AWAKE is now listed on the CERN website, click here to see it.

AWAKE appears on the CERN Accelerator Complex map, click here to see it.

Click here to see a summary of published results.

The goal of the AWAKE project is to study the driving of large amplitude (>1GV/m) accelerating plasma wakefields using proton (p⁺) bunches. Plasma wakefields are usually driven using a single laser pulse or (negatively charged) relativistic electron bunch.

Proton bunches are interesting because they are very stiff, i.e., the value of γm (γ the particles' relativistic factor and m their mass) for individual protons is very large, and they carry large amounts of energy. For example a bunch produced by the CERN super proton synchrotron (SPS) with 1x10¹¹ at 450GeV/p⁺ carries over 7kJ, while a large hadron collider (LHC) bunch at 7TeV and the same p⁺ population carries over 100kJ. As a reminder a typical e^-/e⁺ bunch at the energy frontier (~500GeV, 2x10¹⁰ particles), as considered for example for the international linear collider (ILC) carries 1.6kJ. Therefore, a plasma wakefield accelerator (or PWFA) driven by a p⁺ bunches could produce an ILC-like electron bunch in a single (or just a few) plasma stage(s). The most powerful laser pulses suitable to drive wakefields (TW power level lasers) carry less than 100J (e.g., 40J is 100fs). High energy electron/positon bunches short enough (<100fs) to drive multy-GeV wakefield produced amplitudes also cary less than 100J (65J for 2x10¹⁰, 20GeV particles).

A p⁺-driven PWFA would operate at lower plasma density (~10¹⁴-10¹⁵cm^-3) than the high-gradient (~50GeV/m), high density (~10¹⁶-10¹⁷cm^-3) PWFA currently studied at SLAC FACET. The p⁺-driven PWFA offers therefore two advantages when compared to the e^--driven version.
First, it avoids gradient dilution, i.e., the low average gradient resulting from the long distance necessary between PWFA section to in-couple the new drive bunch and capture and re-focus the accelerated bunch into the next plasma section. This distance between meter-scale PWFA stages could be between 10 and 100m for a ~20GeV e^- drive bunch. The peak gradients is therefore reduced or diluted by the ratio of the distance between stages and the plasma length. This is typical issue of a staged version of a PWFA-based accelerator in which each drive bunch carries much less energy (J) than the final energy of the accelerated bunch.
Second, when operating at lower plasma density the tri-dimensional size of the plasma accelerating structure is larger (~λ_pe~n_e^-1/2). Therefore the size of bunches (drive and witness) fitting in the structure is also larger. The tolerances on spatial and temporal alignment are decreased. While the length of the accelerator may be shorter, the plasma length is longer, bringing a challenge for plasma source development (also see below) .

Numerical simulations have shown that a e^- bunch injected at 10GeV can gain up to more than 500GeV along a ~500m-long plasma with a density of 6x10¹⁴cm^-3 [Caldwell, Nature Physics 5, 363 (2009)]. The PWFA is driven by a 1TeV bunch with a longitudinal rms size of 100μm and a transverse rms size of 430μm with 10¹¹ p⁺. The loaded accelerating gradient is larger than 1GeV/m and the final electrons energy is limited by the dephasing of the p⁺ losing energy along the plasma. The energy transfer efficiency in this non-optimized case is on the order of 10%. This suggests that recycling of the drive p⁺ bunch may be required.
However, such short and dense high energy proton bunches do not exist today.

The bunches produced at the CERN SPS or LHC are 10-12cm long (σ_z). It was demonstrated in simulations that a long p⁺ bunch propagating a dense plasma (σ_z>>λ_pe~n_e^-1/2) is subject to a transverse two-stream instability [Kumar, Phys. Rev. Lett. 104, 255003 (2010)]. This Self-modulation instability (or SMI) modulates the density of the bunch with the period of the plasma wave by alternatively focusing and defocusing the bunch particles. As a result, the periodically modulated bunch resonantly drives large amplitude wakefields.
The instability is the result of the feedback between the bunch density modulation δn_b and the plasma density modulation δn_e and wakefield amplitude. In the linear regime of plasma wakefields the plasma electrons locally adjust their density to cancel out the charged particle bunch fields and maintain quasi-neutrality (δn_e~n_b+δn_b). The resulting wakefield amplitude (sustained by the plasma electrons) is proportional to the plasma electron density perturbation. Therefore, the periodic wakefields periodically modulate the bunch density, which leads to larger (periodically focusing and defocusing) wakefield amplitudes, thus closing the feedback loop.
The figure shows an example for an electron bunch and corresponds to a case that we will study at SLAC-FACET [Vieira, Phys. Plasmas 19, 063105 (2012)].

