Cyclotron Magnets William Beeckman ECPM37 - Groningen October 28 2009 1 Presentation overview (1) Mechanics • • • Orbit in a uniform magnetic field – cyclotron frequency Consequence of mass increase Isochronous cyclotron Stability and Focusing • • • • • • • The need for extra focusing Edge focusing Spiral focusing The alternating gradient principle Betatronic oscillations No 2 sectors cyclotron Resonances 2 Presentation overview (1) The field shaping zoo • • steel modifications coils Types of steel Magnet design • • Basic parameters : C235 example Using 2d and 3d codes Coils … well, 1 hour is short and I needed a victim 3 TRIUMF 520 MeV H- 18m diameter, 4000 tons, 0.46T 4 PSI 590 MeV H+ separated sectors 5 Cyclone 10/5 and cyclone 3D 6 Mechanics 7 Acceleration - The Lorentz force (1) Force generated by an electric field E and a magnetic field on a particle of charge q and velocity v r r r r F = q (E + v × B) r r Felec = q E r r r Fmagn = q (v × B) Electric field along the trajectory supplies energy Force orthogonal to velocity and field Velocity tangent to trajectory No work done No energy supplied Accelerating gaps Electric field orthogonal to the trajectory does NOT supply energy Extraction deflector 8 Acceleration - The Lorentz force (2) Not all forces supply energy BUT All forces give acceleration Change in the velocity as a vector Force tangent to trajectory Tangential acceleration Force orthogonal to trajectory Centripetal (normal) acceleration r v2 r Fcentripetal = m n R Change in the modulus of velocity Change in the direction of velocity 9 Orbit in a uniform magnetic field The motion in the magnetic field is defined by the balance between the magnetic part of the Lorentz force and the centrifugal force v2 qvB=m r Bz V FL The magnetic rigidity Br is proportional to the momentum p=mv H- FC v p Br = m = q q ωr This can also be written Br = m q thus defining the cyclotron frequency qB ω= m (in radians/s) 10 Rigidity Br = 1 T 2 + 2E 0 T qc B(T), r(m), q(C), c(m/s), T(J), E0(J) From this formula, we can derive the 2 extreme cases Classical limit: T << 2E0 : T = B2r2q2/2m0 Ultrarelativistic limit: T >> E0 : T = Brqc 11 Consequences of the mass increase B qB q B = Cst = B0 then B = B0 γ ω = Cst if = ω= γ m m0 γ ω ⋅m 0 2π f rev E 0 B(T), m(kg), f(Hz), ω(radians/s), E(J) B0 = = 2 q qc Due to the increase of mass, the revolution frequency decreases as acceleration proceeds. For protons at rest: 1.6 10 -19 6 B(T) 15 . 24 10 B(T) f(Hz) = = - 27 2π 1.627 10 2 solutions • the RF frequency varies to follow the decreasing revolution frequency during acceleration : synchrocyclotron • the RF frequency is fixed throughout the whole energy range and the mass increase is compensated by increasing the magnetic field in the same proportion: isochronous cyclotron • The classical cyclotron can be viewed as limit of one of these 2 12 Isochronism (1) Isochronism (Iso=same, chronos=time) is achieved when the particle's revolution frequency is the same for all energies. The accelerating system frequency (or a multiple of it) is then made identical to this constant revolution frequency and, turn after turn, every time the particle shows up at the entrance of the accelerating gap, the voltage value is the same and acceleration takes place. f RF = h.f rev h is the harmonic number. If this condition is not met, the particle will show up in the gap too soon or too late, and will, under certain circumstances, finally experience a voltage of the wrong polarity (deceleration). Once the machine is isochronous, a particle that once was accelerated with a given dee voltage will keep receiving the same voltage throughout the whole acceleration. In other words, the particle doesn’t change its phase wrt the RF system. 13 Isochronism (2) If we compare that particle to one that is at another phase, the difference is in the energy they receive in each gap. The one that gets the largest acceleration will remain such all the time and thus make less turns in the machine. Its phase is kept constant by the fact that, having gained more energy, its path becomes longer, keeping the transit time between gaps constant. All accelerating phases (-89.99° to +89.99°) could in principle be used … with an infinite number of turns at the boundaries. In practice, a phase span of +/-30° about the maximum is used. All particles will reach the final energy but those at or close to +-30° will make twice more turns than the central particle. 