Helicon Plasma Source ...... 16 5.4 Plasma Density Measurements ...... 18 6 Particle Beams Diagnostics ...... 19 6.1 Proton Bunch Diagnostics for SMI ...... 19 6.2 Coherent Transition Radiation Diagnostics ...... 20 6.3 Electro-Optical Sampling ...... 21 6.4 Electron Spectrometer ...... 22 7 Electron Source ...... 23 8 The AWAKE Facility at CERN ...... 24 8.1 Introduction ...... 24 8.2 Experimental Area ...... 26 8.3 Proton and Electron Beam Lines ...... 29 8.4 The SPS Proton Beam ...... 32 9 Project Planning ...... 35 9.1 Timeline ...... 35 9.2 AWAKE Physics Program ...... 35 10 Summary ...... 36

3 4 1 Executive Summary New acceleration technology is mandatory for the future of particle physics. A promising approach is to exploit the properties of plasmas. Past research has focused on creating large-amplitude plasma waves by injecting an intense laser pulse or an electron driver bunch into the plasma. However, the maximum energy gain of accelerated particles in a single plasma stage is limited by the energy of the driver. Proton bunches, being much more energetic, are the most promising drivers of wakefields to accelerate electrons to the TeV energy scale. The objectives of the AWAKE experiment are to understand the physics of the acceleration process, to demonstrate high-gradient acceleration with a proton bunch, and to develop necessary technologies for the long-term perspectives of proton-driven plasma wakefield acceleration. The AWAKE experiment fulfills one of the primary recommendations of the European Strategy for Particle Physics which advocates vigorous R&D in high-gradient accelerating techniques. The AWAKE experiment will use proton bunches for the first time ever to drive plasma wake- fields. A plasma will be used to modulate the long proton bunch into a series of ‘micro-bunches’ which then generate a strong plasma wakefield. Our goal is to accelerate electrons injected into the plasma wakefield to the GeV scale in a few meters of plasma. The evolution of the properties of the proton and electron bunches in the plasma will be studied experimentally using state-of-the-art diagnostic tools and will be compared to predictions from detailed numerical simulations. This information will provide the basis for designing next-generation experiments at CERN. In parallel, we will develop long and uniform vapor, discharge and helicon plasma cells and develop schemes for proton-bunch compression. These are key for the long-term success of proton-driven plasma wakefield acceleration. We will test bunch compression schemes in the CERN accelerators and apply them to the proton bunches used to drive the wakefields. We will also install the different plasma cells in our experimental setup and investigate in detail their characteristics for acceleration. The experiment will inform future larger-scale tests of proton-driven plasma wakefield acceleration and applications to high energy colliders. A detailed comparison of two sites for the AWAKE facility has been performed, based on design studies of the proton beam delivery in the SPS, the primary beam lines, the experimental area, civil en- gineering, general services and infrastructures for the facility as well as radiation protection and general safety aspects. The CNGS facility satisfies best the requirements of AWAKE and we therefore propose to carry out the experiment in the CNGS beam line of the SPS. Assuming approval of AWAKE in mid 2013, first protons could be sent to the plasma cell at the end of 2016. Considering four years for the completion of the electron source system, the electron beam will be operational at the end of 2017. Operation and data taking is planned for 3–4 years. Beam is requested for approximately 4 periods of two weeks per year, with bunch repetition rates of approximately 1/30 Hz. Further experimental efforts will be evaluated based on the results from the initial running. The proposed program requires expertise in proton accelerators, electron accelerators, plasma physics, wakefield acceleration, experimental physics, theory and simulations. The AWAKE Collabora- tion, with the strong backing of CERN, has the expertise to fulfill these requirements.

5 2 Introduction Particle accelerators are the fundamental research tools of the high energy physics community for study- ing the basic laws that govern our Universe. Experiments conducted at the LHC will give us new insights into the physical world around us. Complementing this, future lepton colliders should reach the TeV scale. Circular electron colliders are not feasible at these energies; hence future TeV accelerator designs are based on linear colliders. However, as the beam energy increases, the scale and cost of conventional machines become very large. For a linear accelerator, the size and cost depend on the maximum accel- erating gradient in RF cavities. At present, metallic cavities achieve maximum accelerating gradients around 100 MV/m. To reach the TeV scale in a linear accelerator, the length of the machine is therefore tens of kilometers. It is natural to think about how to make future machines more compact, and plasma acceleration is a possible solution. A plasma is a medium consisting of ions and free electrons; therefore, it can sustain very large electric fields (> GV/m) [1,2]. In the last few decades, more than 3 orders of magnitude higher acceleration gradient than in RF cavities have been demonstrated with plasmas in the laboratory [3, 4]. Beam-driven plasma wakefield acceleration experiments performed at SLAC [4] successfully doubled the energies of some of the electrons in the initial 42 GeV beam in less than 1 m of plasma. Generally speaking, a plasma acts as an energy transformer; it transfers the energy from the driver (laser or beam pulse) to the witness bunch that is accelerated. Current proton synchrotrons are capable of producing high energy protons, reaching up to multi TeVs (the LHC), so that a new accelerator frontier would be opened if we could efficiently transfer the energy in a proton bunch to a witness electron bunch. It has been recently proposed to use a high energy proton bunch to drive a plasma wakefield for electron beam acceleration [5]. Numerical simulations have shown [6] that a 1 TeV bunch, with 1011 protons and an rms bunch length of 100 µm as driver could indeed excite a large amplitude plasma wave. Surfing the appropriate phase of the wave, an electron bunch reaches energies over 600 GeV in a single passage through a 450 m long plasma. Recent studies [7, 8] have shown that similar gradients can be reached with a modulated long proton bunch, opening the path for immediate experimental investigations with the existing proton bunches at CERN. The modulation of the proton density on axis results from the transverse focusing and defocusing field along the bunch. For coherent wakefield excitation, this is equivalent to having a series of ultra-short proton bunches with an effective length and period set by the plasma wavelength. The AWAKE experiment will use proton bunches for the first time ever to drive plasma wakefields. The main physics goals of the experiment are:

– to study the physics of self-modulation of long proton bunches in plasma as a function of beam and plasma parameters. This includes radial modulation and seeding of the instability. – to probe the longitudinal (accelerating) wakefields with externally injected electrons. This includes measuring their energy spectrum for different injection and plasma parameters. – to study injection dynamics and the production of multi-GeV electron bunches, either from side injection or from on-axis injection (with two plasma cells). This will include using a plasma density step to maintain the wakefields at the GV/m level over meter distances. – to develop long, scalable and uniform plasma cells and develop schemes for the production and acceleration of short bunches of protons for future experiments and accelerators.

The results of the experiment will inform future larger-scale R&D experiments on proton-driven plasma wakefield acceleration such as acceleration of electron bunches to ∼ 100 GeV in ∼ 100 m. The results will allow first designs to be made using this technology for future particle physics facilities such as TeV scale ep or e+e− colliders. Since the submission of a Letter of Intent [9] to the SPS Committee in May 2011, the group of involved scientists has been formalized into the AWAKE Collaboration. This document extends the work

6 presented in the Letter of Intent and provides a design of the experiment. The experimental program will be staged with first protons injected into a plasma expected at the end of 2016. This will be followed by an initial 3–4 year program leading to the acceleration of electrons ultimately to the GeV scale in a few meters of plasma and the testing of different plasma source technologies which could be scaled to much greater distances for future experimentation or facilities. In the following main body of the Design Report, brief descriptions of the important elements of the experiment are given, with more detailed descriptions given in the accompanying Technical Notes [10]. The basics of plasma wakefield acceleration relevant to the AWAKE experiment are out- lined in Section 3. The baseline design of the experiment is given in Section 4, including an overview of the experiment and the expectations from simulations. The three different plasma cell designs are explained in Section 5. The equipment to diagnose the properties of the various beams and the plasma, including the electron spectrometer, are discussed in Section 6. In Section 7, the electron source is de- scribed. Section 8 gives details of the AWAKE facility at CERN, including the design of the beam lines, choice of site and layout. The timeline and research program is given in Section 9. The report concludes with a brief summary in Section 10.

3 The Self-Modulation Instability Plasma wakefields are usually driven by laser pulses or particle bunches approximately one plasma wave- length long: σz ≈ λpe. Here λpe = 2πc/ωpe is the wavelength of a relativistically moving plasma wave 2
1/2 with electron plasma angular frequency ωpe = nee /0me in a plasma with density ne. A laser driven accelerator is known as a laser wakefield accelerator (LWFA) [1]. An accelerator driven by a charged particle bunch is known as a plasma wakefield accelerator (PWFA) [2]. The longitudinal ampli- 1/2 tude of the wakefield scales as the wave breaking field [11]: EWB = mecωpe/e ∝ ne . This scaling therefore favors short pulses or bunches and large plasma densities to reach large amplitude wakefields. These wakefields, for intense enough drivers, have accelerating/decelerating longitudinal components (Ez ∼ EWB) and transverse focusing/defocusing components with comparable amplitudes. In the linear wakefield regime, these fields vary periodically behind the drive bunch and have a π/2 phase difference. Unlike recent plasma wakefield accelerator experiments which employed short bunches (σz . λpe) to drive intense wakefields, the AWAKE experiment will use much longer proton bunches (σz & 100λpe) to generate plasma wakefields. The AWAKE experiment will thus operate in the so-called self-modulated plasma wakefield accelerator regime [7]. In this regime, the maximum plasma density for wakefield ex- citation is given by the condition that the plasma return current must flow outside the drive bunch. This condition is satisfied when the bunch transverse size σr is smaller than the cold plasma collisionless electron skin depth c/ωpe or when kpeσr < 1 (kpe = ωpe/c). When this condition is not satisfied the bunch can be subject to the current filamentation instability. The instability breaks the bunch into trans- verse current filaments [12] and prevents the efficient excitation of plasma wakefields. For σr = 200 µm, 14 −3 setting kpeσr = 1 yields ne = 7 × 10 cm , λpe = 1.2 mm and EWB = 3 GV/m. With relativistic bunches, energy gain and loss does not lead to significant dephasing between ∼ 1 ∆γ drive bunch particles over meter-scale plasma lengths: ∆L = γ2 γ L  λpe (for particles with energy γ and γ ± ∆γ and plasma length L). This means that there is no longitudinal bunching. However, the transverse wakefield components can periodically focus and defocus the particles that typically have non relativistic transverse velocities (hv i ∼  c  c where  is the beam transverse emittance and σ its ⊥ σ0 0 waist size) [13].

