A.Poklonskiy (SPbSU, MSU), D.Neuffer (FNAL) C.Johnstone (FNAL), M.Berz (MSU), K.Makino (MSU)

20 December 2004 Taylor Models Workhop

Exploring and optimizing Adiabatic Buncher and Phase Rotator for Neutrino Factory in COSY Infinity

A.Poklonskiy (SPbSU, MSU), D.Neuffer (FNAL)C.Johnstone (FNAL), M.Berz (MSU), K.Makino

(MSU)

20 December 2004 Taylor Models Workshop

Goal & Problems

Goal:• Build neutrino factory to study different neutrino-

related things and/or muon collider… to collideProblems:

• Muons are short-living particles compact lattice, fast beam gymnastics

• Muons are produced with large initial momentum spread cooling

• Energy spread is large energy spread reduction before cooling

• Some desired beam manipulations requires new types of field configuration development of such new elements

• All these small muons production rate (<0.2) and…PRICE!


R&D goal: “affordable” e, -Factory

• Improve from baseline: – Collection

• Induction Linac “high-frequency” buncher

– Cooling• Linear Cooling Ring

Coolers

– Acceleration• RLA “non-scaling FFAG”+ e+ + +

e

– e– + e + and/or

The Neutrino Factory and Muon Collider Collaboration

http://www.cap.bnl.gov/mumu/mu_home_page.html


RF Cavity and Solenoid in Pictures


Adiabatic buncher + () Rotator (David Neuffer)

• Drift (90m) decay, beam develops correlation

• Buncher (60m) (~333Mhz200MHz, 04.8MV/m) – Forms beam into string of bunches

Rotator (~12m) (~200MHz, 10 MV/m)– Lines bunches into equal energies

• Cooler (~50m long) (~200 MHz)– Fixed frequency transverse cooling system

Replaces Induction Linacs with medium-frequency RF (~200MHz)


Longitudinal Motion (2D simulations)

Drift Buncher

(E) rotator Cooler

System would capture both signs (+, -)


Key Parameters

• Drift

– Length LD

• Buncher

– Length LB

– RF Gradients EB

– Final RF frequency RF (LD, LB, RF: (LD + LB) (1/) = RF)

• Phase Rotator

– Length LR

– Vernier offset, spacing NR, V

– RF gradients ER


Lattice Variations

– Shorter bunch trains (for ring cooler, more ’s lost)?

– Longer bunch trains (more ’s survived)?

– Different final frequencies? (200,88,44Mhz)

– Number of different RF frequencies and gradients in buncher and rotator (60…10)?

– Different central energies (200MeV, 280MeV, optimal)?

– Matching into cooling channel, accelerator

– Transverse focusing (150m B=1.25T solenoidal field or…)?

– Mixed buncher-rotator? – Cost/perfomance optimum?

OPTIMIZATION IS NEEDED


COSY Infinity Simulations

COSY Infinity code (M. Berz, K. Makino, et al.)

Where M – map of the equations of motion (flow), obtained as a set of DA – vectors (Taylor expansions of final coordinates in terms of initial coordinates) uses DA methods to compute maps to arbitrary order own programming language allows complicated optimization scenarios writing internal optimization routines and interface to add more provides DA framework which could significantly simplify use of gradient optimization methods model is simple now, but much more complicated in future and COSY has large library of standard lattice elements

0

0

0

000

00

,1

)(,,,,E

EEE

vttl

p

pby

p

paxZ yx

)),()(,()( 00 sZssMsZ ),(),(),( 021021 ssMssMssM


Big Problem

Use of Taylor series leads to tricky way of handling beams with large coordinate spread (and that is exactly the case) Relative coordinates should be < 0.5 (empirical fact)


Straightforward division


Equations of Longitudinal Motion

)sin()sin(

)sin(

))(sin()sin(

)sin(

2

2

2

2

ss

RF

ss

RF

s

s

RFss

RFz

sRFz

mv

qV

dt

d

v

zm

qV

dt

zd

zzz

zqVvztqV

dt

dp

qVdt

dp

nonlinear oscillator

synchronous particle

equations in deviations from synchronous particle

COSY Infinity uses similar coordinates


Consequences of Equations of Motion

• Existence of stable regions, where we have oscillatory motion and unstable regions, and, therefore existence of separatrix (depends on frequency, RF gradient, synch. phase, etc… ). Stable area is called the “bucket”.


