geant4 applications to accelerator...
TRANSCRIPT
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
GEANT4 Applications to Accelerator Physics
V. Daniel [email protected]
February 21st 2002
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
• Accelerators
• G4 tools for beam physics (V. Daniel Elvira, P. Lebrun, P. Spentzouris, CPD/CD Fermilab)
• Applications
• Issues, Plans
Outline
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Magnets: dipoles, quadrupoles, solenoidsBasic elements of an accelerator to bend, focus, capture a beam of particles.
B
VB⊗
1 2
B
Horizontal Vertical deflection
Uniform field provides focusing in the horizontal direction but a deflection in the direction of the field would make the particle spiral away
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Weak Focusing
Cross section of a circular accelerator
B F=q V x B
rB
B 0y =
y
x
Focusing is weak because although there is also radial focusing, By gets weaker as a function of radius (gaps too large, and lots of iron !)
Particles above (below) midplane experience downward (upward) force vertical focusing
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Strong FocusingAlternating gradient focusing (1952) by Livingston at BNL Cosmotron (magnets in a ring)
Quadrupole lens: focus horizontally and defocus vertically or vice-versa
Geometric optics: equal strength convex and concave lenses can produce net focusing (if their distance < focal length)
yB
xB
yy
BBx
xB
B xyxx
yy ∂
∂=
∂∂
∂∂
=∂
∂=
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Solenoid Focusing
Large beam size (σx ~ σy ~1-10 cm), large angles
large aperture, thin lens approximation no longer valid: constx
xB
B yy ≠
∂∂
=
B
solenoid
B
Particle track
Bz=f(r,z), Br=g(r,z) Non-linear equations, horizontal and vertical motion not independent ( x-py & y-px correlations)
Characteristic feature of the Muon Collider/ Neutrino source design. Solenoids focus in two directions simultaneously.
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
The Solenoid Classclass Solenoid: public G4MagneticField {
int mynumptrxy, mynumptz;
double *mycoorrxy, *mycoorz;
double *mybrxy, *mybz;
vector<HepSpline1D> mysplz, mysplrxy;
vector<Sheet> sheetsInSol;
public:
Solenoid ( double minrxy, double maxrxy, int numptrxy, double minz, double maxz, int numptz, vector<Sheet> const &vsheets );
~Solenoid();
void GetFieldValue( const double Point[4], double *Bfield ) const;
}
251 cm
74 cm
Open Inventor
Defines a grid: r, z, Br , Bz
Spline fits of Bzand Br versus z for each r
Set of infinitely thin sheets of current
Linear interpolation of fits in r
Assembles the solenoid, builds grid and spline fits
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
class Sheet: public G4MagneticField {
int myid, mytype; float mythick, myrad, mylen, mycur; ThreeVector mylocation;
public:
Sheet ( ThreeVector location, int id, int type, float thick, float rad, float len, float cur ) :
mylocation(location), myid(id), mytype(type), mythick(thick), myrad(rad), mylen(len), mycur(cur)
{ ; }
void GetFieldValue ( const double Point[4], double *Bfield ) const;
virtual ~Sheet() { ; }
}; // Sheet
The Sheet Class
radius, length, and current
Analytic calculation of the magnetic field of the sheet
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Global Field and Magnet SolidsIn the DetectorConstruction user class we construct:
• a global field result of an array of magnets
• a logical volume for each type of magnet
• physical volumes to place the array
AGlobalEMField* ff = new AGlobalEMField();
fieldMgr->SetDetectorField(ff);
LogSol = new SolenoidG4LogicVol (Solenoid *sol, G4Material *MatCoil,...);
SolenoidG4PhysVol(G4RotationMatrix *pRot, const G4ThreeVector &position,SolenoidG4LogicVol *SolLV, G4VPhysicalVolume *pMother, double fscale, AGlobalEMField *ff )
Pointer to global field
Set global field
Construct logical volume for each type of magnet
Place each magnet and add it to the global field
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
AccelerationResonant cavities provide both acceleration and focusing in the longitudinal direction
L
R
Pillbox cavityπ 2πφs
L: cavity length, R: cavity radius Vp : peak voltage ν: frequency φs: synchronous phaseVs: synchronous V
Vp
( )t?2p fsinrc
?2pJEt)(r,E s00z +
==
Vp
2.405Rc
?2p=
Ez = 0 at r = R (ν fixes radius)
L is fixed by Vz, ν, phase advance
Fast particles get a smaller kick than slow particles (longitudinal focusing)
Vs
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
E-E
nom
inal
(MeV
)
Longitudinal phase space distribution of a “bunch” of particles
Bucket Separatrix
c∆T
Wmax
( )s
nomimax
fp?p2
c3Tc?
