Electron cloud effects the LHC in Giovanni Rumolo

Electron cloud effects the LHC in Giovanni Rumolo, G. Iadarola and O. Dominguez in LHC Beam Operation workshop - Evian 2011, 13 December 2011 For all LHC data shown (or referred to) in this presentation: V. Baglin, H. Bartosik, P. Baudrenghien, G. Bregliozzi, S. Claudet, J. Esteban. Müller, G. Lanza, G. Papotti, F. Roncarolo, E. Shaposhnikova, L. Tavian

Outline Focus of this talk Analysis of the 2011 observations and measurements – How we observe electron cloud in the LHC – Resume of the effect of the scrubbing run with 50 ns beams (1 – 11 April 2011) – Experience and progress with 25 ns beams → Historical: MD sessions from the 29 June 2011 to the 24 October 2011 → Scrubbing of unbaked/uncoated field free regions → Scrubbing of the arcs → Evolution of some beam observables – Concluding remarks 2

Electron cloud observation in the LHC → The electron flux to the chamber wall Fe is revealed through 1) Pressure rise Beam chamber 2) Heat load 3

Electron cloud observation in the LHC → The presence of electrons with density re around the beam causes 1) Beam coherent instabilities, single or coupled-bunch type, for the last bunches of a bunch train Beam 2) Incoherent emittance growth, degrading lifetime, slow losses Obviously, both Fe and re depend on the beam structure and on the surface properties, e. g. R 0 and dmax From the evolution of the observables during scrubbing, we can infer the decrease of dmax ! 4

2011 scrubbing run in one slide! ⇒ A 10 -days scrubbing run took place at the beginning of April 2011, during which 50 ns spaced beams with up to 1020 bunches per beam were injected into the LHC and kept at 450 Ge. V/c. ⇒ It resulted into a very efficient machine cleaning – The dynamic vacuum decreased by one order of magnitude – The heat load on the beam screen in the arcs → was significant at the beginning of the scrubbing run → disappeared at the end of the scrubbing run, even with higher number of bunches injected – The average stable phase over the beam decreased by one order of magnitude – Instabilities and emittance growth, clearly visible at the beginning of the scrubbing run, disappeared at the end even with low chromaticity settings ⇒ After the scrubbing run and first test ramps, the machine became ready to operate for physics with 50 ns beams ⇒ The number of bunches per beam was ramped up to its maximum (1380) within two months 5

dmax in the arcs after the 50 ns scrubbing ⇒ Heat load measured during the ramp in physics fill 1704 compared with the one predicted by numerical simulations ⇒ The measured heat load is compatible with values of dmax=2. 1 -2. 2 for R 0=0. 6 -0. 7 ⇒ The expected dmax threshold is about 2. 2 @ 450 Ge. V and 2. 1 @ 3. 5 Te. V 40 -60 m. W/m Simulation scan Measurement in fill 1704 Simulations by H. Maury-Cuna 6

dmax in the uncoated and/or unbaked sections: estimation technique • The evaluation of dmax is done in the field-free regions in proximity of the pressure gauges – Used Beam 1 data from gauges (Cu): VGI. 141. 6 L 4. B and VGPB. 2. 5 L 3. B – A solution (R 0 , dmax) is found comparing the pressure rises DPi measured at different injections with the electron fluxes Fi from simulations Measured pressures � Baked but uncoated: SEY ~1. 6 -1. 9. � Length 0. 3 m � Pumping speed from NEG and maximum for CH 4 ≈ 10 L/s NEG Simulated electron fluxes 7

dmax in the uncoated and/or unbaked sections: results • Pressure rise measurements with 50 ns beam to estimate dmax in the field-free regions in proximity of the pressure gauges (R 0 ≈0. 2) – Measurements done at the beginning and at the end of the scrubbing run – Measurements done during the 50 ns operation of LHC (19 May) – As expected, we are asymptotically approaching the dmax threshold for 50 ns beams Calculated threshold for 50 ns beam 29 June 2011, date of the first injections of 25 ns beams in LHC On the 29 June, a new story begins, with the 25 ns beams in LHC … 8

