Search for dark matter axions at 1-2 micro-eV with a RADES-DarkQuantum haloscope at CERN

arXiv:2611.07806 · hep-ex, hep-ph · Submitted 2026-11-05 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.

Kai: We're looking at the paper: "Search for dark matter axions at 1-2 micro-eV with a RADES-DarkQuantum haloscope at CERN".

Mira: Here's the gist of the paper. The paper describes a proposal to employ CERN’s M1 superconducting magnet to deploy a haloscope to perform a search for galactic axion and axion-like particle dark matter providing continuous coverage from approximately 1.2 to 1.8 µeV Under conservative assumptions, the experiment can reach the QCD axion band over a significant fraction of the scanned range, while optimistic detector performance would allow sensitivity to benchmark models such as KSVZ and DFSZ over a large part of the explored mass interval With the current long shut-down of CERN’s accelerators providing a unique and timely window, such an experiment will deliver high-impact results even in conservative scenarios and potentially delivering a definitive test of the axion dark matter paradigm in the unexplored 1–2 µeV mass range.

Kai: Experimental Setup The proposed RADES-DQLF experiment follows the standard axion haloscope concept described in section 1.

Mira: The cavity resonance is tuned across the target axion-mass range, while the resulting microwave signal is extracted and amplified by a low-noise receiver.

Kai: The experimental design builds on the cavity concepts and technologies developed within RADES and DarkQuantum, adapted to the specific magnetic, cryogenic and mechanical constraints of M1.

Title and authors: Kai: They lay out the experimental setup by following the standard axion haloscope concept, tuning a cavity resonance across the target mass range and extracting that resulting microwave signal with a low-noise receiver <ref:2611.07806#pg1>.

Mira: The design builds on existing cavity concepts from RADES and DarkQuantum, but they had to adapt it to the specific magnetic, cryogenic, and mechanical constraints of the M1 magnet <ref:2611.07806#pg2>.

Lev: They describe the resonant cavities using a quasi-cylindrical geometry based on the RADESBabyIAXO cavity concept, with DQLF-one having dimensions similar to that largest cavity <ref:2611.07806#pg3>.

Kai: They also discuss options for coating the cavity body; they are starting with stainless steel AISI 316L coated with ten microns of electroplated copper, but they have an option to deposit a niobium titanium layer on top of that copper for a ten-fold increase in cavity quality factor Q0 <ref:2611.07806#pg2>.

Mira: The sensitivity relies heavily on the signal-to-noise ratio given by the Dicke radiometer equation, and their receiver chain focuses on a superconducting quantum interference device based first stage amplifier <ref:2611.07806#pg3>.

Lev: They state that microstrip SQUID amplifiers can operate near the quantum limit with noise temperatures as low as five hundred millikelvin at five hundred nineteen megahertz <ref:2611.07806#pg3>.

The paper's summary: Kai: The paper details the integration at CERN’s M1 magnet, which provides a nominal horizontal magnetic field of three tesla at its center <ref:2611.07806#pg2>.

Mira: They outline the data taking protocol, starting with a setup phase, then tuning the cavity resonance, measuring reflection and transmission coefficients to determine cavity parameters Q and beta <ref:2611.07806#pg4>.

Lev: Their sensitivity projections use the standard haloscope signal-to-noise formalism, showing that even under conservative assumptions they expect to probe unexplored parameter space around one point two to two micro-electron volts <ref:2611.07806#pg6>.

Kai: They also mention using three complementary cavity geometries to achieve continuous coverage over the frequency range from approximately two hundred ninety-four to four hundred forty-five megahertz Improvement one <ref:2611.07806#pg1>.

Mira: And they have two acquisition strategies they support: real time spectral averaging and time domain streaming, which lets them adapt to different physics goals and data rates Improvement two <ref:2611.07806#pg1>.

The paper's improvements: Kai: So, the paper wraps up by saying this RADES-DQLF experiment directly addresses a gap in experimental searching for axions in that mass range <ref:2611.07806#pg7>.

Mira: They conclude that deploying three complementary cavity geometries allows for continuous coverage of the targeted mass range, following the same modular strategy used for the full BabyIAXO implementation <ref:2611.07806#pg7>.

