Nuclear Recoil Scintillation and Ionization Yields in Liquid Xenon Dan McKinsey Yale University Physics Department February, 011 Indirect and Direct Detection of Dark Matter Aspen Center of Physics
Energy Calibration: determine the energy of nuclear recoils energy of nuclear recoils (NRs) top PMT array PMT Array measured signal in # of pe light yield for 1 kev γ in pe/kevee proportional Gas Xe E g WIMP Liquid Xe E nr = S1/L y /L eff S er /S nr relative scintillation efficiency of NRs to 1 kev γ s at zero field quenching of scintillation yield for 1 kev γ s due to drift field direct Gamma! Xe e- E d e- 1 kev γ (Co-57) quenching of scintillation yield for NRs due to drift field bottom PMT array
L eff = scintillation per unit energy for nuclear recoils scintillation per unit energy for electron recoils In practice, we define the denominator based on 1 kev photoabsorption events from Co-57 In the XENON analysis, we assumed an energy-independent Leff of 0.19 for the WIMP search analysis, and for determining our cross-section limits. Uncertainty in Leff was the main source of systematic uncertainty in determining the cross-section limits. New measurement of Leff: A. Manzur et al., Phys. Rev. C 81, 05808 (0). (Also measured Sn at several drift fields)
Experimental setup water shield polyethylene shield BC501-A organic scintillator water shield neutron generator.8 MeV n θ water shield liquid xenon detector cryostat E R = E n m n M Xe (m n + M Xe ) (1 cos θ) Energies: 4-66 kevr
Selecting single nuclear recoils Quality cuts Q0: remove noise event, high energy events, S1 asymmetry Select neutrons using PSD and time of flight (TOF) Gamma events Counts 3 accidental coincidence time of flight cut for single scatter nuclear recoils multiple-scatter nuclear recoils Neutron events 0 0 30 40 50 60 70 80 90 0 Time-of-flight [ns]
Systematic error polyethylene shield ~1% ~ % a) b) polyethylene shield n θ n θ c) n polyethylene shield 6-16 % θ1 θ a) Multiple elastic scatters b) Outside scatters c) Size and position d) Cross-section database ~ - 4%
Water boxes Cryostat Scintillator Polyethylene Neutron Generator
Comparing data & Monte Carlo E R = kev r χ TOF kevr L eff To compare MC & data: 1 E R E e σ = 3. N phe 3 software + trigger efficiency
Leff results Relative Scintillation Efficiency 0.35 0.3 0.5 0. 0.15 0.1 0.05 0 4 40 Energy [kev ] r Manzur et al,, 009 Aprile et al, 009 Chepel et al, 006 Aprile et al, 005 Akimov et al, 00 Arneodo et al, 000 Relative Scintillation Efficiency 0.35 0.3 0.5 0. 0.15 0.1 0.05 0 XENON ZEPLIN-III 4 40 Energy [kev ] r /kev r ] - Ionization yield [e 0, ] r Energy [kev
L eff model L eff = q ncl q el q esc q ncl nuclear quenching (Lindhard factor), energy goes into heat. q el electronic quenching. Bi-excitonic collisions Escape electrons Xe + Xe Xe + Xe + + e q esc = q el = 1 1 + k de dx N ex + N i N esc Nex 1 + Ni 1 Nesc 1 = α + 1 β α + 1 β 1
L eff model
Leff results Relative Scintillation Efficiency 0.35 0.3 0.5 0. 0.15 0.1 0.05 0 4 40 Energy [kev ] r Manzur et al,, 009 Aprile et al, 009 Chepel et al, 006 Aprile et al, 005 Akimov et al, 00 Arneodo et al, 000 Relative Scintillation Efficiency 0.35 0.3 0.5 0. 0.15 0.1 0.05 0 XENON ZEPLIN-III 4 40 Energy [kev ] r /kev r ] - Ionization yield [e 0, ] r Energy [kev
Nuclear Recoil Ionization Yield and Field Dependence Aprile et al., astro-ph/060155, Phys. Rev. Lett 97, 08130 (006). Energy threshold: kevr Note modest (but not negligible) increase in nuclear recoil ioniza5on yield at higher fields.
