Search for Annual Modulation in Combined ANAIS 112 and COSINE 100 Data using Bayesian Model Comparison
Om Godse 1, Shantanu Desai 2 · 2025
Read Search for Annual Modulation in Combined ANAIS 112 and COSINE 100 Data using Bayesian Model Comparison by Om Godse 1, Shantanu Desai 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-23191.txt.
Introduction
Om Godse and Shantanu Desai test a long-running dark matter controversy with a Bayesian comparison. The DAMA experiment has long claimed an annual modulation in its detector rate. That yearly wave is the signature of Earth moving through the galactic dark matter wind. The ANAIS-112 and COSINE-100 experiments were built to check it with the same target material. The authors combine about three years of data from both checking experiments.
They compare a sinusoidal modulation model against a constant model, trying uniform priors and priors centered on DAMA's best-fit values for the signal's frequency and phase. In both the one-to-six and two-to-six kilo-electron-volt energy windows, the log Bayes factor for the cosine model stayed below about one point one five, far too weak to count as evidence. The verdict is no sign of a dark matter modulation in the combined data.
Catalog record
| CREATOR | Om Godse 1, Shantanu Desai 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 13 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-23191
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Search for Annual Modulation in Combined ANAIS 112 and COSINE 100 Data using Bayesian Model Comparison Om Godse 1, Shantanu Desai 2 Abstract We perform a Bayesian model comparison test between a sinusoidal modulation model and a constant value model using about three years of combined COSINE-100 and ANAIS-112 data. We use both uniform priors and normal priors (based on DAMA best-fit values) for the angular frequency and phase of the cosine signal. We find natural log of Bayes factor for the cosine model compared to the constant value model to be less than 1.15 for the data in both 1-6 keV and 2-6 keV energy intervals. This shows that there is no evidence for cosine signal from dark matter interactions in the combined ANAIS-112/COSINE-100 data. Our analysis codes have also been made publicly available. I Introduction Even though we know for more than 90 years that most of the mass of our galaxy is made up of cold dark matter, the identity of the dark matter candidate(s) is still a mystery 8 . One of the most widely studied dark matter candidate is the Weakly interacting massive particle (WIMP), since such a candidate with weak-scale interactions has a relic abundance which matches that of the observed dark matter 21 . A large number of experiments (refered to as direct detection experiments) have been looking for signatures of dark matter scattering off atomic nuclei for close to four decades 15 . Among the plethora of direct detection experiments, only the DAMA/LIBRA experiment has reported positive evidence for dark matter interactions. The characteristic signature seen in this experiment is an annual modulation in the residual rates, which has the right characteristics expected from dark matter interactions 14 . This signal was first announced in 1998 6 and the statistic significance of the annual modulation using the latest data with a total exposure of 2.46 ton-year covering more than 14 years is about 12.9 \sigma 7 . However, this signal could not be independently confirmed by other direct detection experiments 5 as well as indirect detection experiments 13 . To resolve this long-standing DAMA anomaly, two experiments namely COSINE-100 and ANAIS-112 have been designed using the same target material as DAMA, viz. NaI. Recently, both COSINE-100 and ANAIS-112 presented results on the search for an annual modulated based on joint analysis of their data covering three years 10 (C25, hereafter). The best-fit values for the annual modulation modulation found using least-squares and Markov Chain Monte Carlo based regression analysis were equal to ( -0.0002\pm 0.0026 ) cpd/kg/keV and ( 0.0021\pm 0.0028 ) cpd/kg/keV in 1-6 keV and 2-6 keV intervals, respectively. Therefore, these results do now show any evidence for an annual modulation to support the DAMA claim. In this work, we complement the analysis in 10 , by carrying out a Bayesian model comparison test between the hypothesis that the