FNAL Optimization Update Laura Fields 11 February 2015 1 Outline - - PowerPoint PPT Presentation
FNAL Optimization Update Laura Fields 11 February 2015 1 Outline Results of three optimizations Cylindrical Target Parabolic Horns Reduced engineering constraints Further study of 3-horn optimized design w/ NuMI target
11 February 2015
Laura Fields
1
2
✤ Results of three optimizations ✤ Cylindrical Target ✤ Parabolic Horns ✤ Reduced engineering constraints ✤ Further study of 3-horn optimized design w/ NuMI target ✤ Effect of endcap material ✤ Beam simulation news
✤ Last time I showed early results of a cylindrical target optimization:
3
Best fitness: 1.98 Compared to 1.47 reference and 1.97 (NuMI-style target, optimized)
✤ After running a bit more (and having a lot of grid failures):
4
Best fitness: 2.01 Compared to 1.47 reference and 1.97 (NuMI-style target, optimized)
✤ NuMI-style target optimization, for comparison:
5
Best fitness: 1.97 Compared to 1.47 reference
6
Parameter Lower Limit Upper Limit Unit Horn A: LA 1000 4500 mm 3717 Horn A: F1A 1 99 % 51 Horn A: r1A 20 50 mm 33 Horn A: r2A 20 200 mm 147 Horn A rOCA 200 650 mm 630 Horn B: LB 2000 4500 mm 2551 Horn B: F1B 100 % 37 Horn B: F2B 100 % 12 Horn B: F3B 100 % 2 Horn B: F4B 100 % 16 Horn B: R1B 50 200 mm 186 Horn B: R2B 20 50 mm 47 Horn B: R3B 50 200 mm 179 Horn B: ROCB 200 650 mm 633 HornB: Z position 2000 17000 mm 5453 Horn C: LC 2000 4500 mm 2694 Horn C: F1C 100 % 30 Horn C: F2C 100 % 21 Horn C: F3C 100 % 2 Horn C: F4C 100 % 9 Horn C: R1C 50 550 mm 388 Horn C: R2C 20 50 mm 26 Horn C: R3C 50 550 mm 306 Horn C: ROCC 550 650 mm 620 Horn C: Z Position 4000 19000 mm 17836 Target Length 0.5 2.0 m 1.98 Beam spot size 1.6 2.5 mm 2.1 Target Radius 9 15 mm 7.8 Proton Energy 60 120 GeV 108 Horn Current 150 300 kA 270
7
Parameter Lower Lim Upper Lim Unit Final Optimum Horn A: LA 1000 4500 mm 2815 Horn A: F1A 1 99 % 65 Horn A: r1A 20 50 mm 34 Horn A: r2A 20 200 mm 145 Horn A rOCA 200 650 mm 630 Horn B: LB 2000 4500 mm 3229 Horn B: F1B 100 % 20 Horn B: F2B 100 % 21 Horn B: F3B 100 % 1 Horn B: F4B 100 % 22 Horn B: R1B 50 200 mm 191 Horn B: R2B 20 50 mm 47 Horn B: R3B 50 200 mm 204 Horn B: ROCB 200 650 mm 630 HornB: Z position 2000 17000 mm 3637 Horn C: LC 2000 4500 mm 2816 Horn C: F1C 100 % 36 Horn C: F2C 100 % 16 Horn C: F3C 100 % 3 Horn C: F4C 100 % 5 Horn C: R1C 50 550 mm 398 Horn C: R2C 20 50 mm 45 Horn C: R3C 50 550 mm 310 Horn C: ROCC 550 650 mm 643 Horn C: Z Position 4000 19000 mm 17478 Target Length 0.5 2.0 m 2.00 Beam spot size 1.6 2.5 mm 1.62 Target Fin Width 9 15 mm 13.4 Proton Energy 60 120 GeV 62 Horn Current 150 300 kA 296
Final Optimum Horn A Horn B Horn C
✤ Some conclusions ✤ Optimization run with cylindrical target gives slightly better results than
NuMI-style
✤ Difference in flux is small, and consistent with what I’ve seen before when
I place cylindrical and fin-style target in the same focusing system
✤ One substantial difference between the two optimizations ✤ NuMI-style target is limited by size of target can; cylindrical target is not
constrained and can grow larger
✤ Optimized focusing system is very similar for two options. Primary
differences (other than target):
✤ Cylindrical target prefers lower horn current and higher proton energy