14 Stability and focusing 15 The need for extra focusing Stability in the vertical (also called axial) plane requires that the field decreases with radius. This condition is naturally fulfilled in flat pole magnets since the field leaks outside the magnet. However, in the inside of such a magnet, the field index is very small and hardly ensure axial stability (weak focusing). Synchrocyclotrons are purposedely designed with decreasing fields along radius But, as the particle's mass increases during the acceleration process, one would like to increase the field with radius, in complete opposition with stability requirements. Extra axial stability must be provided in a different way 16 Edge focusing (1) Edge focusing (Thomas focusing) comes from the alternance of high field (hills or sectors) and low field (valleys) regions. Instead of being circular, like in a homogeneous field, the particle path is distorted in a polygon-like shape that depends on the number of hills encountered in 1 turn. The figure below shows an extreme deformation of the orbit going from circular to triangle (3 hills) and to square (4 hills). Of course real orbits are somewhere in between the circle and the polygon 17 Edge focusing (2) On such a polygon-like orbit, the velocity is no longer purely azimuthal but gets a radial component. If this radial velocity interacts with an azimuthal component of the magnetic field, we get a vertical component of the Lorentz force. Such an azimuthal component of the magnetic field exists in the fringing field produced at the interface between a hill and a valley, out of the median plane … which is precisely the place we need it Since both the radial velocity AND the azimuthal field change sign when going from the entrance to the exit edge, both edges have the same focusing effect. 18 Edge focusing (3) 19 Edge focusing (4) 20 Spiral focusing Spiral focusing plays on the other pair (velocity, field), the azimuthal velocity and the radial field. The azimuthal velocity is the main component of the velocity. The radial field is created by giving the hills a spiral shape. BUT, contrary to edge focusing that produces the same effect on both edges, the spiral focusing produces an alternance of focusing and defocusing effects since the azimuthal velocity doesn’t change sign from one edge to the other. 21 Alternating-gradient principle Optics states that an alternating series of focusing and defocusing lenses leads to an overall focusing provided the distance between the lenses is not too large and the focusing strenghs are not too different. This is the alternating-gradient principle. α β β focusing lens defocusing lens defocusing lens focusing lens α 22 Betatronic oscillations (1) The conditions for stability require a restoring force proportional to the displacement. This is the situation of the harmonic oscillator. The particles will perform oscillations around their equilibrium orbit. These oscillations are called betatronic oscillations because they were first observed in an electron accelerator, the betatron. There are 2 such oscillations, one in the horizontal plane and one in the vertical plane. As all harmonic motions, they have a frequency and an amplitude. While the horizontal amplitude may be rather large, the vertical one is restricted by the size of the magnetic gap. The notations for the frequencies are νr (QH) for the horizontal (radial) motion and νz (QV) for the vertical (axial) motion. They are not time inverse values but indicate the number of full oscillation(s) performed by the particle during one revolution. 