3.1 Transverse Modulation of a Long Bunch

When a long and narrow particle bunch travels in a dense plasma; i.e. when σz  λpe, it is subject to a transverse two-stream instability or self-modulation instability (SMI) [7]. The low amplitude transverse wakefields driven by the long bunch modulate its radius with wavelength ∼ λpe. This generates a micro-

7 0.0 0.20 8.3 meters Propagation direction

-0.4 ] ] e0 e0 [n 0.10 [n pe -0.8 radius (r) plasma beam ρ ρ r

-1.2

0.00 Distance in beam (z)

Fig. 1: Example of a self-modulated proton bunch resonantly driving plasma wakefields sustained by the plasma density perturbation. The plasma electron density is shown increasing from white to blue and the proton density increasing from yellow to dark red. bunch pattern as seen in Fig.1. This periodic bunch density modulation then resonantly drives wakefields to larger amplitudes, thereby providing the feedback mechanism for the SMI to develop. The SMI is a convective instability that grows both along the bunch and along the plasma. If the drive bunch propagates in a uniform density plasma, then the instability destroys the micro- bunches soon after the maximum field is reached [14]. The reason lies in the slow motion of the defocus- ing field regions with respect to the bunch. This effect causes a strong decrease in the peak accelerating field, as seen in Fig. 2(a). -13.5 1.2 (a) (b) (c) 1.0 -13.6 25

, cm

GV/m 0.8 20 z-ct -13.7

z

E, 0.6 15 -13.8

0.4 % /10 GeV 0.2 -13.9 5 0.0 -14.0 0 0 246 8 10 0 246 8 10 1.8 1.9 2.0 2.1 2.2 z,m z,m Energy, GeV

Fig. 2: (a) Maximum amplitude of the accelerating field Ez excited along the bunch plotted as a function of position along the plasma. (b) Positions along the bunch (z − ct) where the wakefields are both accelerating and focusing for witness electrons are shown in grey as a function of propagation along the plasma. This position varies over the first 4 m of propagation and remains at the same z − ct position after that.

As the SMI grows, the interplay between bunch radius and wakefield amplitude leads to an ef- fective wakefield phase velocity slower than that of the drive bunch [15, 16] as seen in Fig. 2(b). The figure shows the location of the accelerating and focusing fields along the bunch (z − ct) as a function of propagation distance along the plasma z.

8 Once the SMI saturates, these two velocities become equal. It is possible to avoid the destruction of the micro-bunch structure by a proper step up in the plasma density (Fig. 3(a)) which modifies the instability growth in such a way that the field motion relative to the bunches stops at the optimal moment.

(a) 1.2 (b) stepped-up 1.0 np dne 0.8 0.6

z,max 0.4

E,GV/m 0.2 uniform 0 z 0 20 40 60 80 100 z,m Fig. 3: (a) Plasma density profile used in the simulation for locking the accelerating field close to its peak value. (b) The maximum wakefield amplitude versus the propagation distance for the stepped-up (blue line) and uniform (red line) plasmas for a simulation with an LHC bunch.

The field evolution for the stepped plasma profile is shown in Fig. 3(b) in comparison with the uni- form plasma case for simulations of proton bunches from the LHC. With the density step, the wakefield is preserved for a long distance, at a large fraction of the maximum amplitude. For AWAKE, we expect to achieve a constant gradient of 0.3 − 0.8 of the maximum amplitude field depending on the parameters of the proton bunch. The maximum gradient will depend on the success of our bunch compression ef- 11 forts. For example, for a bunch of 3 × 10 protons compressed so that σz = 3 cm and σr = 50 µm, the maximum gradient would be about 4 GeV/m. Assuming we could maintain 50 % of this peak gradient with a density step, then large energy gains (10 − 100 GeV) would be possible with compressed and modulated proton bunches from the SPS.

3.2 Injection and Acceleration of Witness Electrons The SMI leads to bunch radial modulation that can be observed using coherent and incoherent radiation processes from the drive bunch particles (see Section 6). The longitudinal wakefields can be probed by injecting low energy witness electrons that can be trapped and accelerated. However, since the initial wakefield phase velocity is slower than that of the drive bunch, the injection has to occur at a point along the plasma after the SMI has saturated. Otherwise the electrons would dephase with respect to the wakefields and be defocused by their transverse component and lost. In the case of a single plasma cell the electrons can be injected from the side of the plasma [17]. In this side injection scheme, electrons of a long (σz > λpe), low energy (a few MeV) bunch injected with a small angle (a few milliradians) with respect to the drive beam axis can be trapped and accelerated. A fraction of the electrons reach the beam axis, dephase, accumulate at the peak accelerating wakefield and form short bunches in a few accelerating buckets (see Fig. 4). They are then accelerated to high energies with a narrow energy spread (% level). This side injection scheme relaxes the timing tolerances for injection and has a particle trapping efficiency between 5 and 40%. With two plasma sources, the first source can be used as modulator for the drive bunch. The second one, resonantly driven by the self-modulated drive bunch, can be used as an accelerator for externally injected witness particles. These particles can be injected either sideways, as with a single source, or directly on axis. For on-axis injection, a short (when compared to λpe) bunch is required and the injection timing also has to be within a fraction of the wakefield period. However, on axis injection

9 ct=3.6 m ct=3.8 m ct=4.6 m

0.3

r, mm 0.2

0.1

0.0 -13.8 -13.7 -13.6 -13.8 -13.7 -13.6 -13.8 -13.7 -13.6 z-ct, cm z-ct, cm z-ct, cm

Fig. 4: Maps of the wakefield potential (false colors and moving to the right) at three positions along the plasma. At z = 3.6 m (left hand side image), the low energy electrons (shown as black dots) are injected from the side (top on the figure) towards the wakefield. At z = 3.8 m (middle image) the electrons have reached the wakefield potential wells, some are reflected while some reach the axis and can be trapped. At z = 4.6 m (right hand side image), two trapped electron micro-bunches are visible near the axis (r = 0) and a few electrons are still drifting out radially. also allows for full capture of the witness bunch. Very narrow final energy spreads and larger energy transfer efficiencies can be expected by using accelerator physics techniques such as beam loading of the wakefields, as was shown in PWFA simulations [18].

3.3 Seeding of the SMI The SMI can in principle grow from noise in the plasma and the drive bunch. However, seeding of the instability considerably shortens the plasma length needed for the SMI to reach saturation. Calculations [19] and simulations show that the noise level is very low and that the SMI would not grow to a detectable level over meter-scale plasmas. More importantly, seeding the SMI fixes the phase of the wakefields, a condition necessary to deterministically inject a short witness bunch in the accelerating and focusing phase of the wakefields. The SMI can be seeded by a short laser pulse or particle bunch driving low amplitude wakefields in a preformed plasma ahead of the long drive bunch. A sharp (when compared to λpe) boundary between the drive bunch and the plasma, such as a cut in the bunch current profile or a relativistic ionization front propagating within the drive bunch also seeds the SMI [20]. The ionization front seeding method uses a short laser pulse and has many advantages since it also serves as plasma creation method (see Fig. 5). The SMI leads to a radially symmetric modulation of the bunch charge density. However, there is an asymmetric competing instability known as the hose instability (HI) [21]. This instability is similar to the beam break-up (BBU) instability in RF accelerators. It grows from noise in the transverse dis- placement of the bunch-slice centroids and results in a non-axially symmetric displacement of the bunch along its length. The HI has a growth rate comparable to that of the SMI and the two instabilities directly compete. Simulations indicate that seeding helps the SMI dominate over the HI [15, 16]. The AWAKE experimental program will allow us to investigate the physics of the SMI, electron

10 2 L=0 m laser pulse (a)

1 proton beam

r, mm

0 2 L=4 m electrons (b)

1

r, mm

0 2 L=10 m (c)

1

r, mm

0 -30 -20 -10 0 z-ct, cm 1020 30 (d) (e) Fig. 5: Distribution1.0 of the beams at (a) the entrance to1.0 the plasma, (b) after propagating 4 m in the plasma, and at (c) the exit from the plasma cell. Protons are blue, electrons are red and the laser pulse is the line at z − ct = 0. The laser pulse seeds the SMI for the proton bunch.

r, mm r, mm

0.0 0.0 injection and-13.8 acceleration -13.7 processes over-13.6 a wide range of parameters-13.8 and -13.7 configurations. -13.6 z-ct, cm z-ct, cm 4 Baseline Design The general layout for the baseline design of the experiment is shown in Fig. 6. The proton beam propa- gates through a 10 m long plasma cell, excites the wakefield and becomes modulated by this wakefield. The short laser pulse propagates collinearly with the proton beam and serves the dual function of cre- ating the plasma and seeding the SMI. The electron bunch enters the plasma cell parallel to the proton beam with an initial offset of about 1 cm and is merged into the wakefield several meters downstream as soon as the proton beam is modulated by the SMI. Modulation of the proton beam radius is measured by electro-optical sampling (EOS) and optical and coherent transition radiation (OTR/CTR) diagnostics. The accelerated electron beam is characterized with a magnetic spectrometer. The nominal parameters of the incoming SPS beam are given in Table 1. Beams of other energies were also analyzed in the context of this experiment. It was shown that at substantially lower ener- gies (24 GeV, the energy of the protons in the PS) the excited fields are much lower because of quick emittance-driven blowup of the beam radius [22]. At energies above 75 GeV, the maximum field weakly depends on the driver energy, and the length of the high field region is roughly proportional to the square root of the driver energy (see the Technical Note [23] for more details). Therefore a proton beam energy of 400 GeV, compatible with the CNGS facility, was chosen. The optimum energy and angle for side-injection depends on the wakefield amplitude and phase velocity [17] and were found from simulations (Table 1). The peak accelerating field from simulations with the parameters of Table 1 is above 1 GV/m. Figure 7 shows the energy spectrum of electrons at the exit of the 10 m long plasma cell. Approximately 5% of injected particles are trapped and accelerated to form the electron bunch of energy 2.1 GeV, energy spread 3%, and normalized emittance 170 mm mrad. The average gradient witnessed by the electrons and obtained by dividing the energy gain by the accel- eration length (∼ 6 m) is ∼ 350 MeV/m. Later stages of the experiment will demonstrate higher energy gradients.

11 =%*8.! 85!2)$! .-!2)$! X! 34564789!85! H@0! 85!*+8(:.'/8:8.! E'$#F#$2! Z! ^! >#%2$'*?(! 34! 4;356<89! "#$%&! =%*8.!1)&*8! K! "'()*! !!!Plasma , !!!!!!!!!! 7-10m, 1015 cm-3 + !G)/+! 3! 07E! I((8&8.%?'$! =%*8.!>)/+! @ABCDAB! 12 , + !-.'/!010! J! >#%2$'*?(*! >#%2$'*?(*! 6! W! Y!

