Estimating Solar Background Distributions Dan Pershey Jul 10, 2019
From Last Time ❑ It seems like we capture on 36 Ar too frequently ❑ Second peak in reco visible energy distribution comes from capture on 36 Ar • Peak 1: Capture on 40 Ar (E=6.2 MeV) • Peak 2: Capture on 36 Ar (E=8.7 MeV) ❑ 36 Ar concentration is 0.3% • 36 Ar( n,γ ) = 3.9 b (arxiv 1808.08556) • 40 Ar( n,γ ) = 0.67 b (arxiv 1902.00596, ACED) • Should only see about 1.9% as many captures on 21% of captures! 36 Ar as 40 Ar, but we see 21% in sim ❑ Need a truth study to make sure this is not a reconstruction effect • Possible, given 6-8 MeV is near detection threshold 2
Selecting True Background Events ❑ Three processes dominate the low-E background • 36 Ar(n, γ ) • 40 Ar(n, γ ) • 40 Ar( α , γ ) ❑ Made a TTree of all initial neutrons and α particles that were tracked in the detector along with daughter information ❑ After adding cuts on particle types and energies in the daughters, I can select simulated events from each of these processes without any reco effects ❑ Also add in solar’s fiducial cuts on initial n/ α endpoint to ensure we’re interacting in liquid argon ❑ See 94395 / 2010 / 89 events for 40 Ar(n, γ ) / 36 Ar(n, γ ) / 40 Ar( α , γ ) ❑ Capture rate only 13% higher on 36 Ar – must be seeing significant efficiency effect 3
Smoothing 40 Ar( α , γ ) Distribution ❑ We just made found a sample of 89 40 Ar( α , γ ) events >> than the 5 selected by cuts • That’s a decently high statistics sample • We can bootstrap, make a library of 89 gamma cascades, and throw them uniformly over the detector, smoothing out the efficiency curve ❑ Also wanted to take the opportunity to incorporate energy dependence of the α , γ cross section ❑ Higher energy α’s will also have a higher cross section, and 222 Rn decays have lower energy than two daughter nuclei in its decay chain: 214 Po/ 218 Po • And they’ll have higher energy gamma cascades, which could in principle hurt our hep sensitivity 4
Determining Energy Dependence of 40 Ar( α , γ ) Cross Section ❑ There’s a recent measurement for 40 Ca ( α , γ ) that shows the cross section is relatively high compared to spread in model predicted values ❑ Also gives the shape of σ (E α ), roughly quadratic over 3 MeV ❑ Grabbing the model, we can construct the density of states expected, differential in E α – exactly the distribution of E α expected to interact in the TPC • Set cross section for 40 Ar( α , γ ) to 10 μ b at 5 MeV, which is pretty conservative arXiv:nucl-ex/0509006v2 Normalize to σ (5 MeV) = 10 μ b 5
Energy Dependence of the 222 Rn Chain ❑ For a given isotope, this energy dependence will go up to the initial KE of the decay and then cut off ❑ We sum the density of states for each of the four isotopes in 222 Rn chain ❑ Get a saw-edge pattern, where single isotopes start to kick-in and contribute ❑ In LArSoft, the fraction of α’s that fuse is 5.9e-7 -> which would suggest a cross section of a few mb, the number is more like 1e-9 for 222 Rn decays <P αγ > at 4 MeV 7.40e-11 <P αγ > at 6 MeV 1.18e-9 210 Po 222 Rn <P αγ > at 8 MeV 4.52e-9 218 Po 214 Po Fusion fraction per <P αγ > at 10 MeV 1.13e-8 222 Rn decay chain <P αγ > for 222 Rn 1.71e-9 6
Library Generator Algorithm for 40 Ar( α , γ ) Events ❑ Select a random one of 89 saved library events ❑ Pull an E α from the density of states plot on the last page ❑ Scale all energies in the gamma cascade to match the total energy, 8.85 MeV + E α , of the thrown E α • This isn’t physical, but there’s no reason to really expect GEANT’s handling of the gamma energies is believable, so just need a way to get the right visible energy deposited ❑ Pick a random vertex position within the fiducial volume ❑ Hand the initial 4-vectors for each gamma off to GEANT ❑ Events are generated without other radiologicals to speed up the process, but radio contamination (mostly from 39 Ar) is added in during reconstruction using PDF for pileup energy found for solar events • Rule of thumb, about a 10% chance of adding in an extra ≈ 500 keV of energy 7
