Measurement of the Saturation Length of the Self-Modulation Instability A. Clairembaud, M. Turner, M. Bergamaschi, L. Ranc, F. Pannell, J. Mezger, H. Jaworska, N. van Gils, J. Farmer, P. Muggli, AWAKE Collaboration, C.C. Ahdida, Y. Alekajbaf, C. Amoedo, O. Apsimon, R. Apsimon, T. Bachmann, C. Badiali, M. Baquero, E. Belli, A. Boccardi, T. Bogey, S. Burger, P.N. Burrows, B. Buttenschön, A. Caldwell, M. Chung, C.C. Cobo, D.A. Cooke, D. Dancila, C. Davut, G. Demeter, A.C. Dexter, S. Doebert, A. Eager, D. Easton, B. Elward, R. Fonseca, I. Furno, K. Gajewski, D. Ghosal, E. Granados, J. Gregory, O. Grulke, E. Gschwendtner, E. Guran, D. Harryman, M. Hibberd, P. Karataev, R. Karimov, M.A. Kedves, F. Kraus, M. Krupa, T. Lefevre, T. Lofnes, N. Lopes, K. Lotov, J. Mcgunigal, S.A. Mohadeskasaei, M. Moreira, B. Moser, Z. Najmudin, S. Norman, N. Okhotnikov, A. Omoumi, C. Pakuza, A. Pardons, K. Pelckmans, J. Pisani, A. Pukhov, R. Rossel, H. Saberi, M.V. dos Santos, O. Schmitz, F. Sharmin, F. Silva, L. Silva, B. Spear, L. Stant, C. Stollberg, A. Sublet, C. Swain, G. Tenasini, A. Topaloudis, P. Tuev, J. Uncles, F. Velotti, J. Vieira, C. Welsch, T. Wilson, M. Wing, J. Wolfenden, B. Woolley, G. Xia, V. Yarygova, W. Zhang Abstract The self-modulation (SM) instability transforms a long charged particle bunch traveling in plasma into a train of microbunches that resonantly drives large-amplitude wakefields. We present the first determination of the saturation length of SM using experimental and numerical results. The saturation length is the distance over which wakefields reach their maximum amplitude along the plasma. By varying the plasma length and measuring the radius of the transverse distribution of the bunch, we find that the saturation length of SM decreases with plasma density and initial field amplitude, e.g., when seeding. The saturation length is a fundamental parameter of the instability, and these results are key for understanding SM and designing plasma wakefield accelerators driven by long bunches, such as AWAKE, or by long laser pulses for radiation production. Introduction – Instabilities play a central role in plasma physics (see, e.g., Ref. 15 ). In particular, understanding their development and saturation is essential. A fundamental parameter characterizing many instabilities is the saturation length. An instability can grow in space from an initially small (linear) amplitude. Growth leads to nonlinear effects, and at the saturation length the amplitude reaches a large and essentially constant value, limited by these effects. In general, the effect sought after is maximum at saturation, and its event-to-event variations are also smaller than during growth. For example, in free electron lasers (FELs), a particle bunch modulates to emit light with exponentially growing intensity. The saturation length, L_{\mathrm{sat}} , can be determined by measuring the radiation power as a function of undulator length (see e.g., Ref. 5 ). Determining L_{\mathrm{sat}} is crucial, as it sets the minimum undulator length required for the FEL to reach maximum output power. In plasma, an analogous microbunching process occurs in the self-modulation (SM) instability of a relativistic charged particle bunch. SM develops when the duration of the bunch \sigma_{t} is much longer than the plasma electron period \tau_{pe} 17 . It has been observed with electron 11 ; 13 ; 9 and proton 1 bunches. SM also occurs with long laser pulses propagating in plasmas 10 ; 3 ; 28 and has been observed in experiments 22 . Despite these observations, the saturation length of SM has not yet been measured experimentally. The initial transverse wakefields driven by the long bunch have small amplitude and periodically modulate the bunch density with period \tau_{pe} . Since the modulated bunch drives higher-amplitude wakefields, a feedback loop between density modulation and wakefields is set up. This leads to growth of SM along the plasma and along the bunch. The growth saturates when the bunch is fully modulated, i.e., when the resulting train of microbunches resonantly excites wakefields to the highest amplitudes. Unlike in FELs, where radiation power can be measured directly, wakefield amplitudes are difficult to measure 33 . It is therefore necessary to rely on indirect effects of the wakefields to determine the saturation length of SM. Because SM is a transverse process, the formation of the microbunch train leads to that of a halo of defocused particles surrounding the train. This halo formation is both the cause (modulation of the bunch density) and a consequence (effect of transverse wakefields) of the growth of the wakefields along the plasma 29 . One can expect the formation process to slow significantly once the microbunch train is fully formed, and when saturation of SM is reached, at L_{\mathrm{sat}} . This key parameter