Direct Detection of : Status and Issues

Chris Savage University of Utah

Overview

Aprile et al. (2012)

CDMS Si

LUX Overview

Are any/all of the experiments seeing dark matter? Are the results truly incompatible?

Outline . Dark matter: what is it and how to detect it? (WIMPs) . Basics of direct detection . Experiments & results . Issues • Couplings (particle physics) • Halo model (astrophysics) • Backgrounds • Statistical analysis • Energy calibration Ask questions at any point !

Why Dark Matter?

• Indirect evidence . Velocities of galaxies in clusters (Zwicky 1933) . Galaxy rotation curves (Rubin 1960’s) NASA/WMAP Science Team . Cosmic microwave background . Big bang nucleosynthesis . Structure formation . Gravitational lensing

Figure from astronomynotes.com Colley et al. (HST) What is Dark Matter?

Is it…

• …astrophysical objects? Massive Astrophysical Compact Halo Objects (MACHOs) . Microlensing searches: not significant contribution to DM

• …a modification to gravity? MOdified Newtonian Dynamics (MOND) . Bullet cluster: MOND disfavored

NASA/CXC/CfA/M.Markevitch et al.; NASA/STScI; Magellan/U.Arizona/D.Clowe et al.; ESO WFI Roszkowski (2004) What is Dark Matter?

…Particles!

• . Proposed to solve strong CP problem

• WIMPs Weakly Interacting Massive Particles . “WIMP miracle” . Supersymmetric DM (), asymmetric DM, mirror DM, composite DM, … . WIMP-like: neutrinos (too light)

• SIMP, WIMPzilla, gravitino, etc. How to detect Dark Matter?

Interactions with Standard Model particles

 stuff p p p 

 stuff   p 

Annihilation Scattering Production

Indirect Detection: Direct Detection: Accelerators: Halo (cosmic-rays), Look for scattering LHC capture in Sun (’s) events in detector

Direct Detection See Freese, Lisanti & CS (2012) Direct Detection for a review Goodman & Witten (1985)

Detector • Elastic scattering of WIMP Recoiling WIMP nucleus off detector nuclei

Scatter WIMP • Recoil spectrum: 1 dR (E,t)   F 2(q)  (E,t) dE 2 0 0 2m  (E,t)  dv 1 f (v) vmin (E) v

Particle Physics: Astrophysics: WIMP-nucleus interaction WIMP distribution

CDMS, EDELWEISS, CRESST, COUPP, ZEPLIN, XENON, LUX, CoGeNT, TEXONO, … Annual Modulation WIMP Halo Wind Drukier, Freese & Spergel (1986) 30 km/s

. non-rotating ~300 km/s (to first order)

. Rotating disk (Sun)  WIMP wind

. …+ Earth’s motion • With disk (June) • Against disk (December)

Annual Modulation

NAIAD, DAMA, CoGeNT, DM-Ice, …

Directional Detection

• Determine direction of recoiling nucleus

• Greater sensitivity A. Green (2010) to halo models

DRIFT, … Direct Detection

Non-relativistic velocities O(100 km/s):  O(10 keV) recoil energies . Depend on nuclear & WIMP masses (kinematics) . Requires very sensitive detectors

• Typical signatures of recoiling nucleus . Ionization . Scintillation . Phonons (heat)

• Backgrounds Reduce backgrounds: . Electron recoils: gammas, betas material selection, . Nuclear recoils: neutrons deep underground

Direct Detection

• Basic recoil rate . Background contamination . Background discrimination using multiple signals: detection with only few events

Like hadron collider: • Annual modulation first to see signal, . Most backgrounds do not modulate but messy . Requires large number of events

Like lepton collider: • Directional use for precision . Difficult to reach same target masses measurements . Better characterization of WIMP velocity distribution

Background Discrimination

• Good discrimination Akerib et al. (2004) [CDMS] . CDMS: phonons & ionization . CRESST: phonons & scintillation . XENON: ionization & scintillation

• Poor discrimination . CoGeNT: ionization only . DAMA: scintillation only CDMS • Also: . Signal risetimes (phonons) . Multiple scatters (incl. neutrons)  source (electron recoils) . … n source (nuclear recoils)

Experiments and Results Standard assumptions

Spin-independent, elastic scattering 2 . Cross-section   A p . WIMP mass

Standard Halo Model . Isothermal sphere (Maxwell-Boltzmann) . Non-rotating

D. Dixon, cosmographica.com Experiments

. Aim: higher target mass, lower backgrounds, lower threshold . Every detector is test bed for future detector • e.g. XENON1  XENON10  XENON100  XENON1T

Gaitskell, UCLA DM 2012 Experiments

Raw event rate

good discrimination few backgrounds

poor discrimination many backgrounds

Modulation Low-background analyses

Standard analysis for multi-signal experiments

. Choose cuts to have ~ 1 background event (on average)