25 ns experience in 2011 29/06 DATE 14/10 24/10 SHORT DESCRIPTION 29 June Injections of 9 x 24 b trains per beam with different spacings between them 28 August First attempt to inject a 48 b train: fast instability dumps the beam within less than 1000 turns after injection 07 October High chromaticity (Q’x, y ≈15): Injection tests with trains of 72 -144 -216 -288 bunches from the SPS + ramp & 5 h store with 60 b (12+24+24) per beam 14 October High chromaticity: injection of up to 1020 bunches per beam in 72 b trains (decreasing spacings between trains: 4 -3 -2 -1 ms) 24 -25 October Injection of up to 2100 bunches in Beam 1 and 1020 in Beam 2 (1 ms train spacing) Scrubbing

dmax in the uncoated and/or unbaked sections: results (II) 29/06 • • 14/10 24/10 Pressure rise measurements during most of the 25 ns fills were found hard to be used for the dmax estimation because of beam losses leading to rapidly changing regimes After considerable 25 ns scrubbing, i. e. at the end of the 24/10 MD session, 8 x 72 b batches with different spacings could be injected for Beam 1 into the LHC and remain stable to allow the pressure values to level 10

dmax in the uncoated and/or unbaked sections: results (II) Calculated threshold for 50 ns beam Calculated threshold for 25 ns beam • • Start of 25 ns beams in LHC Scrubbing with 25 ns beam (~40 h) has lowered dmax to 1. 35 ! Again, we are not far from the threshold for 25 ns beams, but further scrubbing is needed 11

dmax in the arcs: estimation technique 29/06 14/10 24/10 Five snapshots in the 25 ns MDs to reproduce heat loads [W/hcell] the. Measured measured heat load by averaged over sectors from simulations! cell by cell data 12

dmax in the arcs: estimation technique fast. BCT + bblength (B 1) 24/10 fast. BCT + bblength (B 2) 13

dmax in the arcs: estimation technique fast. BCT + bblength (B 2) fast. BCT + bblength (B 1) Simulator Py. ECLOUD Simulated heat loads Measured heat load 14

dmax in the arcs: results 29/06 14/10 24/10 R 0 = 0. 7 dmax has decreased from the initial 2. 1 to 1. 55 in the arcs ! Calculated threshold for 25 ns beam (450 Ge. V) Calculated threshold for 25 ns beam (3. 5 Te. V) 15

dmax in the arcs: results Simulations dmax fixed to 1. 55 for the last fill on the 25 October Measurements the energy loss per bunch is obtained from the stable phase shift ZOOM Beam 1 16

dmax in the arcs: results ⇒ Excellent agreement between bunch by bunch synchronous phase shift and simulated energy loss at the saturation of the e-cloud ⇒ Build up phase of the electron cloud still not reproduced by simulations with the same accuracy and level of detail – Simulation underestimates the primary electron generation? – In reality, stronger memory effect between batches → Larger R 0 ? unlikely, because 0. 7 is already a high value → Uncaptured beam between batches? (SPS experience) – Dynamic range of the measurements? – Energy loss from impedance, dominant for the first bunches in each batch and for the last batches in the full train? – Further check with the bunch by bunch position data from BQM ⇒ Model confirms cross-calibration between stable phase shift measurements to measured heat load data

Beam observables: Transverse emittances 14 October batches injected with 4 ms spacing, Q’x, y=15 • Both beams still unstable in the two planes, or anyway affected by emittance growth • Some visible benefits from scrubbing: – The effect of the electron cloud manifests itself later along the trains, in spite of the closer spacing! – October First 1 – 2 trains seem to be hardly now 24 batches injected with 1 affected ms spacing, Q’x=3, Q’y=15 – In general, improvement in vertical • Lowering horizontal chromaticity did not seem to degrade the beam horizontally, but rather it slightly improved it: effect of scrubbing?

Electron Cloud Instability What do we expect ? HEADTAIL simulations by Kevin Li • • • Calculated coherent ECI threshold for central density in dipoles is around re=1012 m-3 for nominal intensity at 450 Ge. V (simulations were run assuming the whole LHC made of dipoles) It can be stabilized with chromaticities Q’x, y>15, but emittance growth due to electron cloud + chromaticity remains! Right plot shows that this could be achieved only for dmax ≤ 1. 5

Beam observables: Losses and lifetimes Beam 1 24 October batches injected with 1 ms spacing Steady improvement visible on 2 nd and 3 rd train Lifetime degrades and then recovers Degrading lifetime Fast losses

Concluding remarks (25 ns) dmax (estimated) dmax (threshold @450 Ge. V) St. St (straight section) 1. 35 1. 25 Beam screen (arcs) 1. 55 1. 45 dmax (threshold @3. 5 Te. V) 1. 37 ⇒ Further scrubbing is needed to suppress the electron cloud ⇒ Tricky, as the efficiency of scrubbing decreases with scrubbing itself… – The electron dose measured in lab to decrease the dmax on Cu by an extra 0. 1 from 1. 55 to 1. 45 is about the same needed to decrease it from 2. 1 to 1. 55. – The flux of scrubbing electrons decreases with lowering dmax ⇒ Instability threshold for 25 ns beams very close to the build up threshold − Not much margin to be in the comfortable situation of scrubbing without significant beam degradation ⇒ Significant extra gain could be boosted by − Multi-train injections from the SPS − Find a comfortably stable filling pattern at 450 Ge. V and ramp to 3. 5 Te. V to benefit from photoelectrons and from the lower electron cloud build up threshold 21