Fig. 6: Baseline design of the AWAKE experiment: The proton bunch extracted from the SPS is injected with the ionizing laser pulse (1). The laser pulse and AL8!+.':'$!M)$(L!8N:.%(:8G!-.'/!:L8!010!O3PQ!#*!/8.28G!R#:L!:L8!#'$#F#$2!&%*8.!+)&*8!O3Q;!AL8!&%*8.!+)&*8!%$G!:L8!+.':'$!the proton bunch travel together through the metal vapor cell. The co-propagating ionization front shown in (2) provides the seeding of the SMI. Growth of M)$(L!:.%S8&!:'28:L8.!:L.')2L!:L8!/8:%&!S%+'.!(8&&;!AL8!('5+.'+%2%?$2!#'$#F%?'$!-.'$:!*L'R$!#$!O6Q!+.'S#G8*!:L8!*88G#$2!'-!the SMI and the resulting self-modulation of the proton bunch occurs over the first 3 − 5 m of plasma (3). The modulated bunch resonantly drives wakefields :L8!07E;!@$&T!:L8!-.%(?'$!'-!:L8!+.':'$!M)$(L!+%.?(#+%?$2!#$!07E!%$G!R%U8V8&G!8N(#:%?'$!#*!*L'R$!#$!O3PQ!%$G!OKPQ!<.'R:L!over the remaining length of plasma (4). The laser pulse is dumped (5) and the proton bunch radial modulation is measured using electro-optical sampling '-!:L8!07E!%$G!:L8!.8*)&?$2!*8&-5/'G)&%?'$!'-!:L8!+.':'$!M)$(L!'(().*!'S8.!:L8!V.*:!J5W/!'-!+&%*/%!OJQ;!AL8!/'G)&%:8G!(EOS) diagnostics (6) and optical transition radiation diagnostics (7). A RF-gun driven by a laser pulse derived from the ionizing laser produces a witness electron bunch (8). The electron bunch can be side-injected into the wakefield after the SMI has saturated (9). Downstream of the plasma, the electron bunch +.':'$!M)$(L!OKPQ!.8*'$%$:&T!G.#S8*!R%U8V8&G*!'S8.!:L8!.8/%#$!&8$2:L!'-!+&%*/%!OKQ;!AL8!&%*8.!+)&*8!#*!G)/+8G!OWQ!%$G!:L8!energy spectrum is measured using a broad acceptance magnetic spectrometer (10). +.':'$!M)$(L!.%G#%&!/'G)&%?'$!#*!/8%*).8G!)*#$2!8&8(:.'!'+?(!*%/+&#$2!OH@0Q!G#%2$'*?(*!OXQ!%$G!'+?(%&!:.%$*#?'$! .%G#%?'$!G#%2$'*?(*!OYQ;!I!.-52)$!G.#S8$!MT!%!&%*8.!+)&*8!G8.#S8G!-.'/!:L8!#'$#F#$2!&%*8.!+.'G)(8*!%!R#:$8**!8&8(:.'$!M)$(L! OZQ;!AL8!8&8(:.'$!M)$(L!(%$!M8!*#G85#$[8(:8G!#$:'!:L8!R%U8V8&G*!O*88!O6QQ\!%]8.!:L8!07E!L%*!*%:).%:8G!O^Q;!I]8.!:L8!+&%*/%! :L8!8&8(:.'$!M)$(L!8$8.2T!*+8(:.)/!#*!/8%*).8G!)*#$2!%!M.'%G!%((8+:%$(8!/%2$8?(!*+8(:.'/8:8.!O34Q;! Table 1: Baseline parameters of the AWAKE experiment. Parameter & notation Value 14 −3 Plasma density, ne 7 × 10 cm Plasma ion-to-electron mass ratio (rubidium), Mi 157 000 11 Proton bunch population, Nb 3 × 10 Proton bunch length, σz 12 cm Proton bunch radius, σr 0.02 cm Proton energy, Wb 400 GeV Proton bunch relative energy spread, δWb/Wb 0.35% Proton bunch normalized emittance, bn 3.5 mm mrad 9 Electron bunch population, Ne 1.25 × 10 Electron bunch length, σze 0.25 cm Electron bunch radius at injection point, σre 0.02 cm Electron energy, We 16 MeV Electron bunch normalized emittance, en 2 mm mrad Injection angle for electron beam, φ 9 mrad Injection delay relative to the laser pulse, ξ0 13.6 cm Intersection of beam trajectories, z0 3.9 m

25 20 15

%/GeV 10 5 0 1.8 1.9 2.0 2.1 2.2 Energy,GeV

Fig. 7: Energy spectrum of electrons side-injected near z ∼ 4 m with 16 MeV energy and accelerated until z = 10 m. The spectrum is narrow (∼ 3% width) and centered around 2 GeV.

Once trapped by the plasma wave, the electron bunch is very sensitive to the wakefield phase in which it resides. The wakefield phase is determined by the plasma density that must be constant with an accuracy ∝ λpe/4σz or ∼ 0.2% for our baseline parameters [24]. To provide this accuracy, instant ionization of a thermal equilibrium gas was chosen as the baseline design of the plasma cell. The final electron bunch parameters are not very sensitive to the injection parameters (see the Technical Note [23] for more details). Injection at angles or energies which are ±30% of the optimum values results in roughly twofold reduction of the bunch charge. The good “window” in the other two parameters is ∼ ±0.25 m for the injection point along the plasma and ∼ ±1 cm for the injection delay with respect to the ionizing laser pulse.

13 5 Plasma Sources For the baseline design, the plasma for the AWAKE experiment must have a number of characteristics:

– length L ≈ 10 m.

– radius Rp larger than approximately three proton bunch rms radii or ≈ 1 mm. 14 15 −3 – density ne within the 10 − 10 cm range. – density uniformity δne/ne on the order of 0.2% or better. – reproducible density. – gas/vapor easy to ionize. – allow for seeding of the SMI. – high-Z gases to avoid background plasma ion motion [25].

We are currently exploring three options for the plasma. The source for the first experiments will be a ru- bidium vapor source ionized by a short and intense laser pulse. The general laser and plasma parameters for the source are listed in Table 2. Because of the laser ionization process this source does not scale well to much longer lengths. Therefore we are investigating discharge plasma sources and a helicon source for plasma lengths longer than ten meters.

5.1 Metal Vapor Plasma Source 14 15 −3 Pure rubidium (Rb) vapor with neutral density n0 in the 10 − 10 cm range can be produced at temperatures between 150 and 200 ◦C (see Fig. 8a) and the Technical Note [26] for more details).The vapor density uniformity is achieved by controlling the temperature of the tube containing the vapor to ±0.5 ◦C (or about 0.2% around 450 K). This is achieved by circulating around the tube containing the Rb vapor synthetic oil from a heater that can stabilize the oil temperature to ±0.01 ◦C [27]. The system can be operated with a single oil temperature. The thermal insulation consists of a vacuum tube surrounding the tube containing the oil. Vacuum suppresses convection, the main cause of heat loss at these temperatures. If necessary, the temperature uniformity will be increased by including an external heated tube (Theat) around the vacuum tube (not shown in Fig. 8). These heating insulation techniques are common in temperature calibration devices where uniformities at the 0.001 ◦C level are achieved, although in smaller-size devices [28]. The Rb vapor will be produced by a commercially available source [29]. The vapor will be con- tained in a ∼ 2 cm radius tube closed at both ends by fast opening/closing valves (a few tens of ms opening/closing time). These valves are being developed in collaboration with industry. In the pressure range considered here (4 × 10−3 − 4 × 10−2 mbar at room temperature) and with the 2 cm radius of the source tube, the Rb is in the molecular to transitional flow regime (see the Technical Note [30] for more details). Calculations show that when the valves open the Rb density simply decreases from the location of the valves towards the tube middle length as a rarefaction wave with a rather slow velocity (∼ 10 m/s). In this regime the atoms inside the tube do not know that the valves open until the rarefaction wave reaches them. This means that within a few tens of milliseconds the Rb density uniformity remains at the required level over most of the 10 m length of the tube, except for a few tens of centimeters at each end. In these end regions the vapor density is very low and does not affect the ionization, acceleration or injection processes. The very uniform neutral density is turned into a correspondingly uniform plasma density by using a threshold ionization process for the first Rb electron, over the barrier ionization (OBI). Since Rb has a 4 low ionization potential (φRb = 4.177 eV), the intensity threshold for OBI (Iioniz ∝ φ ) is relatively low, ∼ 12 −2 15 −2 Iioniz = 1.7 × 10 W cm (when compared e.g. to ∼ 2 × 10 W cm for the first helium electron). A short (30 − 100 fs) Ti:Sapphire laser pulse (λ0 = 800 nm) with 20 − 40 mJ is sufficient to provide both 2 the energy necessary to ionize the atoms in the plasma volume (Eioniz ∼ n0φRbπRpL) and the intensity

14 15 x 10 2 20

) Density 522;4& 52224& -3 !"# Pressure $"#

*369*3!& 1 10 06178370!& 0!50'&

vapor density (cm 0'1& Rbvapor pressure (pa) Rb !"#$"%&'(& 2234&

0 0 )*+$,-./*& 120 140 160 180 200 220 Temperature (oC) 2:34& 22:4& Fig. 8: a) Rubidium vapor density (blue line) and pressure (green line) as a function of temperature. b) Schematic of the rubidium vapor source. The Rb vapor shown in green is produced by heating rubidium ◦ to temperatures TR between 150 and 200 C. The vapor fills a tube kept at slightly higher temperature than the vapor source to avoid condensation. The tube is heated by a hot synthetic oil bath. In this case the oil flowing from both ends to ensure best temperature uniformity has different temperatures (red TH > TL orange) to operate with two different vapor densities n0H < n0L. The hot oil is insulated using a surrounding vacuum tube (white color). The whole system is thermally insulated from room temperature (blue color). The vapor is contained by two valves, one at each end of the vapor column. The figure is not to scale, the hot region is ∼ 10 m long while the vapor tube radius is ∼ 2 cm. to ionize the last atom at the source exit (see the Technical Note [31] for more details.). The laser beam 2 is focused in the midpoint of the vapor source with a long Rayleigh length (ZR = πσ0/λ0 ∼ 5 m ∼ L/2 with σ0 ∼ 1 mm the beam waist size) to minimize the energy required in the pulse. As a result the last laser turning mirror that makes the laser beam co-linear with the proton beam, and therefore also the plasma, is located approximately 20 m upstream of the plasma cell midpoint. The laser intensity and fluence are below the mirror damage threshold. The laser beam angle and position pointing are stabilized by a technique [32] successfully used at the DUKE OK 4 free electron laser. The laser pulse is deflected from the proton beam path after the vapor source by a metallic foil. This is necessary for the optical diagnostics placed downstream of the source. The laser pulse is short compared to the plasma period (30−100 fs  4 ps) and travels co-linearly with the proton bunch (see Figs 5 and 6). The OBI process occurs on a much shorter time scale than the laser pulse length. With this method the very uniform plasma at the time of ionization is allowed to evolve by expansion, recombination, etc. only for the time interval between the laser pulse and the witness electron bunch. This is of the order of the proton bunch rms length, i.e. ∼ 400 ps. The plasma length can be varied by changing the location of the laser beam waist along the vapor source and by decreasing the laser energy. A laser system producing the 1 − 2 TW in 30 − 100 fs and operating at a 10 Hz rate can be purchased as an off the shelf item. These systems have a few % energy stability and are extremely reliable, especially when the oscillator is fiber-based (100 fs option). Simulations indicate that the sharp, relativistic ionization front co-moving within the proton bunch is an effective way of seeding the SMI (see the Technical Note [33] for more details). The laser ionized Rb vapor source satisfies all the criteria for SMI and proton-driven PWFA exper- iments over the 10 m plasma length scale. Based on our current work we see no major issue in building and operating it. An ∼ 3 m long prototype is currently under design and construction and will be tested this year.

15 The discharge and helicon plasma sources considered for later experiments over longer plasma lengths are described in the following sections. Note that a short 1 − 5 m plasma source based on Rb vapor may be used with these sources to allow seeding of the SMI, although schemes based on laser wakefield seeding of the SMI in an argon discharge are also possible (see the Technical Note [34] for more details).

