From naranjo Thu Nov  4 17:46:51 2004
Date: Thu, 4 Nov 2004 17:46:51 +0000
To: michael.loughlin@ukaea.org.uk, APX077@convetry.ac.uk,
        Nigel.hawkes@npl.co.uk, colin.murray.01@bbc.co.uk
Cc: puherman@ritva.physics.ucla.edu
Subject: UCLA neutron detection notes
Reply-To: Brian Naranjo <naranjo@physics.ucla.edu>

Hi,

I put together some notes for the neutron detector.
I'll send everyone the final analysis for BBC 03
later today.  If anyone has a question, please call
me at (310) 825-1842.

Brian


ENERGY CALIBRATION
==================
I use a series of Compton edges (Ba-133 @ 356 keV, Na-22 @ 511 keV,
Cs-137 @ 662 keV, and Na-22 @ 1275 keV) to determine the detector's
electron recoil response.  The conversion between anode charge output
and electron energy is found to be of the simple form y=ax so
that anode charge is measured in 'electron keV' throughout
subsequent analysis.

In detail,
  Step 1 - Monte Carlo generation of gamma response (page 1)

    I used Geant4 to simulate net electron recoil spectra
    of the gamma sources.  The dominant interaction between
    the gamma rays and the liquid scintillator is Compton
    scattering, for which I used Geant's low energy Compton
    extension.  This model uses evaluated cross section data
    from Livermore's EPDL97 to determine the mean free path.

    In the MC data, beyond the the Compton edge, you can see a
    second edge due to double scattering in the detector.
    For example, at the Compton edge, a 662 keV Cs-137 gamma will backscatter
    with an energy of 184 keV.  That 184 keV gamma can then give, at most,
    77 keV to an electron.  Indeed, there is a secondary edge at 555 keV.
    The process continues for triple scatters, etc.

    Ba-133 emits at several energies beside 356 keV which I included
    in the simulation.

  Step 2 - Data collection, analysis, and PMT bias adjustment (page 2)

    One by one, the sources are placed next to the scintillator
    and 500000 pulses are digitized with an 8 bit 1 GS/s Acqiris
    digitizer.  Each pulse is corrected for baseline drift, discriminated
    using a linear fit to the rising edge, and integrated over 500 ns.
    The resulting anode charges are histogrammed.

    To locate the Compton edge in the real data, the MC data
    is convoluted with a gaussian, offset, scaled, stretched, and fit to the
    real data.  The gaussian's sigma is the quoted value of resolution.
    Plotting the anode charge of the four edges against the tabulated
    values of the energies (bottom plot) gives the conversion between
    anode charge and electron keV.

    Prior to each run, I calibrate the detector by adjusting each PMT
    bias so that the Compton edge of Cs-137 is approximately 300 pC.
    This is at approximately -1900 V for our 120 mm Hamamatsu R1250s.


TIMING CALIBRATION
==================
Timing differences between the neutron detector PMTs and SL PMTs
(10 mm Hamamatsu R1463-01) due to differences in cabling and PMT
transit times were measured using the Na-22 source.  I attached a
small plastic scintillator to the face of the SL PMT and
sandwiched the Na-22 source between the plastic and and the liquid
scintillators.  I then triggered on the back-to-back 511 keV positron
annihilation gammas.  The timing differences are shown on
page 3.  Typically, the neutron PMTs are about 10-12 ns later than
the SL PMTs.  This approximately agrees with the PMT spec
sheets and our cabling:
  54 ns (R1250) ~= 24 ns (R1463-01) + 16 ns (extra cabling) + 12 ns (diff)

SL 1 is the PMT used in BBC 01 through BBC 03.  It was found to
have an anomalous drifting dark rate and was replaced by SL 2
in BBC 04 through BBC 06, reducing the dark rate by an order
of magnitude.


MONTE CARLO DETECTOR SIMULATION
===============================

I used Geant4 to simulate the detector's response to 2.45 MeV
neutrons emitted isotropically from the center of the SL cell
(see attached pictures 'bbc_neutron_top.gif' and 'bbc_neutron_side.gif').
The objects specified in the simulation are the quartz SL cell
filled with deuterated acetone, two borosilicate liquid scintillator
cells filled with NE213, and the glass double-paned refrigerator window.
The primary neutron interaction is elastic scattering off nuclei.
Again, I used a high precision Geant4 model incorporating tabulated cross
sections.

  NE213 Scintillator Response (top of page 4)

    I simulated 5000000 neutrons.  For each proton recoil in the liquid
    scintillator, I convert the proton energy into equivalent
    electron light output using the response function of Lee & Lee
    [NIM A 402 (1997) 147].  The total equivalent light output
    is histogrammed in the figure.  From eq. 10 of Lee & Lee, a 2.45 MeV
    proton recoil corresponds 857 electron keV, where the billiard
    ball edge is on the top plot.  Since, according to Lee & Lee,
    recoil protons having energy under 340 keV don't give light
    and 2.1 MeV protons give 690 keV light, the flat region between 690 keV
    and the edge at 850 keV is due to single scatters.  The
    hump in the middle is due to multiple scatters.

    To estimate the measured light response, I convolve the Monte
    Carlo histogram with the earlier obtained electron resolution
    (an approximation, since we want the proton resolution)
    obtaining the smooth curve.  Setting a lower threshold
    of 300 electron keV, the area under the curve corresponds
    to 4.4% of the 5000000 neutrons.  There is another 10%
    cut in detection efficiency due to PSD threshold, so the
    net detection efficiency for 2.45 MeV neutrons in the BBC
    geometry is 4%.
    
  Timing Spectrum (bottom of page 4)

    Times of flight between the SL cell and the first scatter
    in the neutron detector are shown.  This histogram together
    with the timing calibration determine where to look for
    SL-neutron coincidences.  A properly located window of 10 ns
    will capture the vast majority of coincidences.


PULSE SHAPE DISCRIMINATION
==========================
Given a pulse,
 1. measure trace baseline before pulse
 2. discriminate leading edge
 3. integrate total pulse (out to 500 ns), giving total anode charge
    and pulse energy
 4. Integrate "slow light" and "fast light".  The separation point
    is calibrated by minimizing electron contamination of the proton region:
       sep time [ns] = 15.6 ns + 0.013 energy [keV]
    So, for 300 electron keV, it is 19 ns.  For 1000 electron keV, it is 28 ns.
 5. Make plots as shown on page 5.
 6. The proton and electron regions are determine in the PSD calibration.
    For a given energy, the proton region includes 90% of protons.
    Therefore, the total detector efficiency is 4.4% * 0.9 = 4%.
 7. The proton candidates are compared against tabulated waveforms
    and measurement errors.  If the reduced chi^2 is greater than 1.5,
    it is rejected.  Typically, these are pile-up or ion feedback pulses
    in the PMTs.

  AmBe Neutron Source (top of page 5)
  Cosmic Background (bottom of page 5)
