MAGIS: Atomic Interferometry & Beyond

• MAGIS is a suite of atomic interferometry projects that enable ultralight dark matter detection, mid-band gravitational wave searches, and advanced stellar abundance studies.
• It employs vertical light-pulse atom interferometers with large momentum transfer and state-of-the-art noise mitigation to achieve exceptional sensitivity.
• Future designs aim to extend baseline lengths and enhance sensitivity, promising breakthroughs in quantum mechanics, astrophysics, and gravitational sensing.

MAGIS (Matter-wave Atomic Gradiometer Interferometric Sensor) encompasses a suite of next-generation atomic interferometry projects and instrumentation with applications in fundamental physics, quantum sensing, gravitational-wave astronomy, and astrophysical spectroscopy. Multiple research programs share the MAGIS acronym, including large-scale atom interferometry for ultralight dark-matter detection and mid-band gravitational-wave observation (as exemplified by MAGIS-100 and its successors at Fermilab), high-precision abundance surveys in stellar astrophysics (Measuring Abundances of red super Giants with Infrared Spectroscopy), and advanced AI-based frameworks for multi-agent code evolution in software engineering. This article presents a comprehensive technical overview focusing on the atomic-physics-centric MAGIS-100 program and its extensions, with attention to key physical principles, experimental implementations, methodologies, noise mitigation, and future prospects in the broader context of long-baseline atomic sensors.

1. Scientific Motivation and Multi-Domain Applications

The original MAGIS concept as realized at Fermilab is motivated by three intertwined scientific goals: (1) detection of ultralight bosonic dark matter in the $10^{-22}\,\mathrm{eV}$ to $10^{-3}\,\mathrm{eV}$ range via couplings to fundamental constants or new forces, (2) tests of quantum mechanics and coherence phenomena over macroscopic separations and timescales, and (3) gravitational-wave detection in the so-called “mid-band” (0.1–10 Hz), which is inaccessible to both terrestrial (LIGO, Virgo) and space-borne (LISA) laser interferometers (Coleman, 2018, Abe et al., 2021, Balaz et al., 27 Mar 2025). Ultralight dark matter (ULDM) scenarios predict oscillatory ground-state fields that induce time-dependent shifts in atomic energy levels (e.g., via couplings $d_{m_e}$ to the electron mass or $d_e$ to the fine-structure constant) or generate anomalous differential accelerations (probed by comparing distinct isotopes or spin states).

In other contexts, the MAGIS project acronym refers to high-precision stellar-chemical studies exploiting red supergiants (RSGs) as luminous abundance tracers (Taniguchi et al., 17 Jan 2025), and to advanced machine learning multi-agent frameworks for automated resolution of complex software issues (Tao et al., 2024). These parallel lines reflect the acronym’s versatile utility across physics, astrophysics, and computation, though the atomic-physics program is the central referent.

2. Experimental Design and Atom Interferometer Configurations

The core of MAGIS-100 is a vertical light-pulse atom interferometer installed in the 100 m NuMI shaft at Fermilab (Coleman, 2018, Abe et al., 2021, Balaz et al., 27 Mar 2025). Two ultra-cold atomic ensembles (typically ${}^{87}\mathrm{Sr}$, ${}^{87}\mathrm{Rb}$, or ${}^{171}\mathrm{Yb}$) are launched in opposing directions, forming spatially separated, freely-falling proof masses (“gradiometer” configuration). The interferometry sequence employs a Mach–Zehnder geometry with a $\pi/2$–$\pi$–$\pi/2$ light-pulse sequence, realized through single-photon clock transitions ($10^{-3}\,\mathrm{eV}$0 in Sr at 698 nm or corresponding transitions in Yb) or via two-photon Bragg/Raman processes. Large-momentum-transfer (LMT) optics ($10^{-3}\,\mathrm{eV}$1 initially; $10^{-3}\,\mathrm{eV}$2 as a target) substantially enhance the enclosed spacetime area. The atomic states are manipulated and read out using frequency-stabilized, cavity-locked laser systems with sub-Hz linewidth.