can therefore be experimentally determined by characterizing the evolution of the halo of defocused particles along the plasma 8 . In this Letter , we determine for the first time, with experimental and numerical simulation results, the saturation length of the SM process. By varying the plasma length and measuring the radius of the halo at a screen downstream of the plasma, we identify the distance over which this radius increases and saturates as a function of experimental parameters. We find that the saturation length of the radius decreases with increasing plasma electron density, and that seeding the SM process makes this length shorter than in the non-seeded case. Numerical simulation results indicate that the saturation length of the halo radius provides a good estimate for that of SM. There is excellent agreement between numerical simulation and experimental results for the saturation length of the halo radius. Figure 1: Schematic of the experimental setup (not to scale). Inset: Transverse distribution of the bunch at the screen for single events after propagating through (a) Rubidium vapor, (b) 9.5\, m of plasma. Colormap: logarithmic scale. Experimental Setup – Measurements were performed in the context of AWAKE (Advanced WAKefield Experiment) at CERN 14 . A schematic layout of the experiment is shown on Fig. 1 . Figure 2: Radial slices of the transverse distribution of the bunch at the screen for different propagation distances in plasma ( L_{p} ). (a) n_{pe}=(1.06\pm 0.01)\times 10^{14}\, cm -3 , (b) (1.98\pm 0.01)\times 10^{14}\, cm -3 , (c) (3.90\pm 0.02)\times 10^{14}\, cm -3 , (d) (7.42\pm 0.03)\times 10^{14}\, cm -3 . Colormap: logarithmic scale. White symbols: average radius of the halo r_{h} for the corresponding L_{p} , error bars: standard deviation of \sim 10 events. Pink diamonds: saturation point of the radius ( r_{h}=0.9\,r_{h,\mathrm{max}} ). t_{\mathrm{RIF}}=100\, ps ( \sim 0.6\sigma_{t} ). (a) N_{b}=(2.98\pm 0.08)\times 10^{11} , (b) (2.98\pm 0.03)\times 10^{11} , (c) (2.86\pm 0.05)\times 10^{11} , (d) (2.87\pm 0.05)\times 10^{11} . A 400\, GeV proton bunch from the Super Proton Synchrotron, with a population N_{b}\cong 3\times 10^{11} particles, is focused to an rms transverse size \sigma_{r}\cong 200\, µm at the entrance of the plasma. The bunch has a normalized emittance \epsilon_{N}\cong 2.2\, mm-mrad, and an rms duration \sigma_{t}\cong 170\, ps. A short laser pulse with FWHM duration 120 fs and energy \sim 100 mJ co-propagates with the proton bunch. It forms a relativistic ionization front (RIF), and ionizes a rubidium vapor (RbI \rightarrow RbII), resulting in a plasma electron density n_{pe} that can be varied in the range \cong(1-10)\times 10^{14}\, cm -3 24 . This range of densities corresponds to plasma periods much shorter than the bunch duration ( \tau_{pe}=2\pi\sqrt{\frac{m_{e}\epsilon_{0}}{n_{pe}e^{2}}}\cong(11-3) ps \ll\sigma_{t}\cong 170\, ps), enabling the development of SM. When placed within the bunch, the RIF defines the onset of the beam-plasma interaction, and thus the timing and amplitude of the initial wakefields that can seed SM 4 . The laser pulse propagation can be stopped at different locations along the 10.3\, m-long vapor column by inserting a thin ( 200\, µm) aluminum foil (laser blocks on Fig. 1 ). The plasma length ( L_{p} ) over which the proton bunch propagates can therefore be varied ( L_{p}=0.5,1.5,...,9.5\, m), and thus also the distance over which SM develops. To observe the development of SM along the plasma, we record the transverse profile of the bunch on a scintillating screen as a function of L_{p} . The screen is placed 20.3\, m downstream from the plasma entrance, and a camera images the light emitted by the screen after the passage of the bunch. Since the energy spread of the proton bunch caused by interaction with the wakefields is small ( \Delta E/E<2.5\% ), the number of counts per pixel recorded by the camera is proportional to the local, time-integrated proton bunch density. A circular attenuating mask with transmission 17\% and radius 4\, mm is placed in an image plane of the optical line of the camera, and approximately aligned with the center of the bunch (Inset (a), Fig. 1 ). This enables simultaneous observation of the faint halo of defocused particles and of the intense bunch core formed by the train, without saturating the 12-bit camera. Results – The inset of Fig. 1 shows time-integrated transverse profiles of the bunch at the screen after propagation (a) in vapor (no laser pulse), and (b) in a plasma with L_{p}=9.5\, m. When propagating in vapor, the unmodulated bunch fits entirely under the mask and has a Gaussian-like distribution (Inset (a)). After propagating in plasma, the bunch exhibits a halo, indicating that SM has developed 29 (Inset (b)). The halo distribution has a sharp outer edge, the radius