combined COSINE-100/ANAIS-112 data is consistent with annual modulation due to a sinusoidal signal (expected from dark matter interactions), compared to the hypothesis that the data only contain a constant value due to the background. This work is a follow-up to our previous works on Bayesian comparison for annual modulation in multiple direct dark matter detection experiments such as DAMA 17 ; 9 , COSINE-100 18 , ANAIS-112 19 . Other applications of Bayesian model comparison in analysis of DAMA/LIBRA data can also be found in 22 . This manuscript is structured as follows. A brief summary of the combined COSINE-100/ANAIS-112 results discussed in C25 can be found in Sect. II . Our analysis and results of the same data is described in Sect. III . We conclude in Sect. IV . II Summary of C25 We now recap the results from the analysis in C25, where more details can be found. COSINE-100 1 and ANAIS-112 3 are two direct dark matter experiments which use Na(Tl) as the target and have been designed to test the long-standing DAMA annual modulation claim 7 . The COSINE-100 experiment is located in the Yangyang underground laboraory in Korea and consists of five NaI(Tl) detectors with a total mass of 61 kg. COSINE-100 has been taking data since October 2016. Similarly, the ANAIS-112 experiment is located in the Canfranc Underground Laboratory in Spain and consists of nine NaI(Tl) crystals of mass 12.5 kg each, with a total mass of 112.5 kg. ANAIS-112 has been taking data since August 3, 2017. More details of both these experiments along with background radioactive rates can be found in 1 and 3 , for COSINE-100 and ANAIS-112, respectively. The ANAIS-112 and COSINE-100 collaborations have published results from about three years of data and could not support the DAMA/LIBRA annual modulation claim 2 ; 12 . Since both these experiments have similar sensitivity, a combined analysis of both these datasets from the first three years of data was done in C25. Only the single-hit events were used for analysis. Each dataset was background subtracted to obtain the residuals, which were then combined to do a search for annual modulation. A separate search was done in both 1-6 keV and 2-6 keV energy intervals. Details of the background models used for both COSINE-100 and ANAIS-12 can be found in C25 and references therein, and will be skipped here for brevity. The data was sub-divided into 15-day bins for both the experiments. The total exposure analyzed was 485 kg-year. The background subtracted residuals were fit to the following function: R(t)=S_{m}\cos(\omega(t-t_{0})), (1) where t_{0} is the phase (kept fixed at June 2nd), \omega=2\pi/T for T=1 year, and S_{m} is the modulation amplitude. The best-fit values of S_{m} were obtained using both frequentist and MCMC analysis. The best-fit values were equal to ( -0.0002\pm 0.0026 ) cpd/kg/keV and ( 0.0021\pm 0.0028 ) cpd/kg/keV in 1-6 keV and 2-6 keV intervals. Therefore, the modulation amplitudes are consistent with no annual modulation to within 1\sigma . Finally, a similar analysis using a simple combination of six years of data from COSINE-100 11 and ANAIS-112 4 with a total exposure of 1 ton-year was also carried out in C25. These results confirm the previous results of no modulation, with modulation amplitudes having values of ( 0.0005\pm 0.0019 ) cpd/kg/keV and ( 0.0027\pm 0.0021 ) cpd/kg/keV in 1-6 keV and 2-6 keV, respectively. III Analysis and Results III.1 Brief Primer on Bayesian Model Comparison We provide a brief prelude to Bayesian model comparison. and refer the reader to more details in Refs. 25 ; 23 ; 16 ; 20 . To compare the efficacy of a ( M_{2} ) against another model ( M_{1} ), we define the Bayes factor ( B_{21} ) given by: B_{21}=\frac{\int P(D|M_{2},\theta_{2})P(\theta_{2}|M_{2})\,d\theta_{2}}{\int P(D|M_{1},\theta_{1})P(\theta_{1}|M_{1})\,d\theta_{1}}, (2) where P(D|M_{2},\theta_{2}) is the likelihood for the model M_{2} given the data D , and P(\theta_{2}|M_{2}) is the prior for the parameter vector \theta_{2} of the model M_{2} . The denominator refers to the same, but for model M_{1} . If B_{21} is greater than 1.0, then model M_{2} is preferred over M_{1} and vice-versa. The significance can be evaluated using Jeffreys’ scale 16 . For our analysis, M_{2} refers to the hypothesis that the residual data can be described using Eq. 1 , whereas M_{1} refers to the hypothesis