1 Beam e ’s from antineutrinos – Update – David Jaffe, Pedro Ochoa November 13 th 2006 Part 1: from + reweighing Part 2: New ideas.

1

Beam e’s from antineutrinos – Update –

David Jaffe, Pedro Ochoa

November 13th 2006

Part 1: from + reweighing Part 2: New ideas

2

Need to tag antineutrinos coming from + decay:

One of the backgrounds in e analysis: intrinsic beam e‘s

E (GeV)The technique:),(),,()( KK v

datavv

Need high purity at low E

This is what we are trying to measure

Very little contribution from µ+ above this energy (Ecut)

Ecut

True energy of true at the ND

Nearly all come from +→ e+ + e +

Reminder

3

Suggested in last collaboration meeting.

Used carrot and thus required mupi trees (thanks Chris!)

from + reweighting

from +

raw MCreweighed MC

reweighed MCraw MC

Used SKZP “a la Boston” to reweigh the + and K+ parents of the +:

Raw MC Reweighed MC

#events 455.3 472.4

(1.93x1019 POT)

4

pz

pt

Why so little change?

pz

pt

Plotted +,+ weights as a function of pt, pz to make sure no error:

The + parents get weights very close to 1:

parents(# events)

+ parent type(+ ~ 96%)

5

Current status (see minos-doc 2218)

Main idea of scaling methods (cf. minos-doc 1971) is:

),()(),( KECK MCvv

),(),,()( KK vdatavv

No reweighting applied to the MC

Overall technique:

Main idea of fit method is:

),(),( , KK FITMCvv

Scale method 1: C(E) from horn-off data/MC ratio, Ecut < E < Ehigh

Scale method 2: C(E) from horn-off data/MC ratio, Elow < E < Ecut

Stan’s method: C(E) from horn-off data/MC ratio, all E

Scale method 4: C(E) from horn-on data/MC ratio, E > Ecut

Scale method 5 (retired): C(E) from horn-on data/MC ratio, all E

Results in next slide were obtained with Ecut = 10 GeV, Elow = 4 GeV and Ehigh = 16 GeV

6

Current status (see minos-doc 2218)

from + decay

E < Ecut

data-(Fit or Scaled) MC, Ecut < E < 30 GeV

raw MC 375.8 ± 15.1 (stat) 72.8 ± 6.5 (stat)

reweighed MC 373.4 ± 15.1 (stat) 99.1 ± 9.1 (stat)

Scale method 1 1015.6 ± 130.6 (stat) -1636.9

Scale method 2 1001.8 ± 130.7 (stat) -1655.6

Stan’s method 654.8 ± 289.5 (stat) -257.2 ± 298.8 (stat)

Scale method 4 1640.7 ± 126.6 (stat) 132.4± 122.2 (stat)

Fit method 546.4 ± 131.8 (stat) -21.4 ± 124.1 (stat)

“Scale method 5” was removed. See first two backup slides for more details.

Fit method needs to be revisited: SKZP “a la Boston” not very appropriate for

antineutrinos since not much variation in pt,pz space. Considerable fraction of antineutrinos not produced in

target (cf. minos-docs 2042 and 2376)

Should be real nubars from + if

data/MC from horn-off is trust-

worthy in this region

Should be ~0 by construction

Should be real nubars from +

Expected to be highly

negative by construction

Note: le010z185i data POT=1.93x1019

le010z000i data POT=2.77x1018

7

New ideas

How about using the pHE data? Antineutrinos from + are the only ones affected by focusing (?) Can do pHE-LE and extract the two + components that way (?)

KKL

+

Plots scaled to 1.0x1020 POT

All plots until slide 10 are true E of true

antineutrinos.

All available stats for

pHE

LE pME

pHE

8

But also significant differences in the other components:

from -,K-:

LEpHE

LE/pHE ratio

from +

LEMEpHE

Indeed + component is considerably affected by focusing:

from -,K-

9

Where are the -,K- differences coming from?

Plots made by A. Himmel from Caltech

(See backup slide on antineutrino provenance for more information)

10

LE/pHE ratio for plots in previous slide:

Plots made by A. Himmel from Caltech

Note: error bars are probably wrong

11

What about using the pME data?

from -,K-

LEpME

Antineutrinos from -,K- are almost identical in LE and pME !