The above movie (courtesy of J. Vieira, IST) shows the density of a half-cut electron bunch (similar to the above one) in the reddish colors and of the plasma electron density (blueish colors). The beam enters the plasma (blue streaming from the right to the left) and is subject to the self-modulation instability (SMI). The bunch moves from right to left. This view is in a numerical simulation window that moves with the bunch at the speed of light and therefore the bunch appears as stationary in this window. The particles in the de-focusing phase of the wakefields leave the simulation box in the transverse direction. The remaining fraction of the electrons form a bunch train that resonantly drives the wakefields visible as the bubbles in the plasma electron density that sustain them. The bunch and plasma structure are also subject to a weak case of the hose instability resulting in local transverse oscillations of the beam and plasma structure. The movie is the result of a 2D, slab geometry simulation using the particle-in-cell code OSIRIS [Fonseca, Lect. Notes Comp. Sci. vol. 2331/2002, (2002)].

The simulation results shown here above (figures and movie) were obtained with bunches shaped with "sharp rising" front edge. The purpose of the shaping is to seed the SMI. In general an instability grows from noise, i.e., from random fluctuations that have the same characteristics as the mode of the instability (frequency, wavenumber, etc.). Since the noise amplitude is assumed to be small the unstable mode amplitude (the transverse focusing field or force in the case of the SMI) initially grows in the linear regime of instability. That means that the growth rate is proportional to the mode amplitude and the amplitude grows exponentially with time or propagation distance (or both). Note that in the SMI case the growth is not exponential (see Kumar, Phys. Rev. Lett. 104, 255003 (2010)). When the mode amplitude is not small anymore, the growth slows and the instability saturates. Seeding the instability therefore increases the initial amplitude of the unstable mode and reduces the length (or time) needed for the instability to saturate. Also, when there are competing modes or instabilities, seeding one of them may allow for the suppression (or mitigation) of these other modes or instabilities. The generation by a shaped electron bunch of large amplitude wakefields that can seed the SMI was recently demontrated experimentally (see Fang, Phys. Rev. Lett. 112, 045001 (2014)). Seeding can be accomplished by driving wakefields ahead of the drive bunch using a short particle bunch or a short laser pulse. Seeding can also be accomplished by suddenly starting the interaction between the drive bunch and the plasma. Shaping, or cutting the bunch is an effective way of seeding the SMI. This method is appropriate for low energy particle bunches can can be easily manipulated, for example using a masking technique [Muggli, Phys. Rev. Lett. 101, 054801 (2008)]. For high energy particle bunches (such as p⁺ at 400GeV), a sudden creation of the plasma propagating within the drive bunch may be more appropriate. This relativistically moving gas/plasma boundary maybe created by an intense laser pulse ionizing the gas.
This seeding method has been chosen for AWAKE.
It is important that the laser pulse and the density modulation within the proton bunch do not significantly "dephase" along the plasma. Calcutations show that although they travel at (sligtly) different speeds, the propagation distance difference is on the order of 30μm over the 10m-long plasma, small when compared to the plasma wavelength (typically 1mm) and is thus not an issue.

The SMI results from the coupling between the transverse plasma wakefields and the bunch density modulation. It is a particular case of a convective, transverse two-stream instability. There is another (and competing!) beam-plasma convective, transverse two-stream instability: the hose instability [Whittum, Phys. Rev. Lett. 67, 991 (1991)]. It also results from the coupling between the local transverse centroid position of the bunch and the transverse wakefields. Since both instabilities arise from transverse beam-wakefields coupling they can compete with each other. The hose instability leads to large transverse oscillations of the bunch centroid and can destroy the bunch. It must be suppressed or mitigated to allow for the SMI and the resonant excitation of wakefields to occur. Initial theoretical and simulation results indicate that seeding the SMI can mitigate the development of the hose instability [Vieira, Phys. Plasmas 19, 063105 (2012)].

