Giulia Papotti (BE-OP-LHC)
After fruitful discussions with:
M. Lamont, W. Herr,M. Ferro-Luzzi, G. Sterbini, G. Rumolo,
W. Venturini Delsolaro, T. Mertens
Chamonix 2011 LHC Performance Workshop, 27th Jan 2011
Luminosity Analysis
Chamonix 2011 - 2giulia papotti (BE/OP/LHC)
outline
• look only at protons, and only ATLAS + CMS • eases the analysis and allows double-checks
• theory reminders
• historical overviews from 25 bunches to bunch trains• data from end of July (fill 1251)
• luminosity / intensity / emittance lifetimes comparison– for bunch trains
• bunch-by-bunch (bbyb) analysis on a “clean” fill (1440)
• others– fill 1459: 50 ns spaced beams
– fill 1372: hump on and off
• Machine Development ideas
Chamonix 2011 - 3giulia papotti (BE/OP/LHC)
reminder
2221
22
21
210
2 vvhh
b fnNNL
• luminosity decreases with time– beam size
• elastic scattering at IPs and from residual gas, IBS
• noise in PCs, RF (phase and amplitude)
• LR beam-beam
• non-linear resonances
• hump, e-cloud
– intensity• due to interaction of the two beams at the IPs
– luminosity burn-off
– from beam-beam
• other causes – scattering on residual gas
– IBS and Touschek effect (mostly through emittance growth)
– overlap region from orbit drifts
Chamonix 2011 - 4giulia papotti (BE/OP/LHC)
2010 performance
• we have seen these many times
• but they only tell us that we were increasing nb
Chamonix 2011 - 5giulia papotti (BE/OP/LHC)
peak luminosity per collision
• given that rMPP limits the nb, how much Lpeak per bunch?
• luminosity burnoff• L/coll = 0.4 1030 cm-2s-1; 3 collisions / bunch; 100 mb x-section
– dN/dt = 1.2 105 p/s
• Total Losses, after 10 hours: 4.3e9 protons lost…
– 5% (~1 pilot) after 10 hours, or tburnoff ~ 227.5 h
peak luminosity / nb
[1030 cm-2s-1]
pre
pare
d b
y G
. Tra
d
Chamonix 2011 - 6giulia papotti (BE/OP/LHC)
statistics across fills
average bunch
intensity[1011 ppb]
average specific
luminosity[107 cm-2s-1]
calculated emittance
[mm]
fill number
5
2
0.8
1.2
pre
pare
d b
y G
. Tra
d
Chamonix 2011 - 7giulia papotti (BE/OP/LHC)
bunch-by-bunch losses
50 100 150 2005
6
7
8
9
10
11
time [min]
inte
nsity [
1e10 p
+/b
unch]
fill 1299 - beam 1
put in collisions
50 100 150 2005
6
7
8
9
10
11
time [min]
inte
nsity [
1e10 p
+/b
unch]
fill 1299 - beam 2
50 100 150 2000
5
10
15
20
25
30
time [min]
losses [
%]
fill 1299 - beam 1
50 100 150 2000
5
10
15
20
25
30
time [min]
losses [
%]
fill 1299 - beam 2
2h 2h
• take bbyb intensity starting before collisions– calculate % loss from
there• @ time=0: 0% losses
– colour code from head on collisions
• problems start at collisions– before, single beam
lifetimes are excellent
• next: cut 2 hours after collisions to compare fills
IPs: 1 5 2 8 - 1 5 8 - 1 5 2- 1 5 - 2 8 - 8 - 2
Chamonix 2011 - 8giulia papotti (BE/OP/LHC)
bbyb losses summaries – 1
1251 1253 1257 1258 1260 1262 1263 1266 1267 1268 1271 1283 1284 1285 1287 1295 1298 1299 1301 1303 1308 13090
5
10
15
20
25
30
fill number
losses 2
hrs
aft
er
coll.
[%]
• history of fills until 50 bunches/ring (fills longer than 2h)
– tune split until 1258• DQ1x,y= -0.0025; beam 1 always on a bad tune!