Data from Manzur et al TABLE I. L eff and S n values for the different nuclear recoil energies studied. The third column gives the L eff values relative to 1-keV rays. The first error is the statistical uncertainty, and the second error is the systematic uncertainty. E r L eff S n S n S n S n [kev r ] at 0.0 kv/ cm 0.73 kv/ cm 1.0 kv/ cm 1.5 kv/ cm 4.0 kv/ cm 15 66.7 ± 3.3 + 0.018 + 0.0 0.178-0.016-0.009 0.91 ± 0.07 1.11 ± 0.09 0.88 ± 0.06 1 57.7 ± 3. + 0.009 + 0.004 0.18-0.009-0.00 0.95 ± 0.05 0.95 ± 0.06 0.93 ± 0.04 0.93 ± 0.06 95 46.1 ± 4.9 + 0.038 + 0.0 0.158-0.039-0.009 0.97 ± 0.08 0.8 ± 0.08 80 34.9 ±.1 + 0.04 + 0.014 0.133-0.09-0.01 1.33 ± 0.6 1.30 ± 0.5 67 5.7 ±.0 + 0.05 + 0.001 0.13-0.019-0.006 0.95 ± 0.06 0.91 ± 0.1 0.95 ± 0.07 58 19.6 ±.6 + 0.046+ 0.008 0.157-0.036-0.019 0.70 ± 0.06 1.03 ± 0.13 50 15.1 ± 1.5 + 0.030 + 0.019 0.13-0.03-0.014 0.83 ± 0.16 1.0 ± 0.0 1.01 ± 0.15 1.08 ± 0.18 40 9.8 ± 1.3 + 0.07 + 0.09 0.118-0.0-0.0 0.91 ± 0.17 1.64 ± 0.50 0.98 ± 0.18 1.6 ± 0.45 35 7.6 ± 1. + 0.08 + 0.06 0.5-0.0-0.09 0.79 ± 0.8 1.06 ± 0.30 0.79 ± 0.8 3 6.4 ± 1.1 + 0.07 + 0.03 0.094-0.0-0.09 0.9 ± 0.37 1.5 ± 0.45 0.93 ± 0.38 1.38 ± 0.5 30 5.7 ± 1.1 + 0.08 + 0.07 0.077-0.0-0.06 1.35 ± 0.67 1.18 ± 0.61 8 4.9 ± 0.9 + 0.06 + 0.06 0.088-0.03-0.03 1.16 ± 0.45 1.34 ± 0.50 0.87 ± 0.35 5 3.9 ± 0.9 + 0.034+ 0.018 0.073-0.05-0.06 1.19 ± 0.5 1.30 ± 0.38 1.88 ± 0.78
XENON limit ] [cm s WIMP-nucleon -41-4 Leff = 0.19 Leff Leff model -43-44 3 Mass [GeV]
Discussions of the Manzur et al measurement of Leff Discussion driven by tension between XENON and CoGeNT/DAMA and the possibility of a light WIMP signal. See J. Collar,.5187, A. Manalaysay, 07.3746, Sorensen, 07.3549, Sorensen and Dahl, 11.6080 Collar and Manalaysay's concerns essentially center around the trigger/software efficiency, and how it affects the Manzur et al measurement. We certainly are aware of this issue, and took pains to deal with it carefully in our analysis. Note that Manalaysay and Collar have opposite recommendations: they advocate modifying the Manzur et al. values in opposite directions. Paraphrasing Manalaysay: If Manzur et al underestimated their efficiencies, then Manzur et al could be underestimating Leff. Then the real Leff values could be increased, so as to agree with Aprile 009. My answer: These efficiencies could be pulled up, within our stated systematic uncertainties. This would lessen the tension between Aprile 009 and Manzur 0. However, keep in mind that a) Aprile 009 has other significant systematic uncertainties having to do with the subtracting the very large neutron multiple scattering background, and b) theoretical considerations suggest that Leff should fall at low energy, so beware of wishful thinking. Leff very likely decreases as lower energies, as indicated by Manzur et al, ZEPLIN-III, and the XENON calibration re-analysis by P. Sorensen. Paraphrasing Collar: The lower efficiency at low energies makes the Manzur et al experiment less sensitive to the true value of Leff. So Manzur et al could be overestimating Leff, weakening tension between XENON and CoGeNT. My answer: The Manzur et al measurement