. Discrimination worse at low energies:

analysis threshold well above trigger Akerib et al. (2005) [CDMS] threshold

. Best limits for moderate/high WIMP masses

. No sensitivity to light WIMPs

CDMS

CDMS [Ge] CDMS, CRESST & LUX Science 327, 1619 (2010)

CRESST [CaWO4] EPJ C72, 1971 (2012)

No significant excess

LUX [Xe] 1310.8214

Too many events in nuclear recoil band

background-only rejected at 4.7 CDMS, CRESST, LUX & XENON

Aprile et al. (2012)

CRESST CDMS Si

CDMS

XENON LUXLUX CDMS Silicon CDMS [Si] arxiv:1304.4279

background-only rejected at 99.8% Low-threshold analyses

Trade discrimination for lower threshold . Sensitivity to light WIMPs . Weaken limits elsewhere CDMS [Ge] XENON10 [Xe] PRL 106, 131302 (2011) PRL 107, 051301 (2011) [Erratum: PRL 110, 249901 (2013)]

CDMS

XENON10

XENON100 Zn-65/Ge-68 CoGeNT L-shell

• Ionization only (limited discrimination)

excess low energy events

…if dark matter 2012: surface events CoGeNT [Ge] PRL 106, 131301 (2011) Modulation: DAMA

• Modulation search using NaI crystals (scintillation only) . DAMA/NaI: 1996-2002 R. Bernabei et al., Riv. Nuovo Cim. 26N1, 1 (2003) . DAMA/LIBRA: 2003-2009 R. Bernabei et al., Eur. Phys. J. C67, 039 (2010)

Freese, Lisanti & CS (2012)

8.9 annual modulation Modulation: DAMA

Kelso, Sandick & CS (2013)

CoGeNT [Ge] Modulation: CDMS & CoGeNT PRL 107, 141301 (2011)

CDMS [Ge] arxiv:1203.1309

CoGeNT CDMS

CoGeNT: 2.8 modulation Experimental Status

• CDMS (Si), CoGeNT, CRESST & DAMA signals inconsistent with each other …and preferred SUSY region • If any of the signals are from dark matter, CDMS (Ge), XENON, and/or LUX should have had more events Aprile et al. (2012)

CDMS Si

LUX

Issues Issues

What issues can affect interpretation of direct detection results?

• Particle physics (interactions) • Astrophysical uncertainties (halo) • Unknown backgrounds • Statistical analysis • Detector energy calibrations

I will skip many examples (but can come back during question period)

Issue: particle physics Particle Physics Issues

Assumption: single SI cross-section

Other possibilities:

. Spin-dependent couplings

. Isospin-violating dark matter

. Inelastic scattering

. Couplings to electrons instead of nuclei

. Mirror dark matter (Rutherford scattering) See e.g. R. Foot, Phys. Lett. B703, 7 (2011)

. … Spin-dependent

. Coupling to spin rather than mass . DAMA proton-odd, most others proton-even  proton-only SD coupling explains DAMA + null results . 2012: PICASSO closes DAMA SD window

Archambault et al., PLB 711, 153 (2012) Isospin-violating

4 2 2 Leading order only: next order terms affect canceling  SI  Zf p  (A Z) fn   [Cirigliano, Graesser & Ovanesyan (2012)]

. SI coupling to proton significantly different than neutron

neutralino (scalar): fp  fn 2 heavy neutrino: fp  - (1-4sin w) fn

. For fp  -0.7 fn , can exactly cancel SI cross-section for a heavy isotope

. Cannot cancel all CDMS and XENON isotopes at same time

. Fine-tuning!

Particle Physics Issues

• Spin-dependent (no) • Isospin-violating (probably not, fine-tuned) • Inelastic scattering (now excluded*) • Electron coupling • …

Range from well motivated to ad-hoc particle construction. How to connect to larger theory (e.g. supersymmetry)?

Are we throwing away reasons we expect to have WIMPs?

Issue: astrophysics Halo Models

Fiducial case: isothermal sphere . Maxwell-Boltzmann distribution (with cutoff)

D. Dixon, cosmographica.com Actual case . Smooth (virialized) halo . Structure: tidal streams, dark disk, …

Relevant quantities . Local DM density . Local velocity distribution • SHM-like? If so, what parameters? • If not, what?  N-body

D. Martinez-Delgado & G. Perez

Velocity Distribution

• SHM: uncertainty in parameters (vrot, v0, vesc) • Affect compatibility? not significantly

CS, Freese, Gondolo & Spolyar (2009)

See also (e.g.): Hooper & Kelso (2011) N-body simulations

• Via Lactea, GHALO,… Kuhlen et al. (2009) . DM-only . Moderate differences from Maxwell-Boltzmann