– 1260: tune split inverted
– 1263: tunes +0.002
– afterwards: no tune split and chroma corrected to ~1/2 units
– 48 bunches/ring: some bunches have very high losses• bunches with LR (33m) in IP 2 and 8
– then changed to 50 bunch scheme
25 48 50n bunches
be
am
1b
ea
m 2
IPs:1 5 2 8
1 5 81 5 2
1 52 8
82
Chamonix 2011 - 9giulia papotti (BE/OP/LHC)
bbyb losses summaries – 2
• history of 150 ns spacing fills + 50 ns fill (fills longer than 2h)
– 1364, 1366: Alice had wrong polarity
– 1393: b2 a mystery, high losses at collisions
– 50 ns fill (1459) losses out of scale (up to 50% for some bunches)
• nb: little statistics for “luminosity analysis”
– start with 11 fills longer than 8 hours, can use only 6• remove 1366 (wrong Alice pol), 1372 (with hump), 1373 (with hump), 1393
(strange losses), 1450 (RF module trips)
IPs:1 5 2 8
1 5 81 5 2
1 52 8
82
24 56 104 152 200 248 312 368 109
n bunches
Chamonix 2011 - 10giulia papotti (BE/OP/LHC)
• using:
• good agreement on whole fill
– luminosity decay is explained by emittance growth and intensity decay
– orbit stability is good• no overlap issues
lifetimes
tb1 tb2 th tv tL,calc tL,publ
81.3 103.3 44.9* 53.5* 15.911.9o
11.0x
106.1 86.12x30.6o
2x29.5x
2x36.6o
2x35.9x
19.6o
19.3x
18.3o
18.3x
0 100 200 300 400 500 6003.6
3.8
4
4.2
4.4
tota
l in
tensity [
101
3 p
pb]
time after stable beams [h]
0 100 200 300 400 500 6000
2
4
6
time after stable beams [h]
em
itta
nce [m
m]
0 100 200 300 400 500 600100
150
200
250
lum
inosity [
103
0cm
-2H
z]
time after stable beams [h]
beam 1
beam 2
ATLAS
CMS
xATL
yATL
xCMS
yCMS
firs
t 1
00
0 s
(M.
La
mo
nt)
wh
ole fill
vhbbL ttttt
11111
21
* BSRT, o ATLAS, x CMS, all t’s in hours
fill 1440
tota
l in
tensi
ty [
10
13
p]
Chamonix 2011 - 11giulia papotti (BE/OP/LHC)
luminosity lifetime summary
13661372 1373 13751397 14001408 1418 1427 143914401450 145314590
50
100
150
200
fill number
a [
103
2cm
-2H
z]
13661372 1373 13751397 14001408 1418 1427 143914401450 145314590
5
10
15
20
25
30
fill number
b [
h]
fit to y=ae-x/b
ATLAS
CMS
• compare different fills
– initial value (a) follows nb increase
– lifetimes (b) slowly descending
• very similar even removing first 1 h or 2 h
• 50 ns fill: not particularly good
Chamonix 2011 - 12giulia papotti (BE/OP/LHC)
intensity lifetime summary
• compare different fills
– initial value (a) follows nb
increase
– lifetimes (b) beam 2 often worse than beam 1
• very similar even removing first 1 h or 2 h
– lifetime lowest for 50 ns• later: 1 slide on losses
13661372 1373 13751397 14001408 1418 1427 143914401450 145314590
1
2
3
4
5
fill number
a [
101
3p/b
unch]
intensity, fit to y=ae-x/b
13661372 1373 13751397 14001408 1418 1427 143914401450 145314590
50
100
150
fill number
b [
h]
beam 1
beam 2
a [
10
13
p]
Chamonix 2011 - 13giulia papotti (BE/OP/LHC)
intensity lifetimes• single beam lifetimes
– not so easy to recalculate– squeezed + not colliding,
lasts <10 minutes vs tens of hours lifetimes
– most of the time: >100 h• can’t give a number!