Table 2: Metal vapor source plasma and ionizing laser parameters. Parameter Value 14 −3 Plasma Density ne 7 × 10 cm Plasma Density Range 1014 − 1015 cm−3 Plasma Density Uniformity δne/ne ∼ 0.2% (for 12 cm proton bunch) Plasma Wavelength λpe 1.2 mm Plasma Length L 10 m Plasma Length Range 1 − 10 m Plasma Radius Rp ∼ 1 mm Plasma Ions Species Rubidium (< 200 ◦C) Laser Type Fiber Ti:Sapphire Pulse Wavelength λ0 800 nm Pulse Length Range 30 − 100 fs Pulse Energy Range 20 − 40 mJ Rayleigh Length ZR ∼ 5 m Focused rms size σ0 ∼ 1 mm

5.2 Argon Discharge Plasma Source A uniform plasma column with electron density 1014 − 1015 cm−3, diameter 1 − 2 cm and length up to 10 m can be produced by a high-voltage discharge in a dielectric tube filled with argon. This gas can be ionized and heated to a temperature around 1 eV, corresponding to a population of ionized argon (Ar+) close to 100%, by a current density of 1 kA cm−2 with a duration of a few µs. The schematic of the argon discharge plasma source is presented in Fig. 9. A high voltage capacitor C is discharged through the dielectric tube ionizing and heating the plasma (main discharge). The striking of this discharge on a preformed partially ionized plasma produced by a glow discharge reduces the jitter of the main discharge and the voltage required for plasma ignition. The glow discharge is produced by the current leaking through the resistor R3 in parallel with the switch. The interface between the gas–plasma region and the accelerator vacuum pipes is made through the buffer volumes, valves and fast-shutters, shown in Fig. 9, avoiding the existence of windows in the path of the beams. Almost fully ionized uniform plasmas with lengths up to 3 m were obtained in experiments (see the Technical Note [34] for more details). These plasmas were obtained with a jitter corresponding to a fraction of the high-voltage pulse duration, and therefore can be synchronized with the particle bunches and the laser pulses used in the experiment. The development of a plasma source prototype is in progress where the gas containment and the discharge can be tested for the full 10 m plasma source required for the experiment.

5.3 Helicon Plasma Source Helicon waves have been demonstrated to very efficiently generate plasmas with a high degree of ioniza- tion. In contrast to standard plasma wave heating usually done through resonances, in helicon discharges the wave energy is collisionally dissipated [35]. The helicon dispersion relation

16 vacuum discharge plasma vacuum

e beam VC1 VC2 VC4 GV1 VC3 GV2 all CC AC beams

FS1 gas in FS2 HV PS R3 p beam R1 R2 C S

Fig. 9: Schematic of the plasma source based on a high-voltage discharge in argon. The plasma is produced by a high-voltage discharge between the electrodes mounted in the cathode chamber (CC) and the anode chamber (AC). The electric circuit contains a high voltage power supply (HV PS) that charges a capacitor bank (C) through a current limiting resistor (R1), the resistor R3 allows a DC 1 mA glow discharge. The main discharge is externally triggered through the high-voltage switch (S) and is limited in current by the resistor R2.

k kk B = eµ n (1) ω 0 e couples the wavenumber k, frequency ω and ambient magnetic field B to the electron plasma density ne. Here kk is the wavenumber component parallel to B. It has been experimentally verified that this basic scaling of plasma density with ambient magnetic field holds until the wave frequency equals the lower hybrid frequency ω = ωLH [36]. For a typical wave frequency of ω = O(10 MHz) this limits the ambient magnetic field to B ≤ O(10−1 T). However, a more detailed treatment including insulating 15 −3 wall boundary conditions yields maximum achievable plasma densities of ne ≤ 10 cm for ambi- ent magnetic fields of B ≈ 200 mT [37]. Although helicon discharges are routinely used to generate 14 −3 high density plasmas, the nominal plasma density of ne = 7 × 10 cm for AWAKE has not been demonstrated yet. A dedicated experimental program has been established to demonstrate helicon plas- mas providing the necessary plasma density with the required density homogeneity (see the Technical note [38] for more details). A prototype source of 1 m length is currently successfully operated at power densities of 20 MW/m3 using a single excitation antenna. The wave heating system will be extended to multi- antenna operation using four individual RF power-antenna systems with a total power density of 80 MW/m3 by the end of 2013. In independent studies of dual-antenna helicon discharge operation, efficient neutral gas pumping in the plasma center was observed to limit the discharge performance [39]. Thus, advanced gas injection schemes are currently under construction, which can also be used for profile shaping. The key objectives for the source development are:

14 −3 – achieve nominal AWAKE central plasma density ne = 7 × 10 cm in helicon discharges. – demonstrate high axial plasma homogeneity in multi-antenna operation. – validate power balance estimations, particularly radiated power vs. axial end losses to prove scal- ability of the concept.

A large helicon plasma cell that follows a strictly modular concept is depicted schematically in Fig. 10. The modularity applies to all individual parts of the cell: radio frequency (RF) systems, impedance matching devices, magnetic field systems, and vacuum tubes. All individual components

17 Fig. 10: Conceptual layout of a 10 m helicon plasma cell consisting of identical helicon heating systems each having a m = +1 mode helicon antenna, RF amplifier and impedance matching network. are commercially available. The discharge volume is a Quartz glass tube immersed in a set of conven- tional water-cooled magnetic field coils providing a homogeneous magnetic field of B ≤ 200 mT. The impedance matching units and RF power amplifiers operating at 13.56 MHz are standard systems. For the nominal plasma density a total RF heating peak power of PRF ≈ 500 kW is required for an argon plasma 10 m in length and 1 cm in diameter. The power is coupled to the plasma through in total ≈ 40, double helical, m = +1 mode helicon antennas.

5.4 Plasma Density Measurements As mentioned above, a plasma density uniformity better than 0.2% along the beam path is necessary for the resonant excitation of the wakefields and for the trapping and acceleration of the externally injected, low energy electrons. In the discharge and helicon sources the plasma density is evolving as a function of time over ∼ 100 µs. Fast and precise methods for single-shot plasma density measurements with a time window of around 100 ns are therefore under study (see the Technical Note [40] for more details). One method is the cut-off measurement using terahertz time-domain spectroscopy [41]. Here the cut-off of EM waves at the plasma electron frequency is measured. For the AWAKE experiment the plasma density should be tunable in the range 1014 − 1015 cm−3 resulting in plasma frequencies between about 90 and 300 GHz. For a sensitivity of better than 0.2% in plasma density, the plasma frequency spectrum (and hence cut-off frequency) must be measured with a resolution of about 100 MHz 14 −3 for ne = 10 cm . This plasma density diagnostic is currently being tested and will be implemented on the discharge and helicon sources. Initial measurements show good agreement with Langmuir probe density measurements. Alternatively, an optical emission spectroscopy (OES) method is also proposed as a simple, non- invasive and inexpensive technique for measuring the density of the low-temperature plasmas [42]. From the emission spectrum of the atomic species excited through recombination and electron impact pro- cesses, one can extract the electron temperature (Te) and the electron density (ne). The plasma radiation in the visible part of the spectrum is collected by multiple fibers at different locations along the plasma tube, and is sent to a spectrograph with a gated intensified CCD camera. This method is being used as a comparative measurement tool and can be integrated into all three AWAKE plasma sources currently under development.

18 6 Particle Beams Diagnostics Early experiments will focus on the interaction of proton bunches with long plasmas and on the SMI. Later experiments will focus on sampling of the wakefields by witness electrons and on the acceleration of electron bunches, i.e. on accelerator physics. Therefore diagnostics can naturally be split between those for proton and for electron bunches.

6.1 Proton Bunch Diagnostics for SMI As explained in Section 3, the development of the SMI results in the radial modulation of the proton bunch with a longitudinal period given by the plasma wave period (≈ λpe). This temporal modulation of the relativistic proton bunch can be measured using the radiation emitted by the bunch when traversing a dielectric interface or by directly sampling the bunch space charge field. Transition radiation (TR), the radiation emitted by charged particles when entering or exiting for example a metallic foil is suitable as a fast, picosecond scale diagnostic because it is prompt, unlike radiation from phosphor on scintillator screens. It can yield single event information about the bunch self-modulation structure and period.

6.1.1 Optical Transition Radiation Diagnostics Transition radiation is emitted over a very broad spectrum. It is incoherent at wavelengths much shorter than the bunch transverse and longitudinal structure: σz = 12 cm, λpe ∼ 1.2 mm (SMI), σr = 220 µm. In the visible range (400 ≤ λ ≤ 800 nm) optical transition radiation (OTR) is therefore incoherent. The temporal intensity of the OTR still carries information about the bunch longitudinal and transverse structure. An OTR screen similar to those used in previous SPS experiments [43] will be used (see the Technical Note [44] for more details). The backward OTR will be relay imaged onto a CCD for time integrated measurement of the bunch transverse size with and without plasma and onto the slit of a streak camera for time resolved measurements. A first sign of SMI occurrence will be the increase in time integrated transverse size of the bunch on a screen downstream of the plasma cell when the plasma is 14 −3 present. Since the plasma period is relatively long (∼ 4 ps for ne = 7 × 10 cm ) the bunch radius variation can be time resolved with a streak camera. Because of the large disparity between the bunch length (∼ 400 ps) and the plasma period (∼ 4 ps), the radial modulation will be resolved only within a narrow window 20 to 100 ps long. An image similar to that in the zoomed window of Fig. 11(c) can be expected in a narrow time window. With longer time windows the resolution may not be sufficient to determine the modulation period but may reveal the overall radius increase and the growth of the radius along the bunch (compare the bunch evolution in Figs 11(a) without and (b) with plasma). This measurement would also reveal the presence of the hose instability, should it occur (see Section 3). Streaking of the OTR light will therefore provide direct evidence of the occurrence of SMI. Note that if the radiation environment were too intense the light could be brought to a distant streak camera using optical fibers. A single fiber would reveal the time structure near the beam axis (r < 100 µm on Fig. 11), whereas multiple fibers could probe modulation at larger radii. Unlike coherent transition radiation (CTR), incoherent OTR does not carry direct spectral infor- mation about the radial and longitudinal bunch density structure. However, it was demonstrated with a free electron laser that the incoherent radiation shot noise resulting from the discrete nature of the par- ticle bunch carries statistical information about the structure [45]. This effect was measured with single electron bunches with approximately the same length as that of the micro-bunches expected in AWAKE, 1.5 and 4.5 ps. The bunch time structure reveals itself as spikes in the radiation spectrum. Statistically ∼ 2 the width of the spikes is ∆λspike = λ /λpe in the AWAKE case, where λ is the observation wavelength. This characteristic width is obtained either directly from individual event spectra, or from autocorrelation of many spectra acquired from multiple events. Note that while the bunch density modulation is radial

19 2 L=0 m (a)

1

r, mm 0 2 L=4 m (b)

1

r, mm 0 -30 -20 z, cm -10 0

(c) 1.0

r, mm

0.0 -10.7 z, cm -10.6 -10.5 Fig. 11: Simulation result showing (a) the incoming uniform bunch and (b) the self-modulated bunch after 4 m of plasma. (c) Zoomed region of the self-modulated proton bunch, as could be measured using the OTR–streak camera system. The z coordinate is converted to time by the streak camera. The period of the self-modulation is ∼ 1.2 mm or ∼ 4 ps. The r direction is along the camera slit. and not longitudinal, the radial information is indeed included in the tri-dimensional bunch form-factor. ∼ The OTR light will be imaged onto the slit of a ∆λres = 0.02 nm resolution spectrograph equipped ∼ with an intensified CCD camera. This resolution is sufficient to resolve the ∆λspike = 0.9 nm expected ∼ around λ = 745 nm with the self-modulated bunch, but not sufficient to resolve the ∆λspike = 0.002 nm expected with the long incoming proton bunch. Again, the dependency of ∆λspike versus ne will be studied systematically in the experiments (see the Technical Note [46] for more details).

6.2 Coherent Transition Radiation Diagnostics Since TR at wavelengths longer than the characteristic transverse and longitudinal bunch structure is emitted coherently, its intensity is larger than that of OTR by a factor of the order of the number of particles in the bunch. When the proton bunch is self-modulated, radiation around the plasma wavelength is coherently emitted in the 90 to 300 GHz microwave frequency range. Assuming only an effective aperture size of about 2 mm2 of a receiving antenna (e.g. a microwave horn) this radiation has power levels of several Watts at the antenna output. It can be directly detected using a bandpass filter and a following microwave video detector (see Technical Note [47] for more details). Direct detection of this signal would again be evidence of the SMI occurrence. Alternatively the microwave signal could be mixed with a local oscillator signal in a frequency down converter and the intermediate frequency signal detected after filtering and amplification with a fast oscilloscope or spectrum analyzer. This method is more sensitive than the direct detection method and yields direct information about the modulation frequency without any change of the detection filter bandwidth. This is the method commonly used in applications for synthetic aperture radars and radio astronomy. These two CTR measurements will be implemented at the beginning of the experiment.

20 Fig. 12: Comparative measurement of a phase modulated signal using a conventional grating-based optical spectrum analyzer (red) and the DFT method (black).

6.3 Electro-Optical Sampling The bunch space charge field or its corresponding (transverse) CTR field can be used (see the Technical Note [48] for more details) for electro-optical (EO) measurements or sampling (EOS). In EO measure- ments the time and frequency structure of the field is imprinted onto a laser pulse using a crystal with an induced bi-refringence through phase modulation of the incoming laser signal (Pockels or EO effect). The low power laser pulse is produced by a solid-state, 1.5 µm laser fitting in the slot of an instrument rack and transported using optical fibers. This imprint can then be read out with different methods. The dispersive Fourier transformation (DFT) method will be used for single-shot measurements of the bunch modulation in the frequency domain. We use an optical pulse of length ∼ 100 ps, shorter than the temporal proton bunch length. Conversely the Fourier-limited spectral width (≈ 3.5 GHz) is sufficiently narrow for an acceptable frequency resolution. The modulated pulse is transformed into the time domain by means of a highly dispersive optical fiber. The relative position in time of each frequency component is shifted by D · ∆λ with D the dispersion coefficient of the fiber and ∆λ the wavelength difference between the frequency component and the optical pulse. The output signal of the fiber is a convolution of the (dispersion broadened) sampling pulse with the time-scaled shape of the modulation frequency spectrum. It can be detected using a fast photodetector and displayed with a real-time sampling oscilloscope (bandwidth ≤ 8 GHz). This method has been recently tested in the laboratory and yielded very good results (see Fig. 12). When using the DFT setup the phase information of the original modulation signal is lost. For electron beam injection experiments the relative timing of the bunch modulation compared to the ioniza- tion laser is important to optimize the injection phase. To retrieve the phase information an extension of the DFT setup is proposed. The picosecond optical source will be replaced by a femtosecond laser and a dispersive fiber to pre-chirp the laser pulse. The laser pulse will be derived from the ionizing laser oscil- lator. Using the EO modulation scheme above, each frequency component of the dispersion-lengthened optical pulse is modulated by a different time slice of the bunch signal. Again a frequency-to-time conversion is performed by means of a dispersive fiber and detected by a photodetector/oscilloscope

21 combination. Contrary to the DFT measurement, the recorded signal now represents a time stretched image of the modulation signal. The stretching factor is given by M = 1 + D2/D1 where D1 and D2 are the total dispersion of the first and second dispersive fiber, respectively. Assuming M = 100 and an oscilloscope with a bandwidth of 10 GHz one can in principle achieve a sampled bandwidth of 1 THz, or a resolution better than 100 fs. In practice the limit will be in the range of ∼ 400 − 500 GHz. This technique has been successfully demonstrated in [49]. With an operating wavelength of 1.55 µm using conventional cheap and easy-to-handle telecom components both setups are particularly suited for integration in the AWAKE experiment. Making use of optical fibers means that the radiation sensitive electronics parts are placed away from the beam line area. Both EO measurements will be implemented in a later phase of the experiment.

6.4 Electron Spectrometer A state-of-the-art electron spectrometer with a very large energy acceptance and the best possible energy resolution will be developed and installed downstream of the plasma cell to measure the properties of the accelerated electron beam. We have designed (see the Technical Note [50] for more details) a spec- trometer system that will allow us to measure an electron beam exiting the plasma in the energy range of 10–20 MeV (the injection energy) to 5 GeV, depending on whether the plasma is on/off, whether proton bunches are compressed, and depending on plasma parameters. The basic layout of the spectrometer is to have a magnet system placed after the exit of the plasma in which a dipole magnet bends the accelerated electrons onto a screen which is imaged by a CCD camera. The screen is mounted at 45◦ to the proton beam axis and the exit face of the dipole magnet. The CCD camera is mounted at 45◦ to the screen and therefore 90◦ to the beam axis. A LANEX screen is suitable for the needs of the experiment as it is relatively radiation hard, has high light output and is cheap and easy to produce in virtually any size. The camera, already purchased, is an intensified CCD camera iSTAR 340 T model with 1850 × 512 pixels. In order to achieve a reasonable deflection for 1 GeV electrons onto a scintillating screen viewed by a CCD camera, typically a 1 m long dipole with a field of 1 T is required. The aperture must be more than 5 cm vertically and 20 cm horizontally. According to beam–plasma interaction simulations some protons are defocused during the beam self-modulation process in the beginning of the plasma section and can obtain a transverse kick of up to 1 mrad. This leads to ±1 cm of the transverse proton position inside the electron spectrometer which consequently sets the lower limit for the vertical aperture of the electron spectrometer. In order to achieve good energy resolution of the high energy electron beam, additional focusing of the beam exiting the plasma cell is required. Two quadrupoles with an integrated magnetic field gradient of around 7 T/m and an aperture larger than 4 cm are installed just before the spectrometer. This doublet provides focusing in both planes allowing both improvement in the energy resolution (with horizontal focusing) and reduction in the vertical beam size at the scintillator screen. The quadrupoles and the spectrometer dipole exist at CERN and could be used for AWAKE. The CERN dipole magnet was simulated with its maximum field of 1.84 T and the screen posi- tioned with an edge almost touching the dipole magnet. Figure 7 shows the electron energy spectrum obtained from numerical simulations of the baseline design (see Table 1). The particles from the plasma simulation were tracked from the exit of the plasma, through the dipole magnet and to the screen using the particle tracker code GPT [51] and the resulting spatial and intensity profiles expected on the screen are shown in Fig. 13(a). From the spread in position, and the correlation of the position with energy (Fig. 13(b)), an energy profile can be extracted. In Fig. 13(c), the reconstructed spectrum is compared to the original energy spectrum of electrons exiting the plasma. The screen and CCD camera efficiencies have not been accounted for but they will only affect the overall normalization of the spectrum and not the energy resolution. The measured energy spectrum also demonstrates that the emittance of the ac-

22 (a)

270

260 Actual y pixels 250 600 Measured 240 1150 1200 1250 1300 x pixels 500

400 (b) (c) 300

Electrons per pixel 200

100

0 1.7 1.8 1.9 2 2.1 Electron Energy/GeV

Fig. 13: (a) Simulation of impact position of electrons on the scintillator screen, as will be seen by the CCD camera, having passed through the magnetic spectrometer. (b) Correlation of electron energy with screen position compared to expectations (blue line). (c) The electron energy spectrum reconstructed using the energy spectrometer (blue line) compared with the energy distribution exiting the plasma cell (red line), i.e. that shown in Fig. 7. celerated electron bunch has little effect on the resolution of the measurement. Consequently, it can be seen that the energy reconstructed with the spectrometer system agrees well with the input spectrum and demonstrates the suitability of this set-up.

7 Electron Source The electron injector system (see the Technical Note [52] for more details) is a critical component of the AWAKE experiment that will allow the acceleration of an externally injected witness beam of electrons to be demonstrated. The photo-injector is placed in an area adjacent to the experimental area and the 10–20 MeV electron beam is transported along a ∼ 10–15 m long beam line (see Section 8.3.2) before being injected into the front-face of the plasma cell. A preliminary design of the first section of the electron bunch injection system – the gun and booster – is shown in Fig. 14. This injector is based on the ‘ALPHA-X’ 2.5-cell S-band RF gun currently being commissioned at the Versatile Electron Linear Accelerator (VELA) at STFC Daresbury Laboratory, UK [53], a design which has also been used for some years in the PHIL accelerator test facility at LAL, Orsay, France [54], where it has been shown to deliver high quality electron beams. The specified design is composed of a metal photocathode housed in a 2.5-cell normal conducting RF cavity. The use of a metal photocathode (in this case copper) should give reliable operation with the minimum amount of cathode preparation required. The low quantum efficiency of this material is compensated for by the high intensity of the UV laser pulse used for photo-emission. The use of an RF cavity for initial acceleration provides a high electric field, thus minimizing space charge effects. The electron beam from the gun will have an energy of 5–6 MeV, limited by the point at which significant dark current starts to be produced. However, since the optimum energy for electron capture within the plasma wakefield is likely to be higher, a booster linac section will also be required. A 1 m RF section (either traveling or standing wave) will allow energies of 10–20 MeV to be achieved. In addition, by running the booster off-crest, it would be possible to introduce a correlated energy chirp that could be used with a

23

Fig. 14: Cut-away view of the RF photo-injector gun that will produce the electron bunch to be injected in the plasma wakefields. magnetic chicane to carry out bunch compression. Such an arrangement could allow significantly higher beam currents to be achieved for the shorter pulse lengths envisaged for later stages of this experiment. The electron photo-injector gun will produce both long, low current bunches for initial side injec- tion experiments and short bunches for on-axis injection. Note that the electron beam will be produced on the gun photo-cathode using a laser pulse derived from the low power level of the plasma source ionizing laser system. This will ensure that the electron bunch will be injected at a fixed distance (see Section 8.4.2) behind, i.e. phase-locked with, the laser pulse that seeds the self-modulation of the proton bunch. This distance (or phase) will be chosen to be the optimum for acceleration along the plasma. The expected electron bunch parameters are given in Table 3.

8 The AWAKE Facility at CERN 8.1 Introduction In the 2011 Letter of Intent [9], the CERN West Area was proposed for the AWAKE experimental facility. Meanwhile, the CERN Neutrino to Gran Sasso (CNGS) program has been completed, and it has been decided that the CNGS beam will not re-start in 2014 after the CERN Long Shutdown 1 [55]. Therefore, the CNGS facility has also been considered as location for the installation of the AWAKE experiment.