Lev: From my side, I see the real value in how they handle calibration and monitoring; they include periodic calibration via a reference tone measurement which triggers a new full calibration when required to ensure stability of the receiver chain Improvement three <ref:2611.07806#pg3>.

Kai: That systematic check is crucial because it maintains the integrity of those sensitivity projections Improvement three <ref:2611.07806#pg3>.

Mira: And they also have an automatic re-coupling procedure that adjusts the port loop angle if coupling deviates from its target value, making sure each frequency point is measured under controlled coupling conditions Improvement four <ref:2611.07806#pg3>.

Lev: That control over coupling is important because accurate sensitivity calculations depend on knowing exactly how well coupled each cavity is Improvement four <ref:2611.07806#pg3>.

Kai: Finally, they monitor the total added noise temperature Tadd and gain Gtotal in real time, where the slope of Pout versus Tload yields the total gain and the intercept provides Tadd Improvement five <ref:2611.07806#pg3>.

Mira: That real time monitoring lets them immediately detect systematic degradations in their setup Improvement five <ref:2611.07806#pg3>.

Lev: And their data treatment follows a pipeline similar to HAYSTAC but uses a library called halolib, which allows for a transparent protocol and lets them use ROOT fitting procedures early on to find bad runs Improvement six <ref:2611.07806#pg3>.

Kai: So, the full picture of this paper is that they’ve designed a robust system using existing technology at CERN M1 to systematically search for QCD axions in the unexplored one point two to two micro-electron volt mass range <ref:2611.07806#pg2>.

Mira: It suggests that with the current shut down of CERN's accelerators, this experiment will deliver high impact results even under conservative assumptions, potentially testing the axion dark matter paradigm in that unexplored range <ref:2611.07806#pg2>.

Lev: The real implication for hardware development is seeing how these sensitivity projections translate when you factor in the actual overhead of running this on real hardware Table three <ref:2611.07806#pg3>.

Kai: Right, so we're done with this paper, but next up we'll look at some work on chromatic calorimetry <ref:2611.07806#pg7>.

Conclusion: Kai: So we've been looking at the RADES-DarkQuantum haloscope at CERN M1 <ref:2611.07806#pg4> and this paper, "Search for dark matter axions at one-two micro-eV with a RADES-DarkQuantum haloscope at CERN," it's about setting up a way to look for those axions, which are these hypothetical particles that could make up dark matter.

Mira: Exactly. The core idea is using the M1 magnet's magnetic field to create these resonant cavities, tuning them across a specific frequency range from two hundred ninety-four megahertz to four hundred forty-five megahertz <ref:2611.07806#pg4>.

Lev: And from an error correction side, the setup means we're dealing with a very large system here, so the sensitivity projections are key because we have to figure out how much noise that real hardware can handle Table three.

Kai: The authors show that even if they use conservative assumptions about detector performance, they still expect to find something interesting in that one point two to two micro-electron volt mass range <ref:2611.07806#pg6>.

Mira: That range is really important because it's where some of the benchmark models, like KSVZ and DFSZ, predict these axions might live <ref:2611.07806#pg3>.

Lev: If we were to run this on real hardware, the overhead of that SQUID-based low noise readout would be a major factor in how fast we could actually sweep through all those frequency points Table three.

Kai: The setup is pretty modular, using three different cavity geometries to get continuous coverage across that mass range, which they call DQLF-one to DQLF-three <ref:2611.07806#pg7>.

Mira: That modular approach lets them systematically search for the QCD axion band while also looking at those benchmark models simultaneously <ref:2611.07806#pg4>.

Lev: And they've included some really good technical details about how they handle coupling deviations automatically, which is something we always worry about when you build hardware that needs to be perfectly tuned Improvement four.

Kai: They also show how they monitor the total added noise temperature and gain in real time so they can catch any systematic issues right away Improvement five.

Mira: So the conclusion is that this experiment fills a gap, giving us a chance to test the dark matter axion idea with high impact results even if things don't turn out perfectly optimistic <ref:2611.07806#pg7>.

Lev: For me, it shows how much experimental control you need when you are trying to get a signal this faint in a cryogenic environment Table three.

Kai: It’s definitely a solid step forward for searching for these particles in that specific mass window <ref:2611.07806#pg2>.

Mira: Next up, we're looking at chromatic calorimetry and how it validates energy resolution in particle detection Background Paper.