It may sound strange to build an accelerator based on an instability. The SMI is used to transform the long p⁺ bunch into a train of shorter bunches. The SMI is strongly seeded, in the case of AWKAKE by the relativistic ionization front propagating inside the p⁺ bunch itself. The seeding provides a reference phase for the wakefields to start. In order to be able to inject electrons at the proper phase of the wakefields one needs to fix both the electron injection phase with respect to the ionizing laser and the phase of the wakefields with respect to the same laser pulse.
The delay between the ionizing laser pulse and the electron bunch is fixed at the sub-100fs level by using the same laser oscillator pulse to seed the laser amplifiers producing the ionizing pulse and the pulse for the FR-photo-injector.
Preliminary simulation results show that, after the saturation of the SMI, the phase of the wakefields with respect to the ionizing laser pulse is a weak function of bunch parameters variations, despite the growth of the SMI [see Savard, to appear in Proceedings NA-PAC 2016]. Therefore, at this point in time there is no evidence that the use of the SMI may prevent determinisic external injection of electrons into the accelerating and focusing phase of the wakefields. In fact, with sufficient seeding level (as in AWAKE), the developemnt of the SMI can be seen as amplification of the initial wakefields rather than free evolution of an instability with a dominant random character.

While studying SMI is an interesting physics topic, the goal is to drive plasma wakefields to large amplitude to accelerate particles to large energy. These particles, contained in a witness bunch generated either by a conventional injector/accelerator or by a plasma-based one [Muggli, IPAC 2014 Proceedings, 1470 (2014)], must be injected into the wakefields. For them to propagate over a long distance they must be injected into the focusing phase of the wakefields. For them to be accelerated to large energies they must be injected into the accelerating phase of the wakefields. Note that this phase is different depending on the charge (positive or negative) of the witness bunch. In the linear PWFA regime, a quarter of the plasma wake period satisfies these two conditions (focusing and accelerating). This phase must be known or at least it must be possible to determine it experimentally. If the SMI started from noise, position of this phase relative to the bunch position would be random from event to event. Seeding of the instability with a "signal" of known position along the bunch removes this random variation (see above, "Instability Seeding"). Numerical simulation results show that as the SM develops, the phase velocity of the wakefields is slower than that of the drive bunch or of the seed laser pulse or even of the injected electrons [see for example Pukhov, Phys. Rev. Lett. 107, 145003 (2011)]. Therefore, assuming the electrons are injected in the proper phase at the plasma entrance, the wakefields will dephase (slower) and the electrons may be lost. Therefore, side injections was proposed [Lotov, Journal of Plasma Physics 78(04), 455 (2012)]. Optimum side injection must occur near the saturation position of the SMI along the plasma. However, it was recently discovered in simulations that similar trapping and acceleration results can be obtained when injecting the witness electron bunch on axis (colinearly with the p⁺ and laser beams) and at the plasma entrance [Lotov, Phys. Plasmas 21, 123116 (2014)]. With a long particle bunch, the injection process traps a fraction of the injected particles (a few %). These electrons dephase and bunch near the peak of the accelerating field. They can be accelerated and form a narrow final energy spectrum. On-axis injection will therefore be used initially to maximize the probabilities of success of the experiment.

The above figure shows a schematic of the experimental setup. The p⁺ bunch extracted from the SPS (1) is merged with the ionizing laser pulse (1'). The laser pulse and the proton bunch travel together through the metal vapor cell (Rb or Cs). The co-propagating ionization front shown in (2) provides the seeding for the SMI. Growth of the SMI and the resulting self-modulation of the p⁺ bunch occurs over the first 3-5m of plasma (3). The modulated bunch resonantly drives wakefields over the remaining length of plasma (4). The laser pulse is dumped (5) and the bunch radial modulation is measured using electro optic sampling (EOS) diagnostics (6) and optical transition radiation (OTR) and coherent transition radiation (CTR) diagnostics (7). A rf-gun driven by a laser pulse derived from the ionizing laser produces a witness electron bunch (8). The electron bunch is injected into the wakefields (see (2)) on axis. After the plasma the electron bunch energy spectrum is measured using a broad acceptance magnetic spectrometer (9).

This figure (above) shows a technical version of the above schematic, showing the various key items of the experiments (plasma source, laser, rf-gun, etc.) as boxes with realistic size. The layout is based on the floor plan of the CNGS facility at CERN. This first step towards an actual AWAKE experiment was conceived by the CERN team.