Concluding remarks (50 ns) dmax (estimated) dmax (threshold @450 Ge. V) St. St (straight section) 1. 35 1. 63 Beam screen (arcs) 1. 55 2. 2 dmax (threshold @3. 5 Te. V) 2. 1 ⇒ As could be expected, before the 25 ns beams in the LHC, the dmax values were just about the build up threshold for nominal 50 ns beams ⇒ After the 25 ns MDs, the LHC beam chambers have been cleaned to dmax values well below the build up threshold for nominal 50 ns beams ⇒ Simulation work on dmax thresholds as a function of bunch intensity is ongoing, but first results show little dependence at least up to bunch populations of 1. 8 x 1011 ppb ⇒ If we keep the present level of conditioning, ‘ecloud-less’ operation of LHC with 50 ns beams up to high intensities should be guaranteed (bar specific situations in common beam chambers, which need to be checked) 22

Thank you for your attention Very special thanks to G. Iadarola, H. Bartosik, O. Dominguez, J. Esteban. Müller, and F. Roncarolo for their careful off-line analysis of large amounts of MD data and the huge simulation effort that improved the general understanding of electron cloud and scrubbing! Many thanks to V. Baglin, P. Baudrenghien, G. Bregliozzi, S. Claudet, G. Lanza, G. Papotti, E. Shaposhnikova, L. Tavian for all the beautiful data they kindly provided us with and the numerous discussions Thanks to G. Arduini, B. Goddard, V. Kain, K. Li, H. Maury-Cuna, E. Métral, S. Redaelli, B. Salvant, F. Zimmermann, and all those who promoted and/or actively participated in the MDs

Some references ⇒ Past observations, measurements and studies – 2010 experience reviewed in the note “ 50 and 75 ns operation in the LHC: Vacuum and Cryogenics observations”, G. Arduini et al. , CERN-ATS-Note-2011046 MD – 2010 experience + 2011 scrubbing run reviewed in several LBOC/LMC talks and in IPAC 2011 paper and talk “Electron Cloud observation in the LHC”, G. Rumolo et al. , CERN-ATS-2011 -105 – “Observations of Electron Cloud Effects with the LHC Vacuum System”, V. Baglin et al. TUPS 01 in IPAC 2011 – “Electron Cloud Parameterization Studies in the LHC”, O. Domínguez et al. , CERN-ATS-2011 -142 – “Simulation of Electron-cloud Build-Up for the Cold Arcs of the LHC and Comparison with Measured Data”, H. Maury Cuna et al. , CERN-ATS-2011 -135 – “Review of beam instabilities in the presence of electron clouds in the LHC”, K. Li and G. Rumolo, CERN-ATS-2011 -095 – “Injection into LHC of bunches at 25 ns spacing” G. Arduini, B. Goddard, et al. , CERN-ATS-Note-2011 -050 MD – “Benchmarking Electron-Cloud Build-Up and Heat-Load Simulations against Large-Hadron-Collider Observations” H. Maury-Cuna et al. , CERN-ATS-2011218 24

dmax in the uncoated and/or unbaked sections: method of estimation • The evaluation of dmax is done in the field-free regions in proximity of the pressure gauges – Gauges explored (St. St): VGI. 141. 6 L 4. B and VGPB. 2. 5 L 3. B, Beam 1 data – Beams are injected with a filling pattern with different spacings, to try to extrapolate memory effects 3 ms 2 ms 16 ms � Baked but uncoated: SEY ~1. 6 -1. 9. � Length 0. 3 m � Pumping speed from NEG and maximum for CH 4 ≈ 10 L/s NEG 1 ms 25

dmax in the uncoated and/or unbaked sections: method of estimation • The evaluation of dmax is done in the field-free regions in proximity of the pressure gauges – The pressure rise DP is read at the gauge when injecting a new batch into LHC and is assumed to be proportional to the electron flux Fe – Numerical simulations can be used for finding the pairs (R 0, dmax) compatible with the measured pressure rise ratio – By using more measured points, the point/region where the curves intersect will define the pair (R 0, dmax), or narrow range, solution of of our problem Simulated surface Plane of measured ratio (161) Plausibleofsolution Intersection the two R 0 (R ≈0. 2, d)max =1. 35 – 1. 4 in the plane 0, dmax 26

dmax in the arcs: results Simulations dmax fixed to 1. 55 for the last fill on the 25 October Beam 2 Measurements the energy loss per bunch is obtained from the stable phase shift ZOOM 27

dmax in the arcs: results

Beam observables: Losses and lifetimes Beam 2 24 October batches injected with 1 ms spacing Improvement visible on 2 nd and 3 rd train Degrading lifetime Fast losses