Checked that nubar-PID selection does as good in pME as in LE:

For now neglecting ~0.3% difference in purity between

LE and pME

nubar-PID in pME

all

NC

vv

Selected events at 1.9x1019 POT

from -,K-

pME - LE

12

from -,K-

LEpME

(reweighed)

Checked with SKZP reweighing, just in case:

Selected events at 1.9x1019 POT

Idea is to take (pME-LE) data difference and fit with MC shapes using two scaling parameters “parLE” and “parME”:

from +

pME

parME

from +

LE

parLE

pME-LEFit

from -,K-

pME – LE

13

How well could this work? Use fitted shapes instead of histograms:

from -,K- from -,K-

from + from +

pME

pME LE

LE

Selected events at 1.0x1018 POT

14

Assume: infinite MC statistics (pME and LE)infinite LE data statistics

Create fake pME data set for 1e18 POT by fluctuating smooth histograms with Poisson stats. For example:

fluct

fluct

from -,K-

from +

pME

pME

Sum of these two is fake pME data set

15

(pME-LE) fake data set as a function of pME POT:

(pME-LE)SMOOTH at 1e18 POT

pME POT

(pME-LE)FAKE at 1e18 POT

(pME-LE)FAKE at 1e19 POT

(pME-LE)FAKE at 1e20 POT

16

Used TMinuit with MIGRAD for the fit, with two parameters “parLE” and “parME”

parLE and parME are started at 1.0 and cannot be negative.

Fit fake data set

with

Used

This is an example for pME-POT=1e18

bins LEpME

LE

K

ME

K

LEMEFAKE

FAKE

vvparLEvparMEvLEpME2

,,2)(

from -,K-)ME

( from -,K-)LE

from +

pME

parME

from +

LE

parLE

17

Fake data set and fit are repeated 5,000 times.

Could this work with our current amount of pME POT ~ 1e18 ?

Does not work at this POT !

18

What about 1e19 POT ?

19

5e19 POT2.5e19 POT

At other values of pME POT:

7.5e19 POT 1e20 POT

20

What about systematics?

One systematic is our assessment of from -,K-)ME - ( from -,K-)LE:

Need to get this from MC and not from fit (need more pME stats)

Proper way to estimate error might be looking how much variation with reweighing.

Other systematics (cross-sections, … etc) could be assessed by varying shape of spectra.

Had a preliminary look by not correcting for at all:

pME POT 1e19 2.5e19 5e19 7.5e19 1e20

shift in parLE 1.17 1.19 1.19 1.19 1.20

shift in parME 1.10 1.12 1.12 1.12 1.12

from -,K-)ME - ( from -,K-)LE

21

Summary & Ongoing work

Almost no variation observed when reweighted from +

Have our 5 semi-independent methods for assessing ’s from +:

Fit method needs more work. Currently trying to converge on the best fit for antineutrinos in nubar group.

Need more pHE MC statistics to see if we can do something similar with the pHE data.

New idea of using the MC shapes to fit the (pME-LE) difference:

Allows to cancel many unknowns in ’s from -,K-

Preliminary study shows measurement is possible to ~20% with ~2.5e19 POT of pME data

pME data may be useful for other analyses

22

Backup slides

23

In Scale Method 5 C(E) was approximated with

Main idea of scaling methods is:

),()(),( KECK MCvv

Overall method:),(),,()( KK v

DATAvv

),,(),,(

KK

MCv

DATAv

Pol 4th deg

Why “Scaling method 5” was thrown away:

24

Then we have: ),,()1(),(

),,()1(),(

KfK

KfKMCv

MCMCv

DATAv

DATADATAv

thus giving:

),,(

),,(

1

1

),(

),(

K

K

f

f

K

KMCv

DATAv

MC

DATA

MCv

DATAv

MCf

Let be the fraction of + in the spectrum (Data)

Let be the fraction of + in the spectrum (MC)

DATAfMCf

But if then MCDATA ff )(),(

),(

),,(

),,(EC

K

K

K

KMCv

DATAv

MCv

DATAv

This method implied assuming fDATA = fMC

25

Antineutrino provenance: