LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 1
Overview and relative bunch Overview and relative bunch population measurementspopulation measurements
Massimiliano Ferro-Luzzi, on behalf of the BCNWGMassimiliano Ferro-Luzzi, on behalf of the BCNWG
An overview– definitions, methods
what is a van der Meer scan (VDM),
what is the beam-gas imaging method (BGI)
– brief historical reminder of Bunch Current Normalization effort The relative bunch population measurements
– FBCT , vs others (BPTX, DCCT, LDM, Beam-Gas (BG) rates
– Effects on cross section
– Analysis of bunch-by-bunch cross-section data Conclusions
Thanks to the LHC groups and to the ALICE, ATLAS, CMS and LHCb collaborations for their support
Special thanks to Vladik Balagura for his detailed work on the bunch-by-bunch cross section analysis
Special thanks to Vladik Balagura for his detailed work on the bunch-by-bunch cross section analysis
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 2
What is luminosityWhat is luminosity
R = L
Rate Luminosity Cross section
L = f N1 N2 "the beam overlap"
revolution bunch population in
frequency beam1 and beam2
address with VDM scans or BGI, see later
address with VDM scans or BGI, see later
An absolute bunch charge measurement is needed
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 3
Van der Meer’s trickVan der Meer’s trick
Consider single circulating & colliding bunch pair with zero crossing angle
R = L = f N1 N2 1(x,y) 2(x,y) dx dy
With transverse displacements x , y of one beam w.r.t. the other:
R (x , y) = L(x , y) = f N1 N2 1(x-x , y-y) 2(x,y) dx dy
R (x , y) dx dy = f N1 N2 1(x-x , y-y) 2(x,y) dx dy dx dy
= f N1 N2 2(x,y) [ 1(x-x , y-y) dx dy ] dx dy
= f N1 N2 2(x,y) dx dy = f N1 N2
=1
z
=1
x
See refs. [1,2,3,4]See refs. [1,2,3,4]
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 4
residual gas
Beam-gas imaging methodBeam-gas imaging method
Again, luminosity
L = fN1N2 "overlap"
Beam interacts with residual gas around the interaction region
Reconstruct beam-gas interaction vertices
=> sample transverse beam profile
measure individually the 1 and 2 and rebuild the overlap
– even much better if combined with beam-beam interaction vertex distributions
– can measure also , bunch charges, hourglass effect, etc...
Strength with respect to van der Meer method:
(a) non disruptive, do not affect the beams !
(b) can run fully parasitically during physics running time
=> potentially smaller systematics uncertainties
Requires: (1)vtx detector resolution smaller (or at least comparable) to the beam sizes(2)residual pressure & acceptance must be adapted to this method
Requires: (1)vtx detector resolution smaller (or at least comparable) to the beam sizes(2)residual pressure & acceptance must be adapted to this method
pioneering measurement900 GeV, in 2009 [6]pioneering measurement900 GeV, in 2009 [6]
See refs [5,6,7]See refs [5,6,7]
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 5
A bit of historyA bit of history
20092009 1st lumi calib at Pt8, pioneering BGI at 900 GeV
20102010 1st flat top energy calibrations (VDM and BG) in Apr-May 2010:
BCN systematics dominate the cross section uncertainty ~ 10%
itself dominated by DCCT total current normalization
itself dominated by DCCT baseline uncertainty
started BCNWG, and a detailed analysis 4-6% Another lumi calib in Oct 2010
large currents (reduced baseline effects) down to 3-4 % And another in Nov 2010 (PbPb)
– Much more ghost/satellites !! Meanwhile, advanced solid ground work to tackle satellites/ghost
charge issues, in anticipation of DCCT uncertainty reduction
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 6
20112011
Large effort to understand BCN:– DCCT: BE-BI + Colin Barschel (CERN-PH, Aachen Univ.)