23 Betatronic oscillations (2) 24 Betatronic oscillations (3) From the study of the different terms, we can for sure say that the focusing force varies with • the difference between hill and valley fields • the number of hills in one turn • the spiral angle These effects are taken into account in the following formula for the vertical betatronic frequency 2 N 2 ν z = n + 2 F (1 + 2 tan 2 (ξ )) N −1 N 2 3 2 2 2 ν r = 1 − n + 2 F (1 + 2 tan (ξ )) N − 1 N − 4 2 F= B − B B 2 2 is called the flutter and represents the hill to valley field difference 25 Betatronic oscillations (4) N2 2 N −1 in which N is the number of hills per turn, represents the fact that a triangular shape is more favourable to the existence of a radial velocity that a square shape, etc. (1 + 2 tan 2 (ξ )) ξ Takes into account • edge focusing • spiral focusing Remark that they are both multiplied by F is the spiral angle, i.e. the angle between the orbit and the tangent to the pole edge. A mean spiral takes into account the effect from both edges dB r n=− dr B is the field index and n = 1 - γ2 26 No 2 sectors cyclotron From the expression of the horizontal betatronic frequency, we see that if N<2, the flutter dependent term becomes very large and negative, making N 2 3 2 2 ν = 1 − n + 2 F (1 + 2 tan (ξ )) N − 1 N − 4 2 r negative !! Physically, this translates the change from bounded, oscillatory-type solutions to unlimited exponential ones. This is N called the π stop-band and it is reached every time we have ν r = 2 It implies that • N must be larger than 2 (lower limit of the π stop-band) N • there is an energy limit for every N value T = ( − 1)E 0 2 Protons can for example be accelerated to a maximum of 0.5*938 = 469 MeV for N=3 and to 938 MeV for N=4 27 Number of sectors N small low maximum energy distorted orbits – spaced lenses good focusing N large high maximum energy F small and quasi circular orbits -> spiral compulsory • 1 or 2 No way !! • 3 - large gap between poles, 3 dees, only harmonic modes=3, 6 if resonators coupled, more if independent • 4 - two valleys with RF cavities, all harmonic modes possible, two valleys for other devices • >4 - better for SSC 28 Resonances During the acceleration, QH and QV change. The plot of QV vs QH is the working point diagram. Like any oscillatory phenomenon, the amplitude of a betatronic motion can grow uncontrolled whenever an external source excites it with its own frequency. This resonance occurs as the betatronic frequency is a multiple of the "geometrical frequency" of the cyclotron. In this case, any kick given to the particle because of its particular position will be experienced again and again. These are known as systematic resonances. Under proper circumstances and frequency ratios, the 2 oscillators can be coupled and the energy stored in one motion, transferred to the other. These are coupling resonances. They are indicated by lines in the working point diagram, which the particle’s working point curve should avoid or cross as fast as possible. 29 Betatronic oscillations (6) Systematic resonances Working point curve Coupling resonances 30 The field shaping zoo 31 Shaping the field to the goal • Average field value <B(r)> is important • The requested field shape can be obtained by different methods or, more frequently, combination of methods • Warm coils machines: Iron dominated • SC machines: Coil dominated • … and everything in between • Constant gaps : B(r) naturally decreasing The larger the gap, the stronger the decrease Mid-60' synchrocyclotrons e.g. Orsay SC • Coil field : B(r) naturally increasing Important only when iron becomes saturated e.g. AGOR SC cyclotron 32 Shaping the iron contribution Pros • Very effective, even in superconducting machine • Simple and cheap • Very reliable Cons • Highly non linear • Demands accurate computation and machining • Fixed correction: no flexibility Model it before implementing it to avoid unexpected effects 33 The iron shaping methods zoo • Change the ratio of hill/valley spanned angle with radius e.g. "horns" in C10/5, spanned angle in C235 • Prevent field decrease at large radii e.g. Pole end chamfer in C18/9, valley inserts in C235 • Decrease the gap along radius e.g. Elliptical gap in C235 • Lateral edges milling e.g. C18/9 – C235 • Iron inserts e.g. flaps in C10/5 • Change local saturation e.g. trim rods in Chalk River synchrocyclotron 34 1. Hill spanned angle, C10/5 Horns 35 1. Hill spanned angle, C235 36 2. Pole end chamfer, C18/9 Chamfer 37 2. Valley inserts, C235 Valley steps 38 3. Elliptical gap between poles C235 39 4. Lateral edge milling, C18/9 Shimmed pole edge Fixed pole edge 40 4. Lateral edge milling C235 3 2 1 41 5. Movable iron inserts, C10/5 Flaps 42 5. Movable iron inserts, C18/9 Flaps 43 5. Movable iron inserts, C18/9 