• MaGICC g1536 . DM + baryons simulation . Milky Way –like galaxy (~10-15% smaller mass, velocities) . For comparison: DM-only with same initial conditions See also: CS, Valluri, Stinson (in progress) [+ Kelso, Freese] Ling et al. (2009)

1 (vmin )  dv f (v) N-body simulations (eta) vv v min

DM-only DM+baryons N-body simulations (modulation)

DM-only DM+baryons

N-body simulations

• Moderate deviations from Maxwell-Boltzmann, but… . Boost by disk rotation . Kinematics: always have integration over velocities  Maxwell-Boltzmann works reasonably well for direct detection!

• Inferred MB parameters (assuming isothermal): . Bad for DM-only simulations . Good for DM+baryon simulations

• On the continued use of DM-only results: why? . Inner galaxy dynamics not DM dominated . Baryons: larger effect than precision of simulations

See e.g.: Astrophysics Issues Pato, Strigari, Trotta & Bertone (2012)

• Local halo dominated by smooth background . N-body: Maxwell-Boltzmann close enough? . Does not alter experimental compatibility

• Structure . Can have significant impact in certain cases, even when small . Predicted by some simulations, but severely limited by others . Difficult to make general conclusions regarding compatibility, but…

• Halo model independent analyses Fox, Liu & Weiner (2011); Frandsen et al. (2012); Gondolo & Gelmini (2012) . Use conservative bounds on halo kinematics behavior . Severely constrain astrophysical explanation of experimental results

Issue: unknown backgrounds Unknown Backgrounds

• Low energy, low rate detectors . Backgrounds often not well characterized/understood . Novel detectors sometimes present new and unexpected sources of background events

• Potential source of “signal” in CDMS (Si), CoGeNT, CRESST, and DAMA

• Example backgrounds . Muon-induced events in DAMA . Lead recoils in CRESST . Surface/zero-charge events in CDMS, CoGeNT

Muons in DAMA

• Cosmic-ray muon flux known to modulate . Depends on height of atmosphere, which changes with seasons

• Muon vetoing: muons cannot be direct cause. Indirect cause? . Showers . Delayed phosphorescence [Nygren (2011)]

J. Pradler, UCLA DM 2012 Muons in DAMA

Any muon-induced events (direct or indirect) strongly disfavored

. Phase of muon flux lags DAMA modulation

. DAMA has consistent phase, muon flux does not (and is not expected to)

. Statistical arguments: Poisson fluctuations should be much larger than observed fluctuations in DAMA

Lead Recoils in CRESST Kuzniak, Boulay & Pollmann, Astrop. Phys. 36, 77 (2012) • Background: 210Po  206Pb +  (at surface) • Monte carlo simulations: flat vs. rough surface  underestimating background events!

• A

Surface Events (CoGeNT) Aalseth et al. (2010)

Calibration:

Data: Increasing contamination at low energies

Surface Events (CoGeNT)

• J Collar @ TAUP 2011:

significant number of Kelso et al. (2011) surface events

Aalseth et al. (2012): middle case

Kelso et al. (2011) CDMS: Trigger Threshold

Agnese et al. (2013) • Are there potential Silicon populations of events below trigger threshold?

Ahmed et al. (2011) • Answer: YES Germanium LE . Zero-charge events . …

XENON: also has known population of events below S1 trigger Unknown Backgrounds

• Significant fraction of CoGeNT “signal” now attributed to surface events . Still claim excess events . Still have modulation

• Very reasonable CRESST background explanation

• Many potential modulation backgrounds in DAMA have been excluded . Not easy to match all DAMA data

• …but new backgrounds often uncovered. What are we missing?

• Be cautious near thresholds!

Issue: statistical analysis Statistical analysis issues

• Weak statistics • Flawed/misleading statistics • Missing/incomplete statistics • (Overly-)conservative statistics

• Examples: . Threshold and single bin/counts-only analyses (weak statistics) . DAMA binning (weak statistics) . CoGeNT 2010 analysis (flawed/misleading statistics) . Collar & Fields (2012) reanalysis of CDMS low-energy data (missing/incomplete statistics) . XENON100 energy calibration (conservative) [later]

Single bin / counts-only analysis

Fit single composite bin

fit to all bins

Total modulation (counts) compatible, spectrum is not!

CS, Gelmini, Gondolo & Freese (2008) DAMA binning

• 2 goodness-of-fit test affected by binning (can make stronger)

Kelso, Sandick & CS (2013) CoGeNT (2010)

Statistical issues:

• Cut away regions that were “uninteresting” (misleading)

• Improperly calculated regions (flawed) . Less than 0.5 result (much less 90%) . Black box numerical routine?