• in collisions, add all causes:
INJECT RAMP SQUEEZE
FILL 1393
INJECTRAMP
SQUEEZECOLLIDE
COLLIDE
FILL 1381
1 h
1 h
125 h
125 h
beam 1 – beam 2
– calculated tburnoff ~230 h
– for tgas need a measurment– 100 h in Design Report
– losses on collimators, Mike predicts tcoll ~ 120 h• from D. Wollmann’s presentation in Evian, proton loss rate ~ 1.e8 p/s
– still missing: longitudinal losses, …
– but only tburnoff and tcoll give already tb1 ~ 80 h
• need controlled experiment: measure single beam lifetime
...1111
1
collgasburnoffb tttt
Chamonix 2011 - 14giulia papotti (BE/OP/LHC)
136613721373137513971400140814181427143914401450145314592
2.5
3
3.5
fill number
emittance from lumiregions
ah [m
m]
136613721373137513971400140814181427143914401450145314590
20
40
60
fill number
bh [
h]
fits to y=ae-x/b
136613721373137513971400140814181427143914401450145314592
2.5
3
3.5
fill number
av [m
m]
136613721373137513971400140814181427143914401450145314590
20
40
60
fill number
bv [
h]
CMS
ATLAS
growth summary• compare different fills, from lumi regions
– tv (bv) generally worse than th (bh)• v would be better than h if IBS dominated• hump probably• something else maybe? with
humpgood
fill
Chamonix 2011 - 15giulia papotti (BE/OP/LHC)
• emittance growth times from Design Report (7 TeV)– longitudinal: tIBS~ 63 h– horizontal: th,IBS~ 105 h
– from theory:• factor 2 from g• smaller than nominal
– scaling from DR: th,IBS~ 37.5 h
• from V. Lebedev th,IBS(2.5mm, 3.5 TeV) ~ 38.6 h
IBS
***0
1
lvh
ppbN
gt
136613721373137513971400140814181427143914401450145314592
2.5
3
3.5
fill number
emittance from lumiregions
ah [m
m]
136613721373137513971400140814181427143914401450145314590
20
40
60
fill number
bh [
h]
fits to y=ae-x/b
136613721373137513971400140814181427143914401450145314592
2.5
3
3.5
fill number
av [m
m]
136613721373137513971400140814181427143914401450145314590
20
40
60
fill number
bv [
h]
CMS
ATLAS
• IBS does not explain the whole blow up– noise? beam-beam? hump?
– need controlled experiment: estimate IBS at least
Chamonix 2011 - 16giulia papotti (BE/OP/LHC)
0 2 4 6 8 10100
120
140
160
180
200
220
time after stable beams [h]
lum
inosity [
103
0cm
-2H
z]
fill 1440 - fit to y=ae-x/b
R2=98.178%
b=18.3 h
a=196.4 h
ATLAS
fit
luminosity lifetime
• fill 1440, from ATLAS total instantaneous luminosity
• long fill and very clean data sets
– exponential fit: not very good fit
– overall: lifetime ~18.3 hours• first 30’: lifetime ~10.9 h• at around 2h: lifetime ~13.8 h• at around 6h: lifetime ~24.0 h• at around 10h: lifetime ~31.4 h
Chamonix 2011 - 17giulia papotti (BE/OP/LHC)
luminosity lifetime
0 5 10100
150
200
time after stable beams [h]
lum
inosity [
103
0cm
-2H
z]
fill 1440 - fit to y=ae-x/b
R2=98.178%
b=18.3 h
0 5 10100
150
200
time after stable beams [h]
lum
inosity [
103
0cm
-2H
z]
fit to y=ae-x/b+ce-x/d
R2=99.997%
b=3.0 h
d=28.7 h
0 5 100
50
100
150
200
fit to y=ae-x/b+ce-x/d
lum
inosity [
103
0cm
-2H
z]
time after stable beams [h]
a=44
c=164
• e.g.: from ATLAS total lumi
• much better with 2 exponentials instead of 1
– y = a e-x/b + c e-x/d
• a+c = Lpeak
• b, d time constants
• …can do the same on bbyb lumi data
– works also for single beam lifetimes
Chamonix 2011 - 18giulia papotti (BE/OP/LHC)
bbyb lumi lifetimes
0 1000 2000 30000.08
0.1
0.12
0.14
0.16
a [
103
0cm
-2H
z]
fill 1440, ATLAS
25ns slot
0 1000 2000 30002.6
2.8
3
3.2
3.4
3.6
b [