Flaps 44 5. Movable iron inserts, C18/9 H– measured magnetic field D Two measured magnetic field in C18/9 cyclotron Difference between magnetic fields provided by iron insert (flap) movement. Flap effect 45 6. Trimming rods, Chalk River SC 46 Shaping the coil contribution • • • • • • • • • Pros Great flexibility Multiparticles, multienergy machines Reliable to the level of the power supply Linear … but iron makes it not so linear Cons Very weak except in superconducting machines Trim coils increase the gap Demands accurate computation Insulation in high vacuum (H-), connection outputs Complex and expensive Model it before implementing it to avoid unexpected effects 47 The coil shaping methods zoo • Choose the coil position at nominal current e.g. all SC cyclotron • Coil subdivision e.g. VECC • Trim coils around poles e.g. C70, AGOR • Trim coils on poles e.g. MC40 • Harmonic coils for beam centering e.g. C235 48 2. Coil subdivision, VECC β coils α coils 49 3. Trim coils around poles, C70 <B> B0 B0γ(H-) B0γ(D-) <r> 50 3. Trim coils around poles, C70 51 3. Trim coils around poles, AGOR 52 4. Trim coil on poles, MC40 53 4. Trim coil on poles, MC40 54 5. Harmonic Coils, C235 55 Construction 56 Parameters for choice • • • • • • • • • • Cast Large parts, low deflection, poor mechanical properties Chemical composition and thermal treatment paramount Risk of defects, porosities, sand inclusions, etc Accuracy from machining, no accuracy loss from assembly Laminated Limited thickness : 300 mm max, usual 200 mm Good magnetic and mechanical properties Slight anisotropy Forged Very good magnetic and mechanical properties Expensive (casting + forging) Large parts possible 57 Assembly Many possibilities . From Annual Report of China Institute of Atomic Energy 2005 CYCIAE 100 MeV 58 Fully cast 59 Fully cast 60 Cast and assembled 61 Designing the magnet 62 Strategic decisions of design To decide • • • • • magnetic field level on poles, Bhill pole mechanical radius gap between poles valley gap and field, Bvalley number of sectors, N 63 Magnetic field on poles Determined by the type of accelerated ions negative ions, H- and D• low field then large pole radius (EM stripping) • easy, 100% extraction by stripping fully stripped positive ions • high field, small pole radius • complicated extraction system or self-extraction 64 Gap between poles Compromise between small gap • reduced number of At of coils • pole radius reduced, orbits close to outer edge • no space • very sensitive to errors : vertical losses large gap • large space: injection, extraction, probes • easier vacuum pumping • lower field 65 Valley gap Compromise between small gap • low flutter • axial focusing problems • spiralization of poles necessary • more field large gap • deep valley design 66 Well designed cyclotron magnet • • • • • • • Isochronous magnetic field during acceleration Axial (vertical) and radial focusing of beam Operation point far from resonances Fast passage through resonance(s) zone(s) Possibility to install all subsystems Efficient use of power Good cooling 67 Design steps • • • • • step0: Squeeze requirements and extract juice step1: Get starting numbers from hand calculation step2: 2d global model step3: 3d global model step4: 2d cuts for detailed local objects Steps 2, 3 and 4 are combined in an iterative process and 2d calculations must be preferred. 68 Fixed energy 235 MeV Non superconducting coils Isochronous Machine Dimensioning Magnetic rigidity Bρ = 2.35 T.m Maximum field limited by coil power efficiency B(r) = B0*γ(r) High B obtained using ELLIPTIC gap Bmax ~ 3T close to coil B min ~ B coil ∆Β = saturation magnetization 2.1 T Small gap and little axial focusing possible thanks to the small current (300 nAmps) required for treatment Azimuthal field modulation (flutter) Spiral shape of poles 2 2 N ν z = n + 2 −1 ⋅ F ⋅ (1+ 2 tan(ξ )) N 2 2 d B r n = − = 1−γ dr B For best RF efficiency, k=0.5 BUT to decrease machine dimensions k >0.5 (more hill, thus more field) CHOICE : k=0.67 (60° hills) <B> near coil is thus 0.67*3 + 0.33*(3-2.1) = 2.31 T 2 F= B − B γ(235 MeV) = 1.25 thus central field B0 ~2.31/1.25 =1.8 T Coil dimension, position and current to match these values, taking into account power consumption and cooling Return yoke dimensions