Beam Evolution


Clever Division• Use central

energies as centres of boxes, use RF period as ranges for the box

• Add “jumping” between intervals after each step. We change buckets and particles could be lost in one bucket and re-captured in another


Still a Problem

• 50 central energies x 60 RF cavities in Buncher = 300 maps…

• We are using DA arithmetic, everything is a DA-vector, including elementary functions (sin), so we need relatively high expansion order. Use 2 times more intervals + 5th order leads to natural advantage in buncher… maybe.

• Use COSY’s ability to generate parameter-dependent maps with ease and special law of bunches central energies distribution (smaller energies tends to be closer to each other)

40-50 maps 12-15 maps Potential calculation time reduction. Implementing.• 6000 particles, 50 central energies, 70 RFs

– 1st order: 0 hrs 10 min– 7th order: 8 hrs some mins

OPTIMIZATION?

nT

1


Sin Taylor expansion

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

3rd order

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

5th order

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

7th order

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

9th order

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

11th order

-2-1.5

-1-0.5

0 0.5

1 1.5

2

-3 -2 -1 0 1 2 3

13th order


Different Order Simualtions


Conclusions

• Model is implemented in COSY Infinity and checked for consistency with other codes

• Some removal of the obstacles is done

• Brute-force optimization still seems to be infeasible. Looking for some more sophisticated method.


Yet Unanswered Questions

• Should longitudinal motion be studied separately, or should it be included on the very early stages?

• Are there any map-dependent criterias which could be used for map-based optimization?


Second Optimization Approach

• From synchronism condition one could derive following relation for kinetic energies of synch particles:

nTn n

cn

,111

1

111

1)(

20

cc

c

n

WnT


Final Kinetic Energy Relation

– From the rotator concept one could derive amount of energy gained by n-th synchronous particle in RF

– So for final energy n-th particle has after the rotator consists of m RFs we have

)2sin()( nN

EnT RF

)()(),( nTmnTmnT


Evolution of central energies shape T(n,m)


Energies shape in buncher and amount of kick they get in rotator


Energy Shape Evolution in Rotator


Objective Functions

– The idea of the whole structure is to reduce overall beam energy spread and to put particles energies around some central energy. So we have general objective function:

– First, we can set and get

2)),(( cn TmnTcI

21 )),(( cTmnTI

nCn ,1


Different optimized paremters (n vs T_fin)


Different optimized paremters (T_0 vs T_fin)


Evolution of central energies shape (unoptimized)


Evolution of central energies shape (optimized)


Different optimized paremters (T_0 vs T_fin) + energies distribution


Objective Functions

– For calculating we can use particle’s energies distribution

22 )),(( cn TmnTcI

nc

n energy particles %---------------------------------------------- -12 963.96 1023 17.050000 -11 510.85 692 11.533333 -10 374.64 537 8.950000 -9 302.98 412 6.866667 …


Different optimized paremters (n vs T_fin)


Summary

Model of buncher and phase rotator was written in COSY Infinity

Simulations of particle dynamics in lattice with different orders and different initial distributions were performed

Comparisons with previous simulations (David Neuffer’s code, ICOOL, others) shows good agreement

Several variations of lattice parameters were studied

Model of lattice optimization using control theory is proposed

Model of central energies distribution is developed. Some results for parameters have been obtained


Future Plans

Finish central energies optimizations, try changing more parameters, check optimized parameters for whole distribution (COSY, ICOOL?)

Develop some criteria for simultaneous optimization of central energies and energies of all paritcles in a beam or use control theory approach for the whole longitudinal motion optimization

Study transverse motion and particles loss because of decay and aperture, final emittance cut

Different lattices for different cooling sections/targets/whatever proposed (project is on R&D status)

A.Poklonskiy (SPbSU, MSU), D.Neuffer (FNAL) C.Johnstone (FNAL), M.Berz (MSU), K.Makino (MSU)

Documents

Transcript of A.Poklonskiy (SPbSU, MSU), D.Neuffer (FNAL) C.Johnstone (FNAL), M.Berz (MSU), K.Makino (MSU)