h]
25ns slot
0 1000 2000 30000.35
0.4
0.45
0.5
0.55
c [
103
0cm
-2H
z]
25ns slot
0 1000 2000 300024
26
28
30
32
d [
h]
25ns slot
0 1000 2000 300099.992
99.994
99.996
99.998
100
R2 [
%]
25ns slot
fit to y=ae-x/b+ce-x/d
• factor 10 between time constants
– fast component with effect from collisions schedule
• superpacman bunches
• fast+slow component also in other fills
IPs: 1 5 2 8 - 1 5 8 - 1 5 2 - 1 5
Chamonix 2011 - 19giulia papotti (BE/OP/LHC)
0 1000 2000 3000
0
0.5
1
1.5
a [
101
1ppb]
fill 1440, ATLAS
25ns slot
0 1000 2000 3000
0
50
100
150
b [
h]
25ns slot
0 1000 2000 3000
0
0.5
1
1.5
c [
101
1ppb]
25ns slot
0 1000 2000 3000-500
0
500
d [
h]
25ns slot
0 1000 2000 300099.85
99.9
99.95
100
R2 [
%]
25ns slot
fit to y=ae-x/b+ce-x/d
intensity lifetimes in collisions
• need the 2-exp model– single exp not good
• much more variation– no simple correlation
with collision schedule
– some correlation with injection pattern
• a lot of variation across different fills
IPs: 1 5 2 8 - 1 5 8 - 1 5 2 - 1 5
Chamonix 2011 - 20giulia papotti (BE/OP/LHC)
bbyb growth – 1
• bbyb luminosities from experiments
• combine with fBCT to get specific luminosities
• unfold emittance
– assumes very small orbit drifts• verified in few occasions with end-of-fill lumi scans
0 200 400 600
0.1
0.2
0.3
0.4
0.5
lum
i per
bunch [
103
0cm
-2H
z]
time after StableBeams [min]
fill 1440, CMS
0 200 400 6001.5
2
2.5
3
3.5
4
4.5
5
specific
lum
i [1
07cm
-2H
z]
time after StableBeams [min]
Nbu
= 368 bu/ring
0 200 400 6001
2
3
4
5
6
calc
ula
ted e
mitta
nce [m
m]
time after StableBeams [min]
growth = 4%/h (0.09 mm/h)
mean = 3.35
mean = 2.31
IPs: 1 5 2 8 – 1 5 8 – 1 5 2 – 1 5
Chamonix 2011 - 21giulia papotti (BE/OP/LHC)
bbyb growth – 2• ATLAS vs CMS: good agreement on average emittance
– use end-of-fill lumi scans for emittance cross-check– quite some spread on bbyb luminosity, from experiments
0 200 400 600
0.1
0.2
0.3
0.4
0.5
lum
i per
bunch [
103
0cm
-2H
z]
time after StableBeams [min]
fill 1440, CMS
0 200 400 6001.5
2
2.5
3
3.5
4
4.5
5
specific
lum
i [1
07cm
-2H
z]
time after StableBeams [min]
Nbu
= 368 bu/ring
0 200 400 6001
2
3
4
5
6
calc
ula
ted e
mitta
nce [m
m]
time after StableBeams [min]
growth = 4%/h (0.09 mm/h)
mean = 3.35
mean = 2.31
0 200 400 600
0.1
0.2
0.3
0.4
0.5
lum
i per
bunch [
103
0cm
-2H
z]
time after StableBeams [min]
fill 1440, ATLAS
0 200 400 6001.5
2
2.5
3
3.5
4
4.5
5
specific
lum
i [1
07cm
-2H
z]
time after StableBeams [min]
Nbu
= 368 bu/ring
0 200 400 6001
2
3
4
5
6
calc
ula
ted e
mitta
nce [m
m]
time after StableBeams [min]
growth = 4%/h (0.09 mm/h)
mean = 3.30
mean = 2.26
AT
LA
SC
MS
IPs: 1 5 2 8 – 1 5 8 – 1 5 2 – 1 5
Chamonix 2011 - 22giulia papotti (BE/OP/LHC)
estimation vs lumireg
0 100 200 300 400 500 6001
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
time after StableBeams [min]
em
itta
nce [m
m]
fill 1440, ATLAS, fit to y=ae-x/b+ce-x/d
lumireg: a=2.9mm b=-44.5h c=-0.5mm d=1.9h R2=90.0%
lumi+fBCT: a=2.7mm b=-55.3h c=-0.5mm d=3.3h R2=100.0%
0 100 200 300 400 500 6001
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
time after StableBeams [min]
em
itta
nce [m
m]
fill 1440, CMS, fit to y=ae-x/b+ce-x/d
lumireg: a=2.9mm b=-36.1h c=-0.5mm d=1.6h R2=63.9%
lumi+fBCT: a=2.8mm b=-61.2h c=-0.5mm d=4.1h R2=100.0%
• compare
– emittance calculated from luminosity + fBCT
– luminous region (average)
• good agreement!