function of Flux to transport Allowed iron saturation Radially focusing Axially defocusing Extra axial focusing needed At high radii, <B> = k*Bhill+ (1-k)*Bvalley k = Staking factor = hill angle/90° Radius close to coil is R~ 2.35/2.31 = 1.02m B increases with radius B 2 2 CHOICE : ν z = 0.2 Flutter and spiral not too large Field gradient can be strong Spiral angle of pole completely determined since n, N, F and ν z are known 69 Computing a dipole field (1) How can we compute the field in the air gap between the poles ? r r r ∂D r 4th Maxwell's equation ∇ × H − =J ∂t (Ampere's law) here: no charge => no displacement current. r r r r r r B .dl = ∫ J .ds In integral form ∫ H .dl = ∫ µ0 µ r r S r r r r B B r ∫iron µ 0 µ r .dl + air∫ µ 0 .1 .dl = ∫S J .ds In iron, µr >>1 and the whole contribution of the path in iron is negligible compared to the path in air (µr =1) B .h = nI where h is the gap height, I the current, µ0 and n the number of turns in the solenoid 70 Computing a dipole field (2) From this formula, we see that the field B = • • • µ 0 nI h is proportional to the total current in the solenoid, is inversely proportional to the magnetic gap and is independent on the pole surface, a rather counter-intuitive fact to most people. We also see that it is very important that the magnetic circuit is made out of very high relative permeability material. Since the total current is shared between the path in the air and the path in iron, it is important to reduce the contribution from the latter to make the best possible use of the solenoid current. 71 Permeability of ferromagnetics Material Initial relative permeability 0.002 Wb/m² Maximum relative permeability Steel (0.9% C) Iron (99.8%) Iron (99.95%) 78 Permalloy (78.5% Ni, 21.5% Fe) Superpermalloy Cobalt (99%) Nickel (99%) Mumetal (75% Ni, 2% Cr, 5% Cu , 18% Fe) 50 150 10,000 8,000 100,000 70 110 20,000 1000 5000 200,000 100,000 1,000,000 250 600 100,000 72 Permeability for a good steel (Poisson) Permeability (log scale) as a function of Induction permeability ( µ ) 10000.0 1000.0 100.0 10.0 1.0 0 0.5 1 1.5 2 2.5 B [T] 73 2D Modelling in early study The 3D geometry is modelled with a 2D code in axisymmetry using pseudo-materials The stacking factor is the proprtion of the circle occupied by the real material. Each pseudo-material is defined by a modified B-H curve B pseudo = µ H + k ⋅ (B − µ H ) 0 0 74 3D Modelling (1) 75 3D Modelling (1) 76 Field Vs Radius for Different Azimuths 77 Play every possible trick to use 2d instead of 3d 78 3D Modelling (2) 79 2D Modelling for details study Strong iron saturation allows to use axi-symmetrical models to compute the field along a radial cut. The 2d (axi) model is simply given the shape of the radial cut, with all its features. It allows to study details in a much more quick and efficient way. The technique is remarquably accurate but it requires an initial "calibration" of the models with angles to cope for different offsets. It was used for instance in the study of the gradient corrector in C230. This is a tiny slab of iron, a few mm thick with thickness variations of 0.1 mm in an overall structure with typical dimensions of a couple of meters 80 2D Modelling : Gradient Corrector 81 3d tips You all know this but it is worth repeating it anyway! 82 Meshing • Codes compute potentials and we want 1st and 2nd order derivatives for fields and tunes Prepare your model to have a proper mesh where you compute values, the rest of the model is subsidiary YOU must be in control of the mesh, not the code • A limited amount of quadratic elements is much more effective to accuracy than many linear elements 83 Betatron Frequencies 84 Boundaries • Is the rest of the universe far enough ? TEST IT! Don't be happy with default far field boundary conditions. Solution must be stable with normal far field conditions • Use symmetry boundary conditions where possible 85 Coils and steel • The farther from saturation you are, the more you have to care about correctly describing your steel (BH curve) • Correction coils and saturated steel won't necessarily behave the way you expect they would • Codes are very good in the computation of small changes between 2 models but less good at absolute values. 86 Thank you 87
© Copyright 2026 Paperzz