– Relative bunch populations, satellites, ghost: BCNWG
Three notes (being) written:– DCCT (Total current normalization): "RESULTS OF THE LHC DCCT CALIBRATION
STUDIES", CERN-ATS-Note-2012-026 PERF
– Relative populations: "STUDY OF THE RELATIVE LHC BUNCH POPULATIONS FOR
LUMINOSITY CALIBRATION", CERN-ATS-Note-2012-028 PERF
– Ghost and satellites: "STUDY OF THE LHC GHOST CHARGE AND SATELLITE
BUNCHES FOR LUMINOSITY CALIBRATION", CERN-ATS-Note-2012-029 PERF
The two main topics of this session :
I) The results of these BCN studies
II) The results of the expts' lumi calibrations
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 7
The scan seriesThe scan series
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 8
Beam and bunch populations Beam and bunch populations NN
Ntot = Nmain + Nghost ( + Npilots )
For each beam: (we suppress here the beam index j=1 or 2)
Nmain = Ni iM
sum over the set of all the main bunches (colliding or not)
total beam population (every charge in the ring)
NiPi = Nmain
population fractions: ( Pi = 1 ) iM
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 9
Main devices used in BCNMain devices used in BCN
Total current: Total current: DCCT
Individual bunch charge: Individual bunch charge:
Fast Bunch Current Transformers (FBCT)
IP1 button pickups (BPTX) Longitudinal Density Monitor (LDM) Beam-Gas rates in IP8
the main device:the main device:
but also:but also:
See the talk of
<= Colin Barschel
<= here + Jean-Jacques Gras (session 3)
<= Gabriel Anders
<= Adam Jeff
<= Jaap Panman
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 10
Definitions: slots, RF bins and all thatDefinitions: slots, RF bins and all that
25ns 25nsslot i+1slot i
responseRF bin
2.5ns
LHC = 3564 slots = 35640 RF bins (1 bin contains 1 RF bucket)
RF bin i10-9
LHCb beam-gas rate allows us to measure the "ghost" charge in the nominally non-filled slots (40 MHz electronics, integrating over 25ns...), but cannot discriminate in the nominally filled slots
FBCT integrates over ~24ns and applies a threshold (cut noise)
phase undefined ??
typical ghost/satellites in protonsand in Pb runs
... i-2 i-1 i i+1 i+2 ...
threshold
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 11
Definition of ghost and satellite charges Definition of ghost and satellite charges
Given that:– Total beam population is measured with the DCCTs
– The relative populations are measured (mainly) with the FBCTs, which which have a ~25ns integration time, while LHC RF buckets are 4 ~ 1ns
long, contained inside RF bins of 2.5ns
which have a 25ns slot threshold below which the slot is assigned 0 charge
– The main tool for measuring the particles invisible to the fBCT has been the LHCb BG rates, which also have a 25ns granularity
we choose to define...
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 12
... (among many possible definitions)... (among many possible definitions)
Ghost charge: charge in all slots which are not nominally filled (total charge below the FBCT threshold it is not "seen" by the FBCT )– can be estimated from LHCb beam-gas rate
caveat: trigger efficiency over 25 ns !
– can be estimated from LDM, <0.1ns time resolution ! caveat: normalization to nominal bunches! (emittance, position...)
definitions of phase relative to FBCT
Satellites: the charge in those neighbouring RF bins which are within 25ns "across" a nominally filled RF bin– they participate negligibly in luminosity (if crossing angle >0)
– are integrated in the "slot population" by FBCT what is the integration efficiency ?
what is the definition of phase (is the nominal bin centered ?)
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 13
The FBCTThe FBCT
The data for the relative bunch population measurements come from a ‘Slot Sum’ mode of acquisition with a specified number of 900 samplings for averaging (taken in 3600 turns with one sampling every fourth turn).
The Slot Sum returns 3564 bunch slot intensities on the high bandwidth channel each averaged over the specified number of turns.
