S-Band Side Coupled Linac Maria Rosaria Masullo and Vittorio Giorgio Vaccaro

Transcription

1 S-Band Side Coupled Linac Maria Rosaria Masullo and Vittorio Giorgio Vaccaro Istituto Nazionale di Fisica Nucleare - Section of Napoli Università degli Studi di Napoli Federico II, Napoli Thanks to all the students, doctorate students and post-doctored who have collaborated to this work with enthusiasm and intelligence. 1

2 Where are now all these young persons?????? Rita Buiano working as Engineer Sara Lanzone working at Adam in Geneva Roberta Mansueto working as Engineer in a company Alessandro D Elia working at CERN with a contract at Daresbury Giovanni De Michele working as Ph.D. Student at CERNERN Claudio Serpico working at Electra (Trieste, Italy) 2

3 The steps of the talk Historical background A new mechanical design of SCL PALME : Full linac structure from 30MeV up to 230 MeV. An extensive analysis of the design. ACLIP: A Proof of Principle. Design, fabrication and tests of the first module from 30MeV to 35 MeV. 3

4 Which accelerator for particle therapy? Synchrotron Cyclotron Linac Cyclotron (low energy) + Linac (ad hoc designed, the most compact possible ). This system was originally proposed by Amaldi (1993) with idea of finding solutions which have better performances and are less expensive with respect to distinct devices having the same functions. 4

5 Cyclotron + Linac Why a linac? Modularity - the energy can be easily and quickly varied: few msec are required for varying the particle energy The energy variation doesn t degrade the beam quality. 5

6 Cyclotron + Linac Which linac? (1994) Mario Weiss, at CERN, proposed a Side Coupled Linac (SCL) from 62 MeV (same frequency of electron linacs used in conventional radiotherapy: high gradients and thus a relative short accelerator) A collaboration headed by Amaldi designed, built and tested an SCL (named LIBO from LInac BOoster) from 62 MeV to 73 MeV. (2002) A 62 MeV proton beam was successfully accelerated at LNS of INFN in Catania!! 6

7 Cyclotron + Linac Which linac? [2 nd round] It appeared that it would be less expensive to have a cyclotron of lower energy (say 30 MeV) and a linac. However the performances of a SCL worsen with decreasing energies (mainly the power needed per energy gradient increases) unless one resorts to ad hoc remedies. This problem has been tackled by different research lines. In our approach we left behind the idea to compact the linac at any cost. 7

8 In standard SCL, designed for low beam energy, the cavity aspect ratio (surface/volume) is quite unfavorable: because of mechanical reasons, the septum thickness cannot be smaller than 3.6 mm. In this new configuration one may get rid of that constraint! From now on the septum thickness is set at 1.8mm. The new tile is named BBAC (Back to Back Accelerating Cavity). Standard tiles New tiles: BBAC 8

9 The basic tile BBAC 9

10 MANY ADVANTAGES -the septum between two adjacent cavities is no longer 60 obtained by setting two tiles back 55to back, the manufacturing 50 is simplified and the septum thickness 45 can be reduced 40 without impacting the quality of the 35 brazing process; 20 -the increase of the volume/surface 15 ratio in the cavity 10 produces an increase of the shunt impedance; Z_sh [MOhm/m] Kinetic energy [MeV] BBAC Tile: Septum = 1.0 mm BBAC Tile: Septum = 1.4 mm BBAC Tile: Septum = 1.8 mm Standard Tile: Septum = 3.6 mm -the larger width of the accelerating cavity and its asymmetric cut allow to easily insert the frequency rod tuners; -the reduced number of tiles required in the assembly provides a more relaxed mechanical tolerances and a general reduction in machining costs (estimated as 40%). 10

11 Two distinct SCL projects 1) PALME project: full linac structure from 30MeV up to 230 MeV, low field gradient in order to minimize the total feeding power. Exhaustive studies on the beam dynamics, activation and thermomechanical behavoiur 2) ACLIP project: devoted only to performe the PoP with high gradient in order to maximize the energy gain. Exhaustive studies on the beam dynamics. Only two modules have been mechanically designed in detail, one fabricated and completely tested (tuning,high power test, acceleration). 11

12 PALME and ACLIP Collaboration between University fo Bari, Milano and Napoli, INFN sections of Bari, Milano and Napoli; External collaboration with NRT and e2v. 12

13 PALME project The system is formed by 25 modules for a final energy of 230 MeV. Each module is made of two tanks connected by a Bridge Coupler The mean accelerating field on axis in a cavity is 16.5 MV/m The RF power per tank is about 1 MW Duty cycle between 0.05 % and 0.1% It is assumed a mean injection current of 150 µa From the beam dynamics we will know that only 8% of the injected current will be transmitted The available mean current for therapy is : 6 na-12na 13