Results ❑ Generated 10 4 events in this way ❑ Produces a much smoother, and more physical-looking, energy distribution ❑ 563/10 4 events selected – efficiency is right- on what you’d expect from our initial 5/89, though differences in energy distribution change this number ❑ Unfortunately, we now have events with reco energies up to 20 MeV – which is within the optimal range for measuring φ ( hep ) ❑ Standard deviation is 2.9 MeV, versus 1.8 MeV for the true energy thrown and 0.3 MeV for the 5 events selected in vanilla MCC11 files 8
Repeating the Process for 36 Ar(n, γ ) ❑ The procedure works great, and I generated roughly 1000x MCC11 stats for the 40 Ar( α , γ ) in a couple hours of a single gpvm – I figured I could do the same for neutron capture on 36 Ar ❑ Not quite so beneficial here, less visible energy translates to lower efficiency ❑ 177/10 4 events selected vs 32/2010 in the MCC11 dataset • But, it’s still an easy 6x increase in stats ❑ Similar strategy could smooth out captures on 40 Ar, but would require grid computing ❑ Or 10x increase in 40 Ar( α , γ ) or 36 Ar(n, γ ) could give very fine resolution on the distributions 9
Normalizing New Distributions ❑ I want to pipe these into our standard solar spectra plots ❑ The rate of capture on 36 Ar is fixed to the capture rate from standard LArSoft, with the 13% reduction in relative cross section discussed earlier ❑ GEANT is known to over-represent 40 Ar( α , γ)’s by at least a factor of 100, so we can’t just fix the rate to the standard rate predicted • Instead calculate out from first principles • alpha rate x detector mass x 10 years of seconds x avg fusion prob x efficiency • (0.01 Bq/kg) x (40x10 6 kg) x (10 x 3.15x10 7 s) x (1.71x10 -9 ) x (0.0563) = 12130 • Pipe in the new distribution with this normalization for 400 kt-yrs of exposure • Calculated at the relatively conservative σ (5MeV) = 10 μ b • Ended up being beneficial GEANT overestimates this by 100x – otherwise who knows when we wouldn’t have noticed an extra bkg component with 12k events 10
Summary Old version ❑ New bootstrapping method bolsters our bkg statistics by 1000x / 6x for severely stats-limited 40 Ar( α , γ ) / 36 Ar(n, γ ) bkgs ❑ Simulating reasonable energy dependence shows that 40 Ar( α , γ ) will be a non-trivial bkg for hep measurement • 222 Rn bkg larger than hep or 8 B neutrinos for E reco > 17 MeV New version • Probably affects DSNB search, too • Indirectly through biasing hep measurement • Need to understand cross section up to 214 Po energies! ❑ Also makes distribution easier to explain to people outside of the collaboration 11
Proposed Plots for DPF (1/2) ❑ I’ll have a poster on solars at DPF, and want to show four previously un-blessed plots • Due to travel, I need the poster printed by next Tuesday, can we get new plots through? ❑ Draft of poster attached to agenda ❑ Selected efficiency for CC events, which I’ve been showing for a few months • Zoomed in to just the region relevant for solars ❑ Nearby activity makes this CC-specific ❑ Shaded histograms give 8B and hep fluxes, normalized to each-other 12
Proposed Plots for DPF (2/2) ❑ A measure of energy reconstruction performance using energy estimator using visible energy from electron + gammas ❑ Bias looks bad below 8 MeV, but you’re mostly seeing a threshold effect ❑ Above 8 MeV threshold, bias is ≈ 100 keV ❑ Want a sample overview plot wrapping up various bkgs with signal ❑ Capture on 40 Ar set to nominal Capture on 36Ar reduced by 48% ❑ Stress this is reco neutrino energy ❑ Make a comment about hep as DSNB bkg 13
Recommend
More recommend