The electronics were configured such that the gain was switched at around 2.3·1010p per bunch (averaged over all populated bunch slots)– 2011 pp scans = LO GAIN, while PbPb scans = HI GAIN
Block schematicssee Ref. [11]Block schematicssee Ref. [11]
thanks to D. Belohrad/ JJ Grasthanks to D. Belohrad/ JJ Gras
toroid trafo
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 14
FBCT subtletiesFBCT subtleties
FBCT uses two integrators (odd and even slots)
Gain was equalized by BE-BI experts
Possible systematic effects between odd/even channel not further investigated in BCNWG
Phase is regularly adjusted to have an optimum between good signal response and low "spill-over" effect
During VDM scans, this is generally done after reaching flat top (before STABLE BEAMS)
signal of the same bunch with the 2 integrators (skipping one clock cycle)
signals while change the fine phase delay
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 15
Typical example of FBCT measurements (Oct 2011)Typical example of FBCT measurements (Oct 2011)
fill 2234fill 2234
courtesy of V. Balaguracourtesy of V. Balagura
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 16
FBCT vs BPTX comparisonFBCT vs BPTX comparison
More of this in Gabriel Anders' presentation Compared also to:
– LDM => Adam Jeff
– BG rates => Jaap Panman
– DCCT , see next
BPTX is a completely different device, in a different location, with different electronicsBPTX is a completely different device, in a different location, with different electronics
fill 1653 mar 2011fill 1653 mar 2011 fill 2234 oct 2011fill 2234 oct 2011
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 17
Comparing FBCT sum vs DCCTComparing FBCT sum vs DCCT(Pb-Pb Nov 2010)(Pb-Pb Nov 2010)
Total current (DCCT) changes by a factor ~0.8 over the duration of the fill
How about the ratio R ?
R = FBCTsum / DCCT
R0 = R at time 0 (~arbitrary)
summing signals of all nominally filled slotssumming signals of all nominally filled slots
PbPb Nov 2010PbPb Nov 2010
sumFBCT(squares)
DCCT A and B(circles and triangles) 1%
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 18
Comparing FBCT sum vs DCCTComparing FBCT sum vs DCCT(Pb-Pb Nov 2011)(Pb-Pb Nov 2011)
PbPb Nov 2011PbPb Nov 2011
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 19
Comparing FBCT sum vs DCCT Comparing FBCT sum vs DCCT (pp 3.5TeV)(pp 3.5TeV)
pp May 2011pp May 2011 pp Oct 2011pp Oct 2011
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 20
Comparing FBCT sum vs DCCTComparing FBCT sum vs DCCT(pp 1.38TeV)(pp 1.38TeV)
pp Mar 2011pp Mar 2011
Klystron B2 trip
Here: more pronounced change of R .
Why ? Non-optimized longitudinal plane (RF) increasing ghost charge (see J. Panman) Beam2: RF klystron trip
Klystron B2 trip
1%
FBCT response depends on bunch length !FBCT response depends on bunch length !
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 21
Fill 1795, check on FBCT bunch length sensitivityFill 1795, check on FBCT bunch length sensitivity
Analyzed the start of ramp– strong change
of bunch length
– stable orbit Evaluate the
slope (sensitivity) of FBCT vs bunch length
energyenergyDCCTDCCT
DCCTDCCT
FBCTsumFBCTsum
FBCTsumFBCTsum
bunchlengthbunchlength
bunchlengthbunchlength
thanks to N. Bacchetta / JJ Grasthanks to N. Bacchetta / JJ Gras
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 22
FBCT response sensitivity on bunch lengthFBCT response sensitivity on bunch length
(dN/N)Slope:
dl
Slopes :
B1 ~ 0.37 % / 0.1ns
B2 ~ 0.48 % / 0.1ns
around nominal bunch intensity
Future: could be refined: make a bunch-by-bunch analysis
ll
thanks to N. Bacchetta / JJ Grasthanks to N. Bacchetta / JJ Gras
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 23
Typical bunch length spread during STABLE BEAMS Typical bunch length spread during STABLE BEAMS
Spread of ~ 0.02 ns (envelope) => expect ~0.1% max effect on FBCT-measured populations
NB: this effect was further reduced in the end-of-June 2011 technical stop by addition of a 70MHz filter
May 2011 VDM scans, fill 1783May 2011 VDM scans, fill 1783 thanks to N. Bacchetta / JJ Grasthanks to N. Bacchetta / JJ Gras
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 24
Position dependence of FBCT responsePosition dependence of FBCT response
Special MD (machine development): – 30 June 2011, fill 1910, 12 bunches, 450 GeV only
beam1 Y move
beam1 Y move
beam1 X movebeam1 X moveb
ea
m1
FB
CT
be
am
1 F
BC
T
be
am
2 F
BC
Tb
ea
m2
FB
CT
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 25
Worst beam/plane combination from this studyWorst beam/plane combination from this study
Beam 1 horizontal: FBCT slope (sensitivity) = ~ 1%/mm (at ~1e11 p)
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 26
Effect on lumi calibration experiments ?Effect on lumi calibration experiments ?
Minimal ! beams stay within 0.1mm at FBCT
during VDM scans
=> effect must be at most 0.1% NB: did not measure the individual
bunch positions at FBCT (only beam average)
Beam2 Y X Beam1 Y
X
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 27
Bunch-by-bunch cross section analysisBunch-by-bunch cross section analysis
Since Nov 2010, usually, (many) more than one colliding pair per experiment were available during VDM scans
=> each pair "i" gives one cross section measurement
Inspect the "consistency" between all bunch pairs If one observes an "inconsistency" (beyond statistics), one may try
to model and apply a correction
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 28
Response of bunch charge measuring deviceResponse of bunch charge measuring device
Ideal
Nmeasured
Ntrue
Nmeasured
Ntrue
Real ??
(overly exaggerated here)
still, can be considered locally linear (affine) to a good approximation
overestimation of population fractionoverestimation of population fraction
underestimation of population fractionunderestimation of population fraction
pure
pro
porti
onal
ity lin
e
pure
pro
porti
onal
ity lin
e
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 29
Simple fit modelSimple fit model
Define the "correct(ed)" populations:
Define the deviations from a local (arbitrary) point:
Assume errors are:
Because of , one of the parameters can be fixed:
=> Minimize:
"errors"measured
e.g.
set of all Main bunches (1...m)
set of all Colliding main bunches
1 par (), 3-par (12 ) or 5-par (121 2)
beam j=1,2i= index for BCID
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 30
Example: LHCb Mar 2011 (1.38 TeV)Example: LHCb Mar 2011 (1.38 TeV)
each point is one colliding bunch pair (IP8)each point is one colliding bunch pair (IP8)
PRELIMINARY
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 31
Other Mar 2011 scansOther Mar 2011 scans
PRELIMINARY
PRELIMINARY
PRELIMINARY
PRELIMINARY
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 32
ATLAS/CMS May 2011 (pp 3.5 TeV, 1.5m)ATLAS/CMS May 2011 (pp 3.5 TeV, 1.5m)
PRELIMINARY
PRELIMINARY
PRELIMINARY
PRELIMINARY
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 33
ALICE/LHCb May 2011 (pp 3.5 TeV)ALICE/LHCb May 2011 (pp 3.5 TeV)
PRELIMINARY
PRELIMINARY PRELIMINARY
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 34
Table of fit resultsTable of fit resultsthese are 1 by definition
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 35
ObservationsObservations
Affine fit generally improves the 2/ndf for pp, but it remains large for the higher-stat experiments (ATLAS, CMS) – improves especially in Mar 2011, not so in May 2011 ...nor Nov 2010
(PbPb, CMS)
Quadratic fit improves a further bit the 2/ndf for Mar 2011, but does nothing good for May 2011
For May 2011: have also a complete data set from BPTX– similar analysis done with BPTX data, similar observations
Cross section results are stable: change by less than 0.4%
Questions: How do the extracted fit parameters (1, 2 for the affine fit) compare
from IP to IP ? Especially: how do they compare between CMS and ATLAS in the
same fill ?
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 36
Comparing fit parameters...Comparing fit parameters...
"moderate" agreement – NB: only IP1 and IP5 use the same
bunch pairs!
But interesting observation on the difference FBCT - BPTX when compared with direct FBCT vs BPTX fit ...
shaded bands from direct FBCT vs BPTX fitshaded bands from direct FBCT vs BPTX fit
Hypothesis:slope - 1 FBCT - BPTX Hypothesis:slope - 1 FBCT - BPTX
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 37
Comparing IP1/IP5Comparing IP1/IP5
See a correlation between IP1/5 in both Mar and May 2011 In Mar 2011, the affine (and even quadratic) fits reduce correlation In May 2011, they don't
r = unweighted Pearson’s sample correlation coefficient
B1 (B2) bunch collides in IP1&5 and...noneIP8 (IP2)IP2 (IP8)
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 38
Bunch-by-bunch responseBunch-by-bunch response
Nmeasured
Ntrue
Remember: here, for illustration, the effects are disgustingly exaggerated
variations of the individual bunch properties can introduce changes ("fluctuations") in the response (bunch-by-bunch)
In our analysis, we did not (could not) take into account (and even less correct for) variations in bunch length, positions, and ...