8.1.1 CERN Site Selection A detailed comparison of the sites for the AWAKE facility has been performed, covering studies on the design of the primary beam lines, the experimental area, civil engineering, general services and

24 Table 3: Design parameters for the electron bunch to be injected in the plasma wakefields.

Parameter Nominal value Beam Energy 10 − 20 MeV Energy Spread (rms) < 1% Bunch Length 0.3 − 10 ps Laser / RF Synchronization 0.1 ps Synchronization to Experiment 0.1 ps Free Repetition Rate 10 Hz Synchronized Repetition Rate 0.03 Hz Focused Transverse Size < 250 µm Angular Divergence < 3 mrad Normalized Emittance 0.5 mm mrad Bunch Charge 1 − 1000 pC

infrastructures for the facility as well as radiation protection and general safety aspects. As a result of these design studies, the cost, manpower and schedule details of the two alternative sites have been worked out and are summarized in Table 4 (see the Technical Note [56] for more details). Unlike CNGS, the West Area is a surface facility, posing strong radiation constraints and consequently the need for civil engineering and heavy shielding. This imposes strict geometric limits on the design of the primary proton beam line and limits the maximum energy to 300 GeV. In addition, in the West Area the proton beam line as well as general service systems such as ventilation, cooling, electrical network systems need to be fully refurbished. Important changes in the logic of the beam interlock system, also used for the LHC operation, would need to be applied.

Table 4: Cost and manpower needs and earliest beam start for the AWAKE experiment in the CNGS facility and in the West Area. The cost includes material, industrial services and fellow/project associates. AWAKE installed Cost PY (person years) Earliest start of experiment at CNGS 8.5 MCHF 34.2 end 2016 at West Area 37.7 MCHF 87.1 end 2017

In summary housing the AWAKE experiment in the CNGS facility is nearly a factor five less costly than in the West Area and first beam to the experiment could be sent end 2016. Consequently the CNGS facility is the best suited location for the AWAKE experiment.

8.1.2 Overview of the CNGS Facility The CNGS facility [57] is a deep-underground area and is designed for running an experiment with high proton beam energy, such as AWAKE, without any significant radiation issue. The facility is fully opera- tional, with a 750 m long proton beam line designed for a fast extracted beam at 400 GeV as also needed by AWAKE. Installing the AWAKE experiment upstream of the CNGS target (see Fig. 15) is possible with only minor modifications to the end of the proton beam line; these include changes to the final fo- cusing system and the integration of the laser and electron beam with the proton beam. General services such as cooling, ventilation, electricity, radiation monitoring and access system exist, are operational and need only some modifications to be adapted to the AWAKE experimental setup. Some civil engineering modifications are necessary to be able to combine the electron beam and the laser pulse with the protons in the plasma cell. The AWAKE facility will be separated from the radioactive area located downstream

25 of the CNGS target area by a shielding wall.

AWAKE experiment

Fig. 15: The AWAKE experiment in the CNGS facility.

8.2 Experimental Area 8.2.1 Layout The AWAKE experiment will be installed in the CNGS facility as shown in Fig. 16. The plasma cell is housed in the downstream end of the CNGS proton beam tunnel. The beam diagnostics for the outgoing protons as well as the electron spectrometer system are installed in the area upstream the CNGS target. The CNGS storage gallery will be modified to become a dust-free (consequently over-pressurized) and temperature regulated area to house the laser system. The electron source, including the klystron is installed in the ventilation chamber. A major part of the electronic racks is installed in the service gallery parallel to the target area. For the AWAKE facility radiation protection studies on material, cooling and soil activation, prompt radiation and airborne activity were performed and showed no or acceptable radiological con- straints. In addition the general safety requirements of the experimental set-up are considered in the design of the facility. The AWAKE area is separated from the downstream part of the CNGS target area by an 80 cm thick concrete chicane shielding and dust-proof separation doors. This keeps radiation dose in the AWAKE experimental area minimal. The CNGS target area is also separately ventilated in order to avoid corrosion of the CNGS equipment inside (target, horns). The high power laser beam used for plasma ionization and bunch-modulation seeding is trans- ported through a new dedicated tunnel (0.5 m diameter, 4 m length) connecting the laser area to the

26 proton beam tunnel. For distortion-free transport through lenses and vacuum viewports, the laser pulse duration must be above 100 ps. The compression of the pulses will be performed in a vacuum cham- ber coupled to the proton beam line, located at the laser/proton beam junction. The laser beam for the electron injector (red line in Fig. 16) is taken from the uncompressed part of the pulse for the plasma production. All laser beams will be transported within a metal enclosure, with the long straight parts under a few mbar vacuum in order to avoid beam perturbation by air flow. The klystrons powering the electron source are housed in the ventilation chamber, which is a low-radiation area and free of electromagnetic interference (necessary for the klystron). The electrons are transported from the electron source system to the proton beam tunnel along the electron beam line through a new liaison tunnel (7 m long, 1 m wide and 2.5 m high). In the early stage of the experiment a 10 m long plasma cell is installed in the downstream end of the proton tunnel. This area can be modified to house a shorter vapor cell and a helicon and/or discharge source. In order to avoid accidental venting and possible contamination from the plasma vapor to the proton beam line vacuum, a double window system for the proton beam is integrated in the new design, 46 m upstream of the plasma cell. The area downstream of the plasma cell comprises the proton diagnostics system as well as the electron spectrometer. Racks needed for the diagnostics and the plasma cell are also foreseen in this area. The electron beam dump is located immediately after the electron spectrometer. Energy deposition estimates lead to a beam dump design with a 30 cm thick block of iron surrounded by 30 cm thick concrete shielding. Downstream of the diagnostics instrumentation the proton beam vacuum tube goes through the shielding separating the AWAKE area from the CNGS target area. The proton beam exits through a vacuum window and passes the 100 m long target chamber and the 1000 m long decay tunnel before being dumped in the existing CNGS beam dump, a 15 m long carbon–iron block equipped with a cooling system. Considering the proton beam divergence along the proton path between the plasma cell and the beam dump, calculations have shown that there is no significant energy deposition from protons in the existing equipment, CNGS target and horns, which can therefore stay at the present location. Only the collimator in front of the CNGS target has to be replaced with a collimator of larger diameter to let the proton beam pass through.

8.2.2 Access to the Experimental Area Additional local access doors for the laser room, the electron source and the upper floor above the electron source are required to allow stand-alone operation for the different systems. In order to give access to the ventilation chamber and the klystron system during the tests of the electron source in stand-alone mode, a 30 cm thick concrete shielding wall is built around the electron source. Access to the first floor of the ventilation chamber and to the plasma cell area is prohibited when the electron source is in operation, and the relevant access doors are interlocked to the powering system of the electron source. Special conditions are required for the doors to the laser area: access to this area must be allowed, but for ‘laser specialists only’ when the laser beam is ON. The access system to the AWAKE area during proton beam mode is used in the same way as for CNGS: when the proton beam is switched on, access to the experimental area is prohibited and the corresponding access door is ∼ 900 m upstream in ECA4, the area where the elevator to the surface and material shaft are located. In this area also a control room for the experiment is foreseen.

27 Laser power supplies Access gallery Items in dark blue: ventilation ducts Items in light blue: AWAKE electronic racks Items in cyan: existing CNGS equipment (cable trays, pipes,…) Lasers

Electron gun

Klystron system

2m

2m

Proton beam line 10m

28 Laser & proton beam junction

Electron beam line

Plasma cell (10m long)

Electron spectrometer

Experimental Diagnostics

CNGS target area

Fig. 16: Integration of the AWAKE experimental components in the experimental area. 8.3 Proton and Electron Beam Lines 8.3.1 Proton Beam Line Design and Optics Studies The CNGS beam line was designed for a 400 GeV proton beam. The same energy will be used for the AWAKE project so that only minor changes will be needed in the matching section and in the final focusing part of the primary beam line (see Fig. 17).

Fig. 17: Comparison between the present CNGS and the future AWAKE layout in the final part of the primary proton beam line.

Two main quadrupoles (MQ) will be removed and the remaining seven magnets of the triplet will be redistributed and shifted by a few meters upstream to fit the plasma cell at the end of the CNGS tunnel. A round beam of 200 µm (1 σ), with a 10% tolerance, is required at the entrance of the plasma cell. This ∗ corresponds to a β of 4.9 m (for a normalized emittance εN = 3.5 mm mrad) and a dispersion close to zero in both planes (see Fig. 18).

Fig. 18: The horizontal and vertical β functions (left) and dispersion (right) for the final part of the AWAKE beam line.

The modified final focusing satisfies the AWAKE requirements on β∗ and a dispersion of 0.029 m could be obtained in both planes corresponding to a beam size of 224 µm (1 σ). A laser beam has to be merged with the proton beam at a maximum distance of 20 m from the plasma cell. A chicane in the proton beam line has been designed to integrate the laser by displacing the

29 last main dipole (MB) towards the experimental area and by installing four additional 1.9 m long dipoles (0.7 T) of type B190 (see Fig. 17 and Fig. 19), giving a kick of 1 mrad each.

Fig. 19: The new beam line with the chicane for the laser integration (blue line) in comparison to the existing CNGS beam line (red line). The tunnel walls are also shown.

In the proposed design the mirror, which is used to inject the laser beam into the merging dipole, is located 19.5 m upstream of the plasma cell and at a distance of 12 m from the center of the last B190 dipole. An offset of 24 mm exists between the proton and the laser beam axis. The clearance needed to fit the mirror in the primary beam line without intercepting protons (6 σ beam envelope) and inducing losses is 18.4 mm (45◦ mirror angle), well within the obtained offset at the mirror. The proton and the laser beam have to be coaxial over the full length of the plasma cell; in par- ticular, the 3 σ proton beam envelope (∼ 600 µm) must be contained in the 1 σ laser spot size (∼1 mm). The pointing precision of the laser at the entrance of the plasma cell corresponds to 100 µm; a pointing accuracy of 100 µm and 15 µrad has to be guaranteed for the proton beam. Assuming that any systematic error (i.e. magnet misalignment) could be compensated by the trajectory correctors, the ripples in the main dipole power converter current have to be kept below ∼ 5 × 10−4. A maximum ripple of the order of 1×10−4, in agreement with the AWAKE requirements, was measured for the CNGS main dipoles, quadrupoles and correctors. A detailed analysis of the contribution of random magnetic errors will be performed. The same converters, which were used for the CNGS line, can also be used for AWAKE. The B190 type magnets will be powered, in pairs, by the converter of one of the removed quadrupoles and by an existing spare.

8.3.1.1 Diagnostics The existing CNGS beam instrumentation [58] can be used for the diagnostics of the AWAKE beam with suitable modifications due to the different beam intensity and bunch structure (only one bunch of 3×1011 protons for nominal operation and one pilot bunch of 5×109 protons for beam setup). In particular the electronics of the beam position monitors (BPMs) has to be replaced by a system similar to the LHC transfer lines; this demands re-cabling works, probably with fibres, and the installation of mini-racks in the tunnel. The BPM accuracy along the line must be of the order of 200 µm with a resolution of 100 µm within a radius of 20 mm. Two more precise BPMs (50 µm accuracy) have to be installed, one upstream and one downstream of the plasma cell, to check the pointing precision of the proton beam with respect to the laser during operation. An interlock will be implemented to stop extraction from the SPS in case of a drift of the proton trajectory out of the experiment tolerances. The existing OTR

30 screens (called BTVs) along the proton beam line will be used for profile and emittance measurements. Two additional OTRs will be put around the plasma cell, one upstream and one downstream, for proton beam setup and to measure shot-to-shot variations of the transverse parameters. An interlock preventing the high power laser from pulsing when the OTR screens are in the beam has to be put into operation. Special attention has to be given to maintaining the integrated radiation dose below 400 Gy to avoid damaging the CCD cameras. The CNGS beam current transformers (BCTs) could satisfy the AWAKE requirements (0.5 − 1% accuracy for the nominal beam and 10 − 20% for pilot bunch) for beam intensity measurements, provided some optimization on cable length and signal filtering is performed. Existing beam loss monitors (BLMs) should fulfill the requirements in terms of resolution and interlocking needs.

8.3.2 Electron Beam Line Design and Optics Studies The electron beam will be produced by a 5 MeV RF gun, accelerated to 10 − 20 MeV and transported to the plasma cell through a beam line of about 12 m. The first 1.5 m will be dedicated to the acceleration system (booster linac) needed to reach the target energy. The lattice of the electron beam line is formed by five dipoles and nine quadrupoles. The first two vertical bends form an achromatic dog-leg to overcome a 1 m vertical step. A horizontal achromat is used to bend the electron beam by about 60◦, parallel to the proton beam. A last vertical bend brings the electrons orthogonal to the plasma cell entrance.

The AWAKE baseline design requires σx,y < 250 µm at the plasma cell (see Table 3 for the design electron bunch parameters). Due to the geometry of the line, only the horizontal dispersion could be close ∆p −3 ∼ to zero. Assuming a p = 10 and a normalized emittance εN = 0.5 mm mrad, σx,y = 100 µm can be obtained for a 20 MeV beam, with the lattice shown in Fig. 20. The final beam size is completely tunable using the last 3 quadrupoles, so as to guarantee enough freedom to the experiment without affecting the whole line. The beam size at the entrance of the plasma cell can be kept below 250 µm even at 5 MeV, using the same quadrupole strengths, but the beam would not be round anymore due to the difference between Dx and Dy at the final drift. The maximum βx and βy in the line are 63 m and 90 m respectively. This design allows side-injection and low charge on-axis injection of the electrons into the plasma. For a later phase of the experimental program with compressed electron bunches a re-design of the electron beam line is necessary.

Fig. 20: Lattice functions along the electron beam line.

31 8.3.2.1 Magnetic Elements The magnetic elements used in this beam line design are based on an existing design in use at the FERMI@Elettra facility [59]. The magnetic parameters are given in Table 5.

Table 5: Magnetic element parameters [59]. Parameters Quadrupoles Parameters Dipoles Yoke length [m] 0.170 Yoke length [m] 0.340 Bore diameter [m] 0.08 Gap [m] 0.032 Max current [A] 100 Max current [A] 330 Turns per pole 45 Max R B.dl [T m] 0.2233 Max Integrated grad [T/m] 1.519

For the specified energy range, just 17% and 60% of the available strength will be needed for the dipoles and quadrupoles respectively. The dynamic range that the magnets have to ensure is given in Table 6. A smaller and weaker dipole can be used for the last vertical deflection in order to reduce the magnetic field range, if required.

Table 6: Dynamic range of magnetic elements of the electron beam line. Dipoles 5 MeV 20 MeV Quadrupoles 5 MeV 20 MeV R B.dlmin [T m] 0.00095 0.0038 gradmin [T/m] 0.066 0.27 R B.dlmax [T m] 0.0095 0.038 gradmax[T/m] 0.23 0.91

The electron beam will reach the plasma cell with an offset of 2 cm with respect to the proton beam. Two bending magnets have to be installed around the plasma cell: one to merge the two beams between 3 m and 5 m downstream of the beginning of the plasma cell. The second magnet is used to bend the non- captured electrons for diagnostic purposes and also the electron beam when the plasma is not present (see Fig. 21). The deflection angle needed to merge the two beams can be up to φ = 20 mrad (as requested from the experiments) with a merging point between 2 and 5 m into the plasma cell. The required magnetic and powering configuration has to be investigated in more detail when the specifications are more precise, especially if these parameters need to be varied for different measurements.

Table 7: Proposed dipole design parameters. *Assuming 1 turn per pole. R Length [m] I [A] Max Gap [m] B.dlmax [G m] 1 317.5* 0.3 13.3

8.3.2.2 Diagnostics Standard beam diagnostics at the exit of the RF gun will measure the emittance (pepper-pot technique) and a maximum of 11 monitors (one at each quadrupole plus two just before the plasma cell) will measure the beam position and profile (e.g. YAG screens).

8.4 The SPS Proton Beam 8.4.1 SPS Beam Characteristics The AWAKE experiment requires short, high intensity, small transverse emittance, single proton bunches extracted from the SPS at 400 GeV. The minimum bunch length at extraction from the SPS is determined by the maximum available RF voltage and the minimum achievable longitudinal emittance. The longitu- dinal instability developing during the ramp is the main reason for the longitudinal emittance blow-up in

32 Fig. 21: Proposed configuration for the two dipoles needed around the plasma cell. the SPS. This limitation can be pushed up by reducing the transition energy of the SPS ring, as recently done with a new optics used in the SPS. However, for the same RF bucket size the new optics requires higher RF voltage. There are two RF systems in the SPS with different maximum available voltages: 8 MV at 200 MHz and 600 kV at 800 MHz. The measurements on the flat top presented below were done with the maximum voltage at 200 MHz. The 800 MHz RF system was used during acceleration for beam stabilization in the longitudinal plane. Two methods of bunch shortening can be currently applied for the AWAKE beam: adiabatic volt- age increase to the maximum amplitude or bunch rotation in the longitudinal phase space during one quarter of the synchrotron period. For the AWAKE experiment a fast acceleration cycle (7.2 s) with an acceleration time of 3 s can be used. On flat top (at the desired extraction energy) a few ms are sufficient for bunch shortening before extraction. More time is required for beam synchronization (up to 300 ms), see Section 8.4.2.

8.4.1.1 Bunch Parameters A few machine development sessions were carried out in the SPS in 2012 to study achievable beam parameters for AWAKE and their reproducibility in the two different SPS optics, with nominal and lower transition energy [60]. The best results were obtained so far with the new optics. Nevertheless bunches were unstable for intensities above 3 × 1011 protons/bunch. The bunch length grows with intensity, see Fig. 22 (left). The minimum bunch length obtained after the bunch rotation was ∼ 20% less than using an adiabatic voltage increase. Bunches injected into the SPS had a constant bunch length (longitudinal emittance of 0.35 − 0.4 eV s), but the final bunch length increases with intensity, so that the bunches with intensities of 3.5 × 1011 had an average bunch length of (1.4 ± 0.1) ns while for 2 × 1011 protons/bunch the bunch length (4 σ Gaussian fit) could be as short as 1.1 ns. In the same intensity range, (2 − 3.5) × 1011, the transverse normalized (rms) emittances on the flat top increases from 1.1 to 2.8 µm with non-negligible cycle-to-cycle variation, Fig. 22 (right). The maximum peak current obtained in the SPS for stable bunches is ∼ 59 ± 4 A and for highest bunch intensities it is 67 ± 7 A. The instability threshold decreases with beam energy. In case of a beam extraction at lower energy (e.g. 300 GeV) the bunches will be more stable, but the bucket area for the same RF voltage will be smaller, so that similar bunch lengths are expected at 300 GeV. Bunches can be stabilized by a controlled longitudinal emittance blow-up, routinely applied during the ramp to the LHC beams. This will improve the reproducibility but will lead to an increase of the minimum bunch length.

33 Fig. 22: Bunch length (Gaussian fit, 4 σ) after rotation (left) and normalised rms transverse emittances (right) on the SPS flat top (450 GeV) as functions of bunch intensity in the SPS Q20 optics [60].

8.4.1.2 SPS Ring Diagnostics Longitudinal profiles (bunch length) can be measured in the SPS every cycle after the bunch rotation before the extraction. The acquisition of the bunch profile can be done during each revolution turn with 8−20 GHz sampling frequency. The bunch intensity can be measured during the whole cycle each 10 ms by a DC current transformer (BCT signal). Transverse emittances in each plane can be measured using Wire Scanners on the flat top once per cycle.

8.4.1.3 Further Improvements Bunch shortening by bunch rotation can be further optimized by using a 180◦ jump to the unstable synchronous phase, an intermediate step during bunch rotation. This RF manipulation requires some modification/improvement in the RF beam control which will be done during the Long Shutdown 1, and will be available for tests in 2014. Further bunch length reduction can be expected in the future due to the planned RF voltage increase: the voltage in the 800 MHz RF system will be increased to 1.2 MV after the Long Shutdown 1, and in the 200 MHz system the voltage will be increased to 12 MV after the Long Shutdown 2. Finally, beam dynamics simulations in a double RF system using the realistic SPS impedance model are planned to study further possible improvements.

8.4.2 Synchronization of the AWAKE Experiment with the SPS Proton Beam 8.4.2.1 Synchronization Requirements The AWAKE experiment relies critically on the alignment in space and on the relative timing of the two particle bunches and of the laser pulse. The laser pulse driving the RF photo-injector gun will be derived from the same oscillator as that for the ionizing laser pulse. Providing the RF in the gun and accelerating section of the electron beam line is synchronized with the laser pulse. The ionizing laser pulse and the electron bunch can be timed with respect to each other. This is a standard feature of RF photo-injector guns. Low level RF for the gun klystron(s) at 2.9985 GHz is used to generate a phase- locked sub-harmonic signal. This signal, typically in the 70 MHz range, is then used to mode-lock the laser oscillator through fast cavity length adjustments. Synchronization between the laser pulse and the electron bunch at the 10 fs scale has been achieved. Synchronization at the few hundred femtosecond level, a fraction of the plasma period (∼ 4 ps), will be required for deterministic injection of the witness electron bunch into the plasma wakefields. The timing of the laser with respect to the proton beam should be stable to better than one longi-

34 tudinal rms length of the proton bunch (300 to 400 ps), and thus a synchronization of the order of 100 ps is desirable between the laser and the SPS RF system.

8.4.2.2 Frequency Reference and RF Rephasing Since the laser requires a stable mode-locker frequency reference, it cannot follow the changes in SPS RF frequency during acceleration. Therefore it is the SPS RF which must rephase and lock to the stable frequency reference on the flat top before extraction, using a procedure similar to that employed for beam synchronization towards the LHC. The reference frequency needs to be transmitted on optical fibre from AWAKE to the RF Faraday cage in BA3. The laser mode-locker frequency, which is of the order of 70 MHz and depends on the optical construction of the laser, and the target SPS RF frequency at extraction (200.394 ± 0.001 MHz), must be a harmonic of a common frequency. The limits on the target RF frequency are given at 400 GeV for a maximum allowable radial beam displacement of 5 mm. In addition to the RF frequency reference, the rephasing process requires a pulse train at the SPS revolution frequency and a pulse train at the laser system repetition rate (about 10 Hz) to be transmitted to the BA3 Faraday cage. A warning pulse also needs to be transmitted from the SPS RF to AWAKE to indicate that beam extraction takes place on the next laser pulse. All of these requirements are found to be achievable with present technology.