14 PALME project: Beam Dynamics Simulations with all 25 modules on show that a power of 1MW for each tank is required for a final output energy of 230 MeV. To reduce the output energy MeV) a specified number of last modules can be switched off. Data show that with 10 modules switched off the transmittance is not influenced (we do not loose particles!). Beam transmittances Total Transmittance tank tank 52 All modules ON 7.78% 7.10% 10 modules(20 tanks) OFF 7.16% 7.14% Partial Transmittance ± 1 MeV T o tal T ( % ) # Tank 10 off All on Total T (%) off All on # Tank 14

15 PALME project: Beam Dynamics When some of last adjacent modules are switched off, the energy spread of the output beam is almost similar to the one at 230MeV. This can be verified in the presented example of 10 tanks switched off. The maximum at 131 MeV fits with expectation since we have an energy increase of about 10 MeV per module. Fraction of total transmitted particles (%) off All on W (MeV) Particle energy distribution: total amount Fraction of useful transmitted particles (%) off All on W (MeV) Particle energy distribution : +/- 1MeV 15

16 Energy modulation : PALME project: Beam Dynamics Particular attention must be paied when reducing continuously the output energy. As an example two cases are shown: the field is in the last 4 and 5 modules. Fraction of total trasmitted particles [%] % 50% W [MeV] In these two cases, we could naivly expect an output energy distribution with its maximum at 210 MeV and 205 MeV respectively. Conversely we get 195 MeV and 185MeV. This is because the buckets are going out of step with respect to RF. In addition to this, it is apparent that the beam energy spread dramatically increases for the 5 module case. This phenomenon is even more pronounced at lower energy. If only in the last active module the power is modulated, this phenomenon does not appear. 16

17 PALME project: Beam Dynamics We learned that, if we want to reduce the output energy, this can be done only acting on 1 module. 1 : the expected maximum energy is like what we get (225.18MeV)!! 1 & 4mod off differ about 1% from what we get (185,4MeV instead of 183,4MeV)!!!!! Fraction of total trasmitted particles [%] All ON 50% 50% + 4 OFF Then one must foresee a number of feeders as large as the number of modules. It would be too expensive to have one klystron one module W [MeV] A possible option is the use of magnetrons. 17

18 PALME project: thermo-mechanical analysis Due to power dissipation on the walls, the temperature of the inner part of the cavity increases more than that of the bulky parts. A thermo-mechanical analysis on an infinite sequence of accelerating cavities with adiabatic walls on the top and on the 30 o C on the side walls, mean accelerating field of 16.5 MV/m and a duty cycle of 0.2%. The bulky part does not allow the radial expansion of the inner part (mainly the noses): therefore significant internal stresses may be produced in the metal with effect on the cavity mechanical properties. For standard tiles we obtain a temperature rise on the nose of about 19.7 K For BBAC tiles with a septum of 1.8 mm, we observe a temperature rise of 27.6 K on the nose!!! 18

19 PALME project: thermo-mechanical analysis When the septum thickness becomes smaller, there are two competing effects the Z sh increases, so that the dissipated power decreases the thermal resistance increases. The net effect is an increase of temperature. Design a new BBAC cavity in order to reduce the temperature rise, while keeping the quality indices, like Z sh and dissipated power, within acceptable ranges. 19

20 PALMEproject: thermo-mechanical analysis Flat wall BBAC cavity BBAC cavity with tilted wall Tilt = rad Heat flux distribution The heat flux is negligible on the nose and maximum in the peripheral part of the cavity. Shunt Impedance [MΩ/m] Temperature Rise [K] Standard Tile BBAC flat BBAC tilted The trade-off between the shunt impedance and temperature rise is convenient 20

21 Optimize the PALME project: activation studies Verify that the beam losses are compatible with safety requirements The maximum amount of dose rate at saturation is given by Zn-63, but because of its short half life, it becomes quickly negligible. Dose for present nuclides Nuclide Half life Dose rate at sat. (µgy/h ) Zn61 89s 0.12 Zn63 38m 140 Cu61 3 h 12.6 Zn62 9 h 16.2 Cu64 13 h 19.0 Zn65 244d 32.6 Total 220 The dynamics analysis suggests us that the activation process is restricted to the first two modules. After 3.0 h from the beam stop event the total dose rate falls to a value of the order of 100 mgy/h which is compatible with the possibility of carrying out a maintenance procedure. An emergency intervention of few minutes is also allowed. 21

22 From PALME to ACLIP The conclusion of the PALME analysis is that it is worthwile to pursue the Proof of Principle The system is formed by two modules: the first from 30MeV to 35MeV, the second from 35MeV to 41MeV. Both the modules will be powered by a single RF feeder. Each module is made of two tanks connected by a Bridge Coupler Each tank has 13 accelerating cells and 13 coupling cells The mean accelerating field on axis in a cavity is 20MV/m RF power per module 2.8 MW 22