This may (?) explain why we can't model perfectly with an affine fit
But the fluctuations could also be unrelated to the population fraction measurements!
bunch with different bunch length ?
bunch with different position at FBCT ?
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 39
ConclusionsConclusions
The FBCT provide accurate measurements of the relative bunch populations– however, scrutinizing the high precision data, we start seeing some
systematic effects at the permil level several small effects observed: dependence on bunch length, beam
positions, non-linearity, phase ? (satellites?), ...
Our detailed studies give us confidence that the effect on the cross section normalization are small (<0.5%, see table)– valid for the VDM scans from Nov 2010 onward
Given our current understanding*, we recommend using the "fudged" uncertainties (from the 1-par or 3-par b-by-b fits) as a measure of the systematic uncertainty on the cross section due to the relative bunch population measurements.
*we cannot explain convincingly the non-statistical fluctuations of the
bunch-by-bunch cross section data
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 40
Literature Literature (more here: http://lpc.web.cern.ch/lpc/lumicalib.htm)(more here: http://lpc.web.cern.ch/lpc/lumicalib.htm)
[1] "Calibration of the effective beam height in the ISR" , S. van der Meer, ISR-PO/68-31, 1968 (CERN). link
[2] "Measurement of the luminosity of the p-pbar collider with a (generalized) Van der Meer method", C. Rubbia, CERN p-pbar note 38. link
[3] "Absolute Luminosity from Machine Parameters", H. Burkhardt, P. Grafstrom, LHC-PROJECT-Report-1019 ; CERN-LHC-PROJECT-Report-1019; link
[4] "Notes on Van der Meer Scan for Absolute Luminosity Measurement", V. Balagura, Nucl. Instr. and Meth. A (2011), doi:10.1016/j.nima.2011.06.007. arXiv:1103.1129 [physics.ins-det] , see http://arxiv.org/abs/1103.1129
[5] "Proposal for an absolute luminosity determination in colliding beam experiments using vertex detection of beam-gas interactions", MFL, Nucl. Instrum. Methods Phys. Res., A 553 , 3 (2005) 388-399. link
[6] "Prompt production in pp collisions at s=0.9 TeV", LHCb Collab., R. Aiij et al., Physics Letters B Vol. 693, Issue 2, 27 Sep.2010, Pages 69–80.
[7] "Absolute luminosity measurements with the LHCb detector at the LHC", LHCb Collaboration, 2012 JINST 7 P01010 doi:10.1088/1748-0221/7/01/P01010. link arXiv.
[8] "LHC Bunch Current Normalisation for the October 2010 Luminosity Calibration Measurements" A. Alici et al. (BCNWG note2), CERN-ATS-Note-2011-016 PERF link
[9] "LHC Bunch Current Normalisation for the April-May 2010 Luminosity Calibration Measurements" G. Anders et al. (BCNWG note1), CERN-ATS-Note-2011-004 PERF link
[10] "LHC Lumi Days: LHC Workshop on LHC Luminosity Calibration", CERN-Proceedings-2011-001. link
[11] "The LHC Fast BCT system: A comparison of Design Parameters with Initial Performance" D. Belohrad, L.K. Jensen, O.R. Jones, M. Ludwig, J.J. Savioz, CERN-BE-2010-010, link
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 41
Backup slidesBackup slides
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 42
Error on cross section ought to be small!Error on cross section ought to be small!
Relative error on cross section due to wrong measurement of population fractions: correct
measured
depends on
- correlations between errors in one beam and population fractions in
the other beam, and
- in correlations between errors in population fraction in each beam
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 43
some (semi-quantitative) coarse boundsome (semi-quantitative) coarse bound
Define
Then:
Using as an indicative value, one can evaluate
In case then
unlikely and really bad
LHC Lumi Days 2012 29-Feb-2012 CERN Massimiliano Ferro-Luzzi 44
Other stuffOther stuff
Also tried to look at bunch families: no convincing structure observed
For example, May 2011
different "families" according to collision scheduledifferent "families" according to collision schedule
offset is related to slopeoffset is related to slope
group of 6 points: 6 VDM scansgroup of 6 points: 6 VDM scans
courtesy of V. Balaguracourtesy of V. Balagura