Alberto Arcusa Puentea, Sergio Arguedas Cuendisl, Marcos Buetas Matac, Sergio Calatronib, Cristian Cogollosd Alejandro D’az-Morcilloe Babette Döobrichd Jacob Eggef, Elisa Gabbrielliid Javier Galindo Guarcha Jose Maráa Garc⸚-Barcel༹og Benito Gimenoh, Yikun Gug Sophia Hollickg Igor Garcia Irastorzag David Kittlingerd Adrian Lambies-Asensioh Axel Lindnerf Manuel Meyeri Adriáan Muónoz-Ur’ane Jose R. Navarro-Madride Tobias Ortmannd Sagar Paulc Angel Rodriguezg Kristof Schmiedenj Laura Seguig Wolfgang Wernsdorferc,k Mina Yamig

Instituto Tecnologico de Arag༹on · CERN - European Organization for Nuclear Research · Karlsruhe Institute of Technology · Max-Planck-Institut fur Physik (Werner-Heisenberg-Institut) · Universidad Politecnica de Cartagena · Deutsches Elektronen-Synchrotron DESY · Centro de Astropart༱culas y Física de Altas Energ༱as (CAPA) Universidad de Zaragoza · IFIC Valencia · CP3-Origins University of Southern Denmark · Rheinische Friedrich-Wilhelms-University · Institute of Quantum Materials and Technologies

hep-ex, hep-ph

Submitted: 2026-11-05

Updated: 2026-11-05

Comments: prepared for submission to JHEP, 42 pages

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 82/100

The gist: The paper describes a proposal to employ CERN’s M1 superconducting magnet to deploy a haloscope to perform a search for galactic axion and axion-like particle dark matter providing continuous…

Terminology

Summary

The paper describes a proposal to employ CERN’s M1 superconducting magnet to deploy a haloscope to perform a search for galactic axion and axion-like particle dark matter providing continuous coverage from approximately 1. p. 12 to 1.8 µeV Under conservative assumptions, the experiment can reach the QCD axion band over a significant fraction of the scanned range, while optimistic detector performance would allow sensitivity to benchmark models such as KSVZ and DFSZ over a large part of the explored mass interval With the current long shut-down of CERN’s accelerators providing a unique and timely window, such an experiment will deliver high-impact results even in conservative scenarios and potentially delivering a definitive test of the axion dark matter paradigm in the unexplored 1–2 µeV mass range. p. 1

Experimental Setup

The proposed RADES-DQLF experiment follows the standard axion haloscope concept described in section 1. The cavity resonance is tuned across the target axion-mass range, while the resulting microwave signal is extracted and amplified by a low-noise receiver. p. 5 The experimental design builds on the cavity concepts and technologies developed within RADES and DarkQuantum, adapted to the specific magnetic, cryogenic and mechanical constraints of M1. p. 5

Resonant Cavities and Tuning Mechanism

The cavity designs are based on a quasi-cylindrical geometry derived from the RADESBabyIAXO cavity concept. The first cavity (called DQLF-1) has the same cross-sectional dimensions as the largest RADES-BabyIAXO cavity. p. 5 The tuning mechanism is based on two rotating metal plates of rectangular geometry aligned with the longitudinal axis and placed inside the resonant volume. p. 6

Superconducting Coating Option

The baseline fabrication approach for the DQLF cavities is to manufacture the body of the cavities with stainless steel AISI 316L, coated with an electroplated copper layer of a minimum thickness of 10 µm. The option of depositing an NbTi layer on top of the existing copper layer is planned to achieve at least a tenfold increase in the cavity Q0. p. 9

SQUID-Based Low-Noise Readout

The sensitivity of a microwave haloscope is fundamentally determined by its signal-to-noise ratio (SNR) as given by the Dicke radiometer equation. The receiver chain within the DQLF programme focuses on the implementation of a superconducting quantum interference device (SQUID) based first-stage amplifier. p. 5 Microstrip SQUID amplifiers (MSAs) can operate near the quantum limit, with noise temperatures as low as ∼ 50 mK at 519 MHz. p. 5

Integration at CERN M1 magnet

The M1 magnet is a large-scale superconducting dipole facility located in the H2 test beam area within CERN’s North Area Experimental Hall (EHN1). The magnet provides a nominal horizontal magnetic field (i. p. 3e. along the axis of the coils) of 3 T at its center. p. 9