of which, r_{h} , can be accurately determined following the procedure described in Ref. 30 . We extract a horizontal slice of the distribution of \sim 10 successive events for each value of L_{p} , and plot them consecutively on Fig. 2 for four values of n_{pe} . The instability is seeded, thus ensuring that the process is reproducible from event to event, as shown by the small variations of r_{h} . Under these conditions, varying L_{p} is equivalent to observing the state of the bunch (and related wakefields) at different locations along the plasma for a single SM event. The evolution of the bunch distribution (discussed below for Fig. 2 (d)) can be divided into three distinct phases. During the first phase, from L_{p}=0 to 0.5\, m, the seed fields driven by the unmodulated bunch have small amplitude (few MV/m ) 4 ; 29 , and are predominantly focusing 16 ; 31 . Since the relativistic protons ( \gamma_{0}=427 ) have a large inertia, very little transverse evolution occurs over this short distance in plasma (compare with “No plasma”). During the second phase, occurring from L_{p}=0.5 to 3.5\, m, the periodic, transverse wakefields develop alternating focusing and defocusing regions. As a result, a fraction of the protons are defocused and leave the wakefields transversely, thus forming the halo surrounding the bunch core. This is seen on the image as an increase in charge density at large radii ( r>4\, mm, i.e., beyond the mask) and as a simultaneous decrease under the mask, and thus to an increase in r_{h} , as expected. The observation of the formation of the halo with L_{p} clearly indicates that SM is growing within this section of plasma. We quantify the evolution by measuring the average radius r_{h,\mathrm{avg}} (white symbols), e.g., with an increase from r_{h,\mathrm{avg}}=11.3 to 16.9\, mm for L_{p}=1.5 to 3.5\, m. During the third phase, occurring from L_{p}=3.5 to 9.5\, m, the transverse distribution of the bunch remains essentially unchanged, and r_{h} has saturated at its maximum value r_{h,\mathrm{max}} . This is the signature expected from saturation of the SM process. We note that for these L_{p} , time-resolved images of the bunch (not shown) exhibit deep modulation of the longitudinal bunch density, another signature of the saturation of SM 17 ; 26 . We define the saturation length of the halo radius L_{\mathrm{sat}} as the interpolated L_{p} at which r_{h} reaches 0.9\,r_{h,\mathrm{max}} : a threshold chosen to enable reliable determination of L_{\mathrm{sat}} , i.e., before the growth curve flattens and r_{h} becomes weakly dependent on L_{p} . It is indicated by the pink diamonds on Fig. 2 . These clearly indicate the decrease of L_{\mathrm{sat}} with increasing n_{pe} (( 4.5\pm 0.2) m, to ( 2.9\pm 0.1) m from Figs. 2 (a) to (d)). Figure 3: (a) Saturation length L_{sat} as a function of n_{pe} (same values as on Fig. 2 ) labeled with the quantity it is determined from. Inset: r_{h,\mathrm{avg}}(L_{p}=9.5\,\text{m}) as a function of n_{pe}\,[\times 10^{14}\, cm {}^{-3}] . (b) Normalized transverse wakefield amplitude W_{\perp,\mathrm{max}} along the plasma for the four n_{pe} values of Fig. 2 . Black dashed line: 90\% of peak value. To determine whether the saturation length of the halo radius provides an estimate for that of SM, we perform 2D axisymmetric particle-in-cell simulations using LCODE 27 with parameters similar to those of the experiments. We propagate ballistically the bunch phase space distribution at the end of the plasma L_{p} to a location equivalent to that of the screen in the experiment. This produces transverse proton distributions that can be compared to those obtained in the experiment, e.g., Figs. 1 (a) and (b). We determine the halo radius and the saturation length from these propagated distributions, using the same procedure as for the experimental images. The inset of Fig. 3 (a) shows close agreement between experimental (pink symbols) and numerical simulation (blue symbols) results for r_{h,\mathrm{avg}}(L_{p}=9.5\,\text{m}) , with differences below 1\, mm for radii of more than 12\, mm. The figure shows that the corresponding experimental and simulation saturation lengths of r_{h} are also in very good agreement, with differences of less than 0.5 m. These small differences may be due to differences in the exact input parameters between numerical simulations and experiments 23 , namely bunch transverse size and emittance 12 . Numerical simulation results also provide the amplitude of the transverse wakefields W_{\perp} that lead to the formation of the halo. Their amplitude varies along the bunch, and we plot on Fig. 3 (b) their maximum amplitude as a function of position along the plasma, evaluated at r=c/\omega_{pe} (the plasma electron collisionless skin depth with \omega_{pe}=2\pi/\tau_{pe} ), and in the front of the bunch ( -\sigma_{t}