Chamonix 2011 - 23giulia papotti (BE/OP/LHC)
50 ns spaced beam
50 100 150 200 250 3000
5
10
15
20
25
30
time [min]
losses [
%]
fill 1459 - beam 1 - 108 bu/ring
50 100 150 200 250 3000
5
10
15
20
25
30
time [min]
losses [
%]
fill 1459 - beam 2
50 100 150 200 250 3005
6
7
8
9
10
11
time [min]
inte
nsity [
1e10 p
+/b
unch]
fill 1459 - beam 1 - 108 bu/ring
50 100 150 200 250 3005
6
7
8
9
10
11
time [min]
inte
nsity [
1e10 p
+/b
unch]
fill 1459 - beam 2
• fill 1459
• 9 trains of 12– summer to winter time change in the
middle of the fill, some problems with data extraction
• very sudden losses for some bunches– mid-train bunches
• seem due to initial parameters
– had no extra LR
– e-cloud?
• no statistics: only 1 fill!
IPs: 1 5 2 8 - 1 5 8 - 1 5 2 - 1 5 - 2 8 - 8 - 2
tLumi [h] th [h] tv [h]
ATLAS 13.2 153.9 19.7
CMS 12.1 25.9 15.5
Chamonix 2011 - 24giulia papotti (BE/OP/LHC)
a fill with the hump – 1
0 5 10 150
5
10
15
20
25
30
35
40
inst
lum
inosity [
1030cm
-2H
z]
time after stable beams [h]
fill 1372
0 5 10 150
2
4
6
8
10
time after stable beams [h]
em
itta
nce [m
m]
CMS
ATLAS
x CMS
y CMS
x ATLAS
y ATLAS
• fill 1372
• particularly bad vertical blow up– emittances from lumi region
• clear correlation with instantaneous luminosity slope changes
Chamonix 2011 - 25giulia papotti (BE/OP/LHC)
0 5 10 1510
15
20
25
30
35
40
45
inst
lum
inosity [
103
0cm
-2H
z]
time after stable beams [h]
ATLAS published
calc(N1, N
2,
x,
y)
calc(N1, N
2,
x,
x)
calc(N1, N
1,
x,
x)
12
3
4
0 5 10 150
10
20
30
40
50
inst
lum
inosity [
103
0cm
-2H
z]
time after stable beams [h]
CMS published
calc(N1, N
2,
x,
y)
calc(N1, N
2,
x,
x)
calc(N1, N
1,
x,
x)
12
3
4
a fill with the hump – 2• difference in integrated luminosity with / without extra V blow up
• 3 calculated curves– with h,v from luminous regions (ATLAS or CMS)
– N1, N2 average intensity beam 1 and beam 2
• integral(3) - integral(2) = 25% (ATLAS) / 27% (CMS)
• integral(3) - integral(1) = 15% (ATLAS) / 18% (CMS)
• 20% integrated luminosity lost due to the hump!
Chamonix 2011 - 26giulia papotti (BE/OP/LHC)
used and requests
• data used for this analysis– machine measurements (always available)
• fBCT: bbyb intensities• single beam lifetime
– from experiments (only for stable beams, in LPC folders)• single bunch luminosities• average luminous region size
– looking forward to BSRT bbyb emittance
• requests for Machine Developments– “hands off” squeezed beams for machine time constants
• few hours for tens of hours time constants?
• and IBS studies?– if after a test ramp… only net time for the measurement
– possibly on a day with no hump
– tune scans! • working point almost never optimized so far
– end-of-fill VdM scans for emittance measurement, bbyb emittance and orbit differences
Chamonix 2011 - 27giulia papotti (BE/OP/LHC)
conclusions• 150 ns spaced beam lifetimes
– luminosity lifetime (~20 h)• decay explained by emittance growth and intensity decay• fast and slow component
– fast component dependence on collision pattern
– intensity lifetime (~90 h)• single beam lifetime is excellent, need a hard number• drops after collisions
– dependence on collision pattern
– emittance grows (x: ~ 30 h; y: ~ 20-40 h)• minimum overlap/orbit drifts
– also verified with end-of-fill lumi scans
• need to look at causes– probably not only IBS
• many fills have different characteristics• fill 1459: with 50 ns spacing, very fast losses• fill 1372: hump caused loss of 20% of integrated lumi
• nb: results are preliminary due to the lack of statistics