The experiment was built and mostly commissioned in 2016. At 360° view of the facility is shown below (Image: Maximilien Brice/CERN).

A key component of the AWAKE experiment is of course the plasma. The source must provide a plasma ~10m-long, with a radius on the order of the bunch radius (typically >3σ_r or about 1mm in this case, since σ_r~200μm). The density must be adjustable in the 10¹⁴-10¹⁵cm^-3 range. This range is chosen to keep the beam radius smaller than the cold, non-collisional plasma skin depth: c/ω_pe, also expressed as k_peσ_r<1, k_pe=1/(c/ω_pe). If the bunch radius were smaller it could be subject to the current filamentation instability [Allen, Phys. Rev. Lett. 109, 185007 (2012)]. The growth of the SMI is not strongly dependent on the density uniformity along the plasma. However, once the self-modulated bunch resonantly drives wakefields after the SMI saturation particles must be injected into and maintained in the proper phase of the wakefields. This leads to a rather tight requirement of δn_e/n_e<0.25% [see for example Lotov, Phys. Plasmas 20, 013102 (2013)]. We will use a rubidium gas vapor cell in which the density uniformity is given by the temperature uniformity along the plasma source (δn_Rb/n_Rb=δT/T). This very uniform vapor column is fully ionized of its first electron by a short (~100fs) and intense (I₀>1.2x10¹²Wcm^-2, P=1-2TW) laser pulse co-propagating with the p⁺ bunch. This laser pulse or the ionization method:

-creates the plasma
-seeds the SMI
-provides the shortest time for the plasma density to decrease or become non uniform, i.e., approximatively the duration of the p⁺ bunch.

In the volume where the laser pulse intensity exceeds the ionization intensity the gas/vapor is fully stripped of its first electron. Therefore the plasma density uniformity becomes equal to that of the neutral vapor and meets the requirement, that is δn_e/n_e=δn_Rb/n_Rb [E. Oz and P. Muggli, Nucl. Instr. Meth. Phys. Res. A 740(11), 197 (2014)].
We note that simulations show that a plasma radius somewhat smaller than the bunch radius does not impede the development of the SMI (see Fang, Phys. Plasmas 21, 056703 (2014)).
The picture below shows the vapor source in the AWAKE tunnel (Image: Maximilien Brice/CERN).

The first phase of the experiment will consist in observing the self-modulation of the p⁺ bunch. Recall that with a nominal plasma density of 7x10¹⁴cm^-3 the plasma frequency is about 240GHz (i.e., period of ~4ps) and the relativistic plasma wave wavelength is approximately 1.2mm. Recall also that the SMI leads to a radial modulation of the bunch charge (focusing/defocusing) with a longitudinal period approximately equal to that of the plasma wave.
The angle of the defocused p⁺ can be measured using two screens placed downstream form the plasma [see Turner Nucl. Instrum. Meth. A 829 (2016) 314]. This is the first evidence that the SMI has occured.
The charge modulation can be converted into temporal modulation of photons using transition radiation. Transition radiation has a very broad spectrum.
At wavelengths much shorter than the typical spatial charge distribution (the bunch length or the self-modulation) this radiation is incoherent. In the visible range (400-800nm) this radiation is called optical transition radiation (OTR). The OTR is prompt and can be used to directly analyze the time structure of the bunch. This light can be imaged onto the slit of a streak camera with sup-picosecond resolution to directly measure the modulation period. Test measurements with a ps-resolution streak camera using the beating and the gating of two cw laser beams in a fiber optics have shown that modulation frequencies up to ~400GHz can be measured (see Rieger, Accepted for publication in Reviews of Scientific Instruments (2017)).
At wavelengths much longer than the typical spatial charge distribution this radiation is coherent. It is know as coherent transition radiation (CTR). Because it is coherent it can be used for interferometry. But also, its intensity scales as the square of the number or bunch particles, while that of OTR scales linearly with it. Because of the wide gap between the unmodulated bunch length (~12cm) and the modulation period (~1.2mm) the much larger radiation power in the microwave range between the two cases can be detected and used as evidence of SMI occurrence. The occurence of SMI can be evidenced by a microwave Schottky diode placed after a waveguide in cutt of for the long bunch rdaition (e.g., frequencies below 100GHz or wavelengths longer than ~3mm). The modulation frequency is also measured using a heterodyne measurement scheme.