• Misunderstand degrees of freedom

CoGeNT (2010)

CoGeNT “Region of Interest”

Statistically valid region CDMS low-energy reanalysis

A Maximum Likelihood Analysis of Low-Energy CDMS Data, Collar & Fields, arxiv:1204.3559.

• Recoil energy & charge for each event

• Model backgrounds: . Electron recoils . Surface events . Zero charge (edge events)

• +nuclear recoils (signal)

 Likelihood analysis

Claim: 5.7 excess CDMS low-energy reanalysis

• Missing: goodness-of-fit check

• Likelihood analysis for each of 8 detectors: results incompatible (goodness-of-fit is poor)

• Conclusion: If the backgrounds are correctly modeled …the background only case is rejected at > 5 …but the excess is inconsistent with WIMPs at > 3

• Why? . Modeling of distribution tails does not match data . Incorrectly modeled distribution virtually guaranteed to find “signal” (whether or not it is really there)

Issue: energy calibrations Energy Calibration

• How to reconstruct recoil energy from observed signal (e.g. scintillation)? . Some calibrations based upon poorly measured quantities

• Proper calibration important for sensitivity to light WIMPs since most signal is near threshold

• Examples: . Quenching factor Q in NaI (DAMA) . Scintillation efficiency factor Leff (XENON) . Energy resolution (XENON)

Quenching Factor: NaI

Hooper et al. (2010): Collar, arxiv:1302.0796 using QNa = 0.30  0.13

Higher Q Lower Q Liquid Xenon/Argon detectors: principles of operation

S2: ionization electrons

S1: prompt photons

Aprile et al. (2010) [XENON10] XENON

• Poisson statistics + PMT resolution [e.g. XENON100] . Recoils produce O(50-1000) photons, only ~6% trigger PMT • Absorption by liquid/walls • PMT photocathode quantum efficiency . Analysis threshold: 3 photo-electrons (PEs) • Nominally require 50 photons • But Poisson fluctuations: event with 30 photons has 27% chance to produce 3+ PE

• How much (average) scintillation in event? Leff

• Analysis “energy” range (badly labeled)

XENON Event Energies

S1: observable

Aprile et al. (2012) [XENON100]

Bad energy estimator XENON Event Energies

Poisson fluctuations All events at same energy + PMT response, etc. XENON Event Energies

Sorensen (2012) XENON Leff What is happening here? (stat/systematic uncertainties)

Aprile et al. (2011) [XENON100]

What is happening here? (unknown behavior) XENON Leff

• Unknown Leff behaviour below 3 keV • Conservative assumption: treat Leff = 0 there • Too conservative?

Aprile et al. (2012) [XENON100]

3 keV (Plante et al.)

< 3 keV

Also: calibration studies Caveats: XENON Leff  XENON 100 day exposure with higher threshold  Leff cutoff at 3.9 keV (now non-zero to at least 3 keV)

Savage et al. (2010)

Left to right: constant, falling, zero

Leff below 3.9 keVnr XENON Efficiencies

Fractions of events passing all cuts in XENON10 (2  S1  75 PE)

constant/falling/zero Leff

Caveats:  Lower threshold than XENON100  Leff cutoff at 3.9 keV (now non-zero to at least 3 keV) XENON Leff & energy resolution Also applies to LUX

• Can Leff uncertainties be used to reconcile experimental results?

• Assumptions are already very conservative (and known to be overly conservative) . Constraints almost certainly cannot be made weaker . Constraints are very probably significantly better at low masses

• Conservative Leff + no upward Poisson fluctuations: overkill

• Be skeptical of claims of compatibility using events over 6.7-30.5 keV

Summary and Remarks

• Four (possibly) positive signals for dark matter, numerous negative results

• Difficult to reconcile some experimental results (let alone all of them)

• Possibilities . Particle physics: maybe, but at what cost? . Astrophysics: unlikely . Unknown backgrounds: significant possibility • Modified/unconsidered backgrounds for CRESST, CDMS . Energy calibration: making things worse

Summary and Remarks

• DAMA: most difficult to reconcile, but impervious to postulated backgrounds (so far)

• CoGeNT: how to reconcile with CDMS low-energy results? (same material & energy range)

• CRESST: explained by surface roughness?

• CDMS Silicon: need to lower trigger threshold

• Answers in upcoming results…

Future

Low mass region . SuperCDMS: low-energy analysis with cleaner detectors [this year] . CDMSlite: very low energy, ionization-only [this year] . DM-Ice: southern hemisphere [???] (also ANAIS, KIMS) Arxiv:1201.2402 SUSY “preferred” regions . LUX: 10 improvement in sensitivity [next year] . XENON1T: 100 [2015] . DARWIN: 1000 [2018] . …

…and beyond: solar neutrino background