9 Project Planning 9.1 Timeline Assuming the approval of AWAKE in mid 2013, the first proton beam to the plasma cell can be sent end of 2016 (see Fig. 23). Considering four years for the completion of the electron source system, the electron beam will be operational end of 2017. Operation of the AWAKE facility is envisaged for a period of 3–4 years. Beam is requested for approximately 4 periods of two weeks per year; the time in between beam is required for installation, commissioning and maintenance of the experimental set- up and for a thorough analysis of the data acquired and planning of the next round of experimentation. Further experimental efforts will be evaluated based on the results of the initial running. It is worth noting that civil engineering and installation work for the AWAKE experiment can be done during SPS and LHC operation and is not linked to the machine shutdown schedule.

Fig. 23: Timeline for the construction and commissioning of the AWAKE experiment in the CNGS fa- cility. The project approval is assumed in June 2013. Operation and data taking is planned for 3–4 years.

9.2 AWAKE Physics Program We plan to carry out the following research program:

35 1. Perform benchmark experiments using proton bunches to drive wakefields for the first time ever. To carry out this task, we will: – Develop a highly uniform metal vapor plasma cell of 10 m length. – Develop the requisite diagnostics to measure proton bunch modulation and the production of strong wakefields. – Carry out a detailed program of measurements in 4 periods of two weeks per year. 2. Understand the physics of the self-modulation process in plasmas. For this, we will: – Perform detailed simulations of the proton–plasma interaction, in particular for the given experimental conditions. – Perform detailed comparisons of experimental data with simulations and extract the key physical insights. 3. Demonstrate high gradient acceleration of a bunch of electrons in the wake of a proton bunch, with a narrow energy spread. This implies that we: – Design and construct a new electron source capable of producing both long bunches to study the side-injection of electrons into the proton wake, and short bunches for on-axis injection at a well defined phase. – Design and construct an electron spectrometer to analyze the energy spectrum of accelerated bunches of electrons. 4. Develop long, scalable and uniform plasma cells for future experiments and accelerators: – Develop helicon plasma cells that are scalable in length and reach the necessary plasma density uniformity, and test these in the AWAKE experiment and demonstrate their feasibility for PWFA. – Develop discharge plasma cells that are scalable in length and reach the necessary plasma density uniformity, and test the discharge plasma cells in the AWAKE experiment and demon- strate their feasibility for PWFA. – Operate in split-cell mode (modulation, acceleration) and demonstrate high-gradient, long- distance acceleration (this requires a density step). 5. Develop a scheme for the production and acceleration of short bunches of protons: – Perform detailed studies of proton bunch compression schemes in the accelerating structures available now at CERN. – Perform an experimental program with compressed proton bunches in AWAKE. – Study schemes for the production and acceleration of short bunch of protons, on the scale of 100 µm.

We anticipate needing 3–4 years of experimentation to carry out the program above. At the end of this program, we expect to understand the physics and potential of proton-driven PWFA for future accelerators, and be in the position to propose projects leading to high quality, high charge and high energy electron bunch acceleration.

10 Summary This AWAKE Design Report has outlined a pioneering experiment to demonstrate for the first time proton-driven plasma wakefield acceleration. This technology could lead to future colliders of high energy but of a much reduced length compared to proposed linear accelerators. Using the SPS beam extracted to the CNGS facility, protons will enter a plasma, modulate into micro-bunches and generate a strong wakefield with longitudinal electric fields over 1 GV/m. The wakefield will be harnessed to

36 accelerate a witness bunch of electrons to the GeV scale within a few meters. The baseline experiment has been shown to be realizable and the infrastructure and beams can be made available at CERN to perform the project. The AWAKE Collaboration has been formed and has the resources and expertise to build and run the experiment. The experiment will inform future larger-scale tests of proton-driven plasma wakefield acceleration and applications to high energy colliders.

References [1] T. Tajima and J.M. Dawson, Phys. Rev. Lett. 43, 267 (1979). [2] P. Chen et al., Phys. Rev. Lett. 54, 693 (1985), and 55, 1537 (1985). [3] W.P. Leemans et al., Nature Phys. 2, 696 (2006). [4] I. Blumenfeld et al., Nature 445, 741 (2007). [5] A. Caldwell, K. Lotov, A. Pukhov and F. Simon, Nature Phys. 5 (2009) 363. [6] K. Lotov, Phys. Rev. ST - Accel. Beams 13, 041301 (2010). [7] N. Kumar, A. Pukhov and K. Lotov, Phys. Rev. Lett. 104, 255003 (2010). [8] A. Caldwell, K. Lotov, A. Pukhov and G. Xia, Plasma Phys. Control. Fusion 53, 014003 (2011). [9] E. Adli et al., “Letter of Intent for a Demonstration Experiment in Proton-Driven Plasma Wakefield Acceleration", CERN-SPSC-2011-020, SPSC-I-240, 2011. [10] Technical Notes collected at: https://edms.cern.ch/nav/P:CERN-0000094988:V0/P:CERN-0000097681:V0/TAB3 [11] J.M. Dawson, Phys. Rev. 113, 383 (1959). [12] B. Allen et al., Phys. Rev. Lett. 109, 185007 (2012). [13] C.E. Clayton et al., Phys. Rev. Lett. 88, 154801 (2002). [14] K.V. Lotov, Phys. Plasmas, 18, 024501 (2011). [15] A. Pukhov et al., Phys. Rev. Lett. 107, 145003 (2011). [16] C.B. Schroeder et al., Phys. Rev. Lett. 107, 145002 (2011). [17] K.V. Lotov, J. Plasma Phys. 78, 455 (2012). [18] M. Tzoufras, et al., Phys. Rev. Lett. 101, 145002 (2008). [19] K.V. Lotov et al., Phys. Rev. ST-AB, accepted for publication (2013). [20] D. Gordon et al, Phys. Rev., E 64, 046404 (2001). [21] D.H. Whittum, et al., Phys. Rev. Lett. 67, 991 (1991). [22] A. Caldwell and K.V. Lotov, Phys. Plasmas 18, 103101 (2011). [23] Technical Note: K. Lotov, “Simulations of wakefield excitation and electron acceleration for AWAKE experiment", EDMS Nr. 1276926. [24] K.V. Lotov, A. Pukhov, and A. Caldwell, Phys. Plasmas 20, 013102 (2013). [25] J. Vieira, R.A. Fonseca, W.B. Mori and L.O. Silva, Phys. Rev. Lett. 109, 145005 (2012). [26] Technical Note: E. Öz and P. Muggli, “Rubidium Vapor Plasma Source", EDMS Nr. 1275631. [27] See for example Grant Instruments TXF 200 Heating Circulator, Digital high performance model. [28] See for example Bureau International des Poids et Mesures, Techniques for Approximating the ITS-90, chapitre 2. [29] See for example MBE Komponenten Incorporated. [30] Technical Note: R. Kersevan, “Valve-Induced Density Variation in the Plasma Cell", EDMS Nr. 1275623. [31] Technical Note: P. Muggli, “Ionization Energy", EDMS Nr. 1277527. [32] I. Pinayev et al., AIP Conference Proceedings, 451, 545 (1998).

37 [33] Technical Note: J. Vieira et al., “Seeding self-modulation for the AWAKE proton driven wakefield acceleration experiment", EDMS Nr. 1275629. [34] Technical Note: S.N. Bland et al., “An Argon discharge plasma source for the proton-driven plasma wakefield accelerator experiment at CERN", EDMS Nr. 1275630. [35] D. Arnush, Phys. Plasmas 7, 3042 (2000). [36] J.G. Kwak et al., Phys. Lett. A 267, 384 (2000). [37] F.F. Chen and D. Arnush, Phys. Plasmas 4, 3411 (1997). [38] Technical Note: O. Grulke et al., “R&D towards a Helicon Plasma Cell", EDMS Nr. 1275621. [39] D.G. Miljak and F.F. Chen, Plasma Sources Science & Technology 7, 537(1998). [40] Technical Note: O. Reimann and R. Tarkeshian, “Proposed diagnostics setups for plasma density measurement", EDMS Nr. 1275626. [41] M.C. Nuss and J. Orenstein, Topics in Applied Physics Volume 74, 7 (1998). [42] D. L Crintea, et al., J. Phys. D: Appl. Phys. 42, 045208 (2009). [43] C. Bovet et al., DIPAC 1999 proceedings, pp. 90. [44] Technical Note: P. Muggli, “Time resolved optical transition radiation diagnostic", EDMS Nr. 1275624. [45] P. Catravas et al., Phys. Rev. Lett. 82, 5261 (1999). [46] Technical Note: P. Muggli, “Optical transition radiation shot noise as a diagnostic for proton bunch self-modulation in AWAKE", EDMS Nr. 1275625. [47] Technical Note: O. Reimann and P. Muggli “First Phase Microwave CTR Measurements of the Proton Bunch Modulation", EDMS Nr. 1275628. [48] Technical Note: O. Reimann, S. Mandry and R. Tarkeshian, “Proton Bunch Measurement Using the TCTR Electro-Optical Diagnostic", EDMS Nr. 1275627 [49] F. Coppinger et al., IEEE Trans.Microwave Theory Tech. 47, 1309 (1999). [50] Technical Note: S. Jolly, et al., “Design of the AWAKE spectrometer", EDMS Nr. 1275622. [51] http://www.pulsar.nl/gpt [52] Technical Note: T.C.Q. Noakes et al., “Technical Note on the Specification and Conceptional Design of a Photoinjector for the AWAKE Project", EDMS Nr. 1272164. [53] P. McIntosh et al., “A new electron beam test facility (EBTF) at Daresbury Laboratory for industrial accelerator system development", Proceedings of IPAC2012, New Orleans, Louisiana, USA (2012) THPPR044. [54] J. Brossard et al., “Low energy beam measurements using PHIL accelerator at LAL, comparison with PARMELA simulations", Proceedings of 2011 Particle Accelerator Conference, New York, NY, USA (2011) WEP210. [55] CERN Long Shutdown 1, http://home.web.cern.ch/about/updates/2013/02/long-shutdown-1-exciting-times-ahead [56] Technical Note: CERN AWAKE project team, “The AWAKE Facility at CERN", EDMS Nr. 1272163. [57] E. Gschwendtner et al., “Perfomance and Operational Experience of the CNGS Facility", IPAC2010, 23–28 May 2010, Kyoto, CERN-ATS-2010-153, 2010. [58] L. Jensen, “Beam Instrumentation for the CNGS Facility”, AB-Note-2006-022, CERN, Geneva, 2006. [59] D. Castronovo, R. Fabris, G. Loda and D. Zangrando, “FERMI@Elettra MAGNETS”, Proceedings of IPAC2011, 2011. [60] T. Argyropoulos et al., CERN-ATS-Note-2013-006 MD.

38