8
✤ Last time I also showed early results from a parabolic horn
9
Best fitness: 1.85 Compared to 1.47 reference and 1.97 (NuMI-style target, conical horns)
✤ More recent results:
10
Best fitness: 1.89 Compared to 1.47 reference and 1.97 (NuMI-style target, conical horns)
It doesn’t look like this is going to do better than the conical horn option
✤ I’m also running an optimization with relaxed engineering
11
Best fitness: 1.90 Compared to 1.47 reference and 1.97 (NuMI-style target w/ engineering constraints, final optimum)
Allows longer target, longer target chase, larger conductor radii
12
✤ To do ✤ Debug grid failures which may be affecting speed
✤ Run an optimization with more realistic IC
✤ Tighter constraint on OD radius (due to nickel-
✤ Study optimized fluxes with updated sensitivity
✤ In past talks, when I’ve described the 3-horn optimized system
(NuMI-style target), people have been curious about the importance
✤ The neck radius was constrained to be small < 50 mm, primarily to
help the optimization converge a few months ago when separate problems were causing it not to converge
✤ In future runs, I can relax this constraint. But to understand it’s
impact, I did a scan of this parameter with a wider range than was used in the optimization
CP sensitivity is fairly flat vs. horn C neck radius Some small loss in sensitivity above r = ~120 mm
15
✤ For Horns A and B, I use 2 mm inner conductor and
✤ We know this is underestimating material ✤ Last November, I presented results of a study that
✤ Effect of increasing upstream neck of Horn 1 (in 2-
This made sense…
✤ Effect of 6 mm downstream endcap:
This was surprising. And indeed, when Paul and I poked further, there was a bug in this simulation…
✤ Effect of 45 mm downstream endcap:
~3% flux loss in peak with 4.5 cm endcap (note that this is much thicker than we expect any single endcap to be)
✤ Effect of 6 mm downstream endcap:
<1% flux loss in peak with 4.5 cm endcap (note that this is much thicker than we expect any single endcap to be)
✤ Still to-do: ✤ Study effect of gradually thickening endcap material
✤ For technical reasons, I have to remove the water
✤ Study endcap effect in 3-horn optimized system
✤ As you may have heard, our experiment is called DUNE ✤ A lot of computing stuff still refers to LBNE ✤ The machines known as lbnegpvm0X that many of us use for building and running gl4bnf
are being converted to dunegpvm0X
✤ /lbne/data/ and /lbne/app/ mounts have been renamed /dune/data/ and /dune
app/
✤ /pnfs/lbne appears to be available on the dungpvm’s, but you should probably start
using /pnfs/dune
✤ Our group name is ‘dune’ instead of ‘lbne’, which means that you can possibly not
modify/delete files created on the lbnegpvm’s
✤ I’ve updated the files in g4lbnf/ProductionScripts to deal with all of this ✤ You scan do a git pull to get the new scripts ✤ Let me know if you have problems (test on dunegpvm06-10)
✤ I think we are pretty close to being able to submit
✤ Will give us a lot more grid slots ✤ I’ve installed g4lbnf v3r4p2 in cvmfs (a
✤ Let me know if you are interested in helping push
23
24
25
LA F1A r1A rOCA r2A r1B r2B r3B F1B F2B F3B F4B r1C rOCB rOCC r2C r3C F1C F2C F3C F5C LB LC
✤ Last time I showed preliminary results of a three horn