23 ACLIP project The First Module from 30 MeV to 35 MeV : just out of the oven!!! 23

24 The tuning problem It has already been mentioned that, in order to reduce the manufacturing costs, the machining tolerances have been relaxed. In general, the cavities may be detuned with respect to the nominal value. In order to perform the tuning we must detect which cavities are detuned. Furthermore we need to know some other quite important parameters as the coupling constants. A 5 coupled cavity system In a coupled cavity linac, the cavities loose their individuality. On the other end the only measurable quantities are the resonance frequencies of the system (F m ). From them it is not possible to easily derive the parameters of the structure. This makes difficult the problem of tuning. 24

25 The tuning problem (the lumped circuit ) The coupled cavity system can be modeled, in a narrow frequency band (in our case 4%), as a lumped constant circuit. This representation is extremely fruitful and leads to Nth-order algebraic equation whose zeros are the resonance frequencies of the whole system. We are in the peculiar situation for which we know the zeros but we do not know the equation coefficients, which have a complicated relation with the system parameters. However it is well known that relation exhiste between the zeros of an algebraic equation and the coefficients. 25

26 Which measurements can we perform and which information can we get?? The adopted procedure is a generalization of a method suggested by Luigi Picardi : it consists in varying the resonant frequency of the coupling cells (by means of tuners) and to measure the variation of resonance frequencies (F m ). It can be 2 2 demonstrated that in xy plane the two variables F and move on a 4 + F straight line F F 4 From the straight line coeffiencients one may derive the wanted parameters 26

27 The results of the measurements. The described procedure may be repeated for some other couple of quantities which will move on a straight line. Note: it is not necessary to know the amount of the produced frequency variation!!! Remark the accuracy of the measurements! Structure Parameters (tank 1) k1 (%) 3,39 ± ka (%) -0,67 ± kc (%) -0,04 ± N V fe (MHz) 3 004, 392 ± 0,003 fa (MHz) 3 005, 44 ± 0,02 fc (MHz) 2996, 46 ± 0,03 Measurements F m Fit the straight line coefficients Structure parameters (f a f c f e k 1 k a k c ) If required tune the structure READY FOR BRAZING!!! 27

28 Working Procedure Scheme Metric control Design Mechanical fabrication Measurements Lumped circuit model Electromagnetic design Lumped circuit model Bead pull Parameters evaluation Field uniformity Some measurements will be repeated after brazing Frequency tuning 28

29 The First Module At December 2008 high power RF tests performed at e2v (Chelmsford,UK) with a magnetron/modulator (MPT5839). showed that the module can easily stand a feeding power up to 4MW (much larger than the design figure of 2.8 MW). We worked at the requested nominal frequency of MHz. The high power RF testing lasted nearly 16 hours. The pulses length was regulated from 3 µs up to 7 µs. We rised the feeding power from 1MW to a repetition rate from 20 Hz to 120 Hz in order to test the module robustness. Good vacuum conditions and no remarkable spark events. Case of RF pulse length of 7 µs and a repetition rate of 20 Hz with 3 MW of peak power: the linac pulse shape was well inside the RF one. 29

30 The First Module First acceleration tests at LNS (Catania, Italy) with a magnetron (e2v - MG5193) and its modulator, made available by NRT (Aprilia, Italy), whose maximum deliverable peak power is 2.5 MW. To measure the energy of the accelerated protons we planned to use a nuclear detector (a 25 mm thick NaI(Tl) crystal) because of the very low beam intensity and the low duty cycle.. Particle dynamics simulations have shown that varying the RF power from 2MW to 2.8MW the maximum final energy changes of 2%. 30

31 The First Module One can notice that the bucket is not yet entirely formed since its likely that the particles did not experience a complete cycle of synchrotron oscillation. However, we may assume that the bunch in formation has a peak energy of 34.2 MeV. Simulation: the output beam energy distribution as function of input feeding power values Fraction MW 2.1 MW 2.8 MW E [MeV] 31

32 We had some problems during the first RF tests: spurious pulse was picked up by the detector preamplifier. The short time available at LNS for the tests was not sufficient to fix the problem We decided to change the detection system and we used radiocromic films in order to try to have evidence of the acceleration : Gafchromic film made of polyester sheets, 100 micron thick, and active layer in between. From the number of activated foils we can assume that the energy gain is between 3.25 and 3.5 MeV (lower limit). SATISFIED!!! It is our intention to repeat experiment adding the second module. 32

33 How difficult and heavy is to work manually!!! Thank you for your attention! 33

34 34