Data Taking Protocol and Organization

The planned data-taking sequence for DQLF is summarized in the flowchart of Fig. 17. The run will open with a setup phase, followed by tuning the cavity resonance, measuring reflection and transmission coefficients, and determining the cavity parameters Q and β. p. 5

Sensitivity Projections

The projected sensitivity of the proposed experiment has been estimated using the standard haloscope signal-to-noise formalism. The resulting projected sensitivity is shown in Fig. p. 9 19. Even under conservative assumptions, our proposal is expected to probe unexplored parameter space in the mass range around 1. p. 12–2 µeV. p. 0

Conclusions

The proposed RADES-DQLF experiment directly addresses this experimental gap. The deployment of three complementary cavity geometries allows continuous coverage of the targeted mass range, following the same modular strategy foreseen for the full BabyIAXO implementation. p. 8

Acknowledgments

It is a pleasure to thank the CERN technical services along with several individuals for helping in the preparation of work. This work was supported by the European Research Council (ERC) under the European Union’s Horizon Europe programme (ERC Synergy Grant DarkQuantum, grant agreement No. p. 4 101118911). The work also profited from strong support by CERN’s PBC (Physics Beyond Collider) initiative. p. 2

References

[3] S. Weinberg, “A New Light Boson? p. 2,” Phys. Rev. Lett. 40 (1978) 223–226

[5] M. Tegmark, A. Aguirre, M. Rees, and F. Wilczek, “Dimensionless constants, cosmology and other dark matters,” Phys. p. 4Rev. D73 (2006) 023505

[11] P. Sikivie, “Experimental tests of the invisible axion,” Phys. p. 3 Rev. Lett. 51 (1983) 1415

[23] CAST Collaboration, A. A. Melcon´ et al., “First results of the CAST-RADES haloscope search for axions at 34. p. 067 µeV,” JHEP 21 (2020) 075

[39] D. Larbalestier, “Metallurgy of NbTi. p. 7” https://boulderschool.yale.edu/sites/default/files/files/larbalestier3.pdf, 2000

[48] CERN, “https://magnet-m1.web.cern.ch/Magnetics/Magnetics.html.”

Table 3

Parameter Optimistic Conservative Physical temperature Tphys 30 mK 100 mK Receiver-added noise nadd 3.5 photons 8. p. 35 photons Tadd at 290 MHz ∼50 mK ∼120 mK Cryogenic circulator No Yes Unloaded cavity quality factor 10 × Q0(f) 0. p. 82 × Q0(f) Coupling β βopt(f) 2 Effective exposure 180 days 180 days Confidence level 90% C. p. 4L.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Continuous coverage of axion mass range: The deployment of three complementary cavity geometries (DQLF-1 to DQLF-3) allows for continuous coverage over the frequency range from approximately 294 to 445 MHz by overlapping tuning ranges. This enables a systematic search across the target 1–2 µeV window.

  2. Optimized scanning strategy: The data taking protocol supports two acquisition strategies, including Real-time spectral averaging and Time-domain streaming, allowing the system to adapt to different physics goals and data rates. This flexibility allows for choosing a frequency resolution that is optimized for specific dark matter models.

  3. Robust calibration and monitoring: The protocol includes periodic calibration via a reference tone measurement which triggers a new full calibration when required, ensuring that the stability of the receiver chain is monitored through periodic calibration-tone measurements. This systematic check maintains the integrity of the sensitivity projection.

  4. Automated coupling recovery: The system incorporates an automatic re-coupling procedure that adjusts the port loop angle if coupling deviates from its target value. This ensures that each frequency point is measured under controlled coupling conditions, which is crucial for accurate sensitivity calculations.

  5. System noise characterization: The system monitors the total added noise temperature Tadd and gain Gtotal in real time, as shown by Fig 9e, where the slope of Pout versus Tload yields the total gain Gtotal, while the intercept provides Tadd. This allows for immediate detection of systematic degradations.

  6. Data analysis framework: The data treatment follows a pipeline similar to HAYSTAC collaboration but utilizes a library called halolib. This framework allows for a transparent, collaboration wide protocol and enables the use of ROOT fitting procedures early in the process to identify 'bad runs'.

Sources

Related papers