EW
−≈
∝
The separatrix divides the regions of stable (inside the bucket) and unstable motion
Bucket size depends on the beam energy, and the r.f. frequency and synchronous phase
Example: Enominal=200 MeV
L=32 cm, R=57 cm Vp =16.5 MV/m ν=201.25 MHz φs= 25.7o
Wmax ~ 60 MeV CT∼80 cm
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
The EMPillbox Classclass EMPillBox: public G4ElectromagneticField {
double frequency; double length; double eMaxGradient; double delay; double phaseAccel; double rMaxCavity;
public:
EMPillBox (double freqIn, double length, double eMaxGradIn, double phaseAccIn, double delayIn);
void GetFieldValue(const double Point[4], double *Bfield ) const;
}; // class EMPillBox
time delay for a cavity to match synchronous phase
synchronous phase
Open Inventor
class RFCavityG4LogicVolclass LinacG4PhysVol
Logical volume, place a linac, and add it to global field
Beryllium windows
Copper walls
We can also construct magnets and r.f. cavities from given field maps, using the Classes:
EMrfmap & MagFieldmap
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
The BeamLiouville’s theorem: density in phase space of non interactive particles in a conservative system is invariant
Phase space volume enclosed by the separatrix or any given surface is conserved (shape may change), if acceleration is performed adiabatically
Normalized Longitudinal Emittance: εL= (σE σCT)/mc2
Normalized Transverse Emittance: εx= (σx σpx)/mc2
Lorenz invariant
Beta Function, β(s) is proportional to beam envelope (amplitud) along its path. It is determined by the accelerator lattice
Examples: β(s) = 2pc/eBz (solenoid) oscillating (FODO)
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
The ν SourceNeutrino beam based on µ storage ring (1.7 km long):• 16-24 GeV proton driver ( 1-1.5 MW)
• Carbon target: p π + κ
• Capture, decay into µ, bunching
• Muon cooling ( εx~1 cm, factor~10)
• Linac + recirculating linacs
• µ Storage ring ( 50 GeV), gives 1020 neutrinos/year
Challenges:• target facility• cooling channel
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Muon Ionization Cooling
x x
z zP1
P2
θ1
θ2
θ1
θ2absorber
accelerator accelerator
absorberP3
θ1
+−= ⊥
Rµ3
trans
xx
LmEßß
fLe
dzde
Multiple scattering
Heating term (Mult.Scatt.)Cooling term +
lengthradiation:LBec2p
ß
dEdz
EßLv/cß
R
2trans
=
==
⊥
P2 < P1
All absorber classes inherit from the abstract class
AbsObj
Multiple Scattering is critical !
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Example: Long Solenoid Channel
Cooling cell~2.5 m long
126 cm
All channel2nd flip
On axis On axis
220 m long (87 cooling cells), two field flip regions (3T to –3T & -7T to 7T)
P (GeV) vs Z (cm)
Electric field acceleration
Energy loss in material
No multiple scattering, straggling, or delta rays
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Px versus x E versus CT( Final εx ~ 2 mm ) ( Final εL ~ 2 x Initial εL )
Published in proceedings of Particle Accelerator Conference, Chicago, June 2001 (Fermilab-Conf-01-182-T)
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Example: Helical Channel
Other simulations:
Open Inventor
zxy
72 m long solenoidal + dipole field with wedge absorbers and thin cavities
0zTyx, BBzL2p
sincos,BB =
=
• Alternate Solenoid Channel (sFoFo), published in proceedings of PAC2001 and Feasibility Study II for a Neutrino Factory at BNL (2001)
• Bent Solenoid Channel, presented at Emittance Exchange Workshop, BNL 2000
• Low Frequency r.f. Cooling Channel, presented at International Cooling Experiment Workship, CERN 2001
• Cooling Experiment (MICE) Simulation (in progress)
Published in proc. of PAC 2001 (Fermilab-Conf-01-182-T)
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
GEANT3 to GEANT4 MigrationAll simulations previously shown were performed with a customized version of GEANT4.2.0. Double precision version of GEANT3 had been used previously. Migration started with GEANT4.1.0…………
• Ekin did not change when electric field present, and time dependent electricfields not supported
• Multiple scattering model changed. Default parameters would give results different from G3. Free parameter left to the user to set with SetScatteringParameter(value).
• Stochastic processes were not eliminated completely by turning off multiple scattering and using SetEnlossFluc(false); delta rays would still be produced
• We were confused by the meaning and use of the accuracy parameters: delta_cord, delta_interaction, delta_one_step_value
Twenty library files modified to produce the CPD/CD Fermilab version of GEANT4.2.0. We also had our own DataCards Class. Therefore………………
• update to newer G4 versions painful
• official release of G4 beam tools among ~10 users is a delicate matter
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
G4.4.0 (with Jan patch) IssuesNow G4 supports time dependent electric fields. It is time to interface the beam physics tools to G4.4.0.
No more customization of the GEANT4 library
(cm)
(GeV/c)
To do:
• Test time dependent electric field.
• Investigate accuracy & performance
• Study new multiple scattering model (5 parameters set by the user)
• Replace our DataCard Class by the use of Messenger Classes
• Manipulate multiple scattering and delta rays from Main() using messengers
It works!
Ez = f(z) & constant with time
V. Daniel Elvira, FermilabG4 Users Meeting, February 21st , 2002
Summary & PlansG3 to G4 migration & beam physics tools development has been successful
Useful studies performed, published in proposals, conference proceedings. A group of people ( < 10) are using the beam physics tools
But we are still using an un-officially patched version of G4.2.0
• Work closely with G4 collaboration until the long solenoid example works with an officially patched version of G4.4
• Release the example and the beam tools to the neutrino source (and the G4) collaborations
• Write documentation on the use of the tools
• Offer support to the users