does indeed have poorer sensitivity at low energies, largely because of the decreasing efficiency. But this shows up as a larger statistical error bar because a larger range of Leff values yield consistency between data and Monte Carlo. This statistical uncertainty is already included in our stated results. Overall, none of the stated concerns of Collar or Manalaysay give us reason to change the Manzur et al. Leff values. Recommendation from Sorensen, and Sorensen and Dahl: Use a decreasing Leff, in agreement with Manzur et al. My recommendation: If you wish to be conservative (optimistic?) about how stringently LXe-based dark matter limits constrain your favorite theory, then use the -1s Leff values from Manzur et al. Note that this is also consistent with Collar's recommendation.
Of course, it would be best to measure the detector response at low energy, throughout the LXe volume, and use this to extract the nuclear recoil response. This is tricky, because the LXe is such a terrific self-shielding material, making it difficult to get low-energy gamma-induced events within the fiducial volume. Enter Kr-83m...
LXe scintillation data from Kr-83m dissolved into LXe 4000 3500 9.4 kev line 3.1 kev line 3000 Counts 500 000 1500 00 500 0 0 0 30 40 50 Energy (kev) L. Kastens et al, Physical Review C 80, 045809 (009).
Typical Kr-83m event in -phase Xe PMT Signal Voltage [mv] 8 6 4 0 PMT Signal Voltage [mv] 8 6 4 0 40 60 80 300 30 340 360 380 400 Time [4.0 ns/pt] - 500 3000 3500 4000 4500 Time [4.0 ns/pt] Kastens et al, JINST 5, P05006 (0). Also see Manalaysay et al, arxiv:0908.0616
Repeated Kr-83m introduction into -phase Xe detector 15s Pulse 5s Pulse s Pulse Generator Open 0.5 Generator Closed Generator Open 0. Rate [Arb. Units] 0.15 0.1 0.05 0 0 5 15 0 5 30 35 40 45 50 Time [h]
Particle Identification in Xe at Yale (PIXeY) Goals: a) S/S1 discrimination as a function of drift field b) Energy resolution studies in -phase Xe c) Low-energy electron and nuclear recoil studies Internals Cross-section view.5 meters PMTs LXe volume 1 meter
PIXeY photos
PIXeY Particle Identification in Xenon at Yale Two-phase detector with electric fields for charge collection and counting by proportional scintillation. The drift region is 5.1 cm tall and 8.8 cm on a side, enclosing liters of xenon. PIXeY will have near optimal energy resolution and test γ / n discrimination in electric fields up to 6 kv / cm.
Summary 1) Manzur et al measurement, with comprehensive approach to systematic uncertainties, implies a dropping Leff at low energies in LXe. ) Theoretical models of LXe Leff including Lindhard effect, bi-excitonic quenching, and electron straggling, predict that Leff should indeed decrease at low energies. 3) Electric field quenching of nuclear recoil scintillation light (S n ) is not well constrained at low energies. Need more controlled two-phase measurements at low energies. LXe dark matter searches need to take S n uncertainty into account. 4) Kr-83m calibration allows measurement of the energy scale as a function of position, using a radioactive gas that disperses throughout the detector volume.