Instrumental parameters for MAGIS-100 are summarized as follows:

• Baseline $10^{-3}\,\mathrm{eV}$3 m, extendable in future detectors to 1 km (MAGIS-1000) or to 40,000 km in space-based proposals (MAGIS-space).
• Interrogation time per half-interferometer $10^{-3}\,\mathrm{eV}$4–$10^{-3}\,\mathrm{eV}$5 s (scaling up to $10^{-3}\,\mathrm{eV}$6–$10^{-3}\,\mathrm{eV}$7 s for ultimate designs).
• Atom fluxes $10^{-3}\,\mathrm{eV}$8 atoms/s, with shot-noise-limited phase sensitivity targeted at $10^{-3}\,\mathrm{eV}$9 (Coleman, 2018).

Beam-propagation aberrations and wavefront errors are systematically monitored and minimized using a custom CMOS-based profiling system with principal component analysis (PCA) post-processing, achieving $d_{m_e}$0 profile-reconstruction errors and sub-pixel centroid stability (Jachinowski et al., 2022).

3. Theoretical Framework and Signal Characterization

The atom interferometer’s response to external perturbations is set by the phase-shift formula: $d_{m_e}$1 for acceleration $d_{m_e}$2, where $d_{m_e}$3, and $d_{m_e}$4 is the half-interferometer time. For differential measurements across separated ensembles,

$d_{m_e}$5

for a passing gravitational wave of strain $d_{m_e}$6 in the long-wavelength regime ($d_{m_e}$7), and

$d_{m_e}$8

for DM-induced time-dependent accelerations or frequency shifts.

Key observables include:

• Acceleration sensitivity per shot: $d_{m_e}$9.
• Strain-noise spectral density: $d_e$0.
• For scalar dark matter, periodic variations in fundamental constants modulate atomic transition frequencies; for vector-coupled ($d_e$1) dark matter, differential accelerations scale with $d_e$2; for axion-like DM, spin-torque terms can be probed via Ramsey–Bordé configurations (Zhou et al., 2024, Badurina et al., 1 May 2025).

Long interrogation times and large baselines enable sensitivity to GW strains as low as $d_e$3–$d_e$4 (MAGIS-100) and ultimately $d_e$5 (MAGIS-1000) in the mid-band (Coleman, 2018, Abe et al., 2021).

4. Environmental Noise, Systematics, and Mitigation Strategies

Environmental and instrumental noise sources represent critical limitations for all long-baseline atom interferometers (Mitchell et al., 2022, Glick et al., 2023, Jachinowski et al., 2022). Key noise mechanisms and mitigation methods include:

• Laser Phase Noise: Suppressed by common-laser interrogation (differential readout) and ultra-stable cavity-locked lasers; the gradiometer configuration strongly rejects common-mode noise (Coleman, 2018).
• Seismic and Platform Vibrations: Active and passive vibration isolation platforms, plus underground installation, reduce vibration-coupled phase noise. Direct vibration noise maps to strain noise as $d_e$6, with $d_e$7 the displacement spectrum.
• Gravity-Gradient Noise (GGN): GGN is due to fluctuating local gravitational fields from ambient seismic and atmospheric density waves, modeled by Rayleigh-wave couplings and characterized by suppression factors ($d_e$8–$d_e$9 with multi-interferometer subtraction) (Mitchell et al., 2022). Site selection (depth, shaft design) and auxiliary seismometry are essential for control.
• Magnetic Fields and Light Shifts: Magnetically insensitive (clock) states, shielding, and state alternation suppress Zeeman and AC Stark contributions.
• Wavefront Aberrations: Spatial filtering, in-situ beam diagnostics, and atom trajectory matching minimize induced systematic errors.
• Coriolis Forces: A tunable optical “pivot-point” geometry simultaneously adjusts the beam and atom-velocity vector, eliminating misalignment and associated phase errors even over 100 m baselines (Glick et al., 2023).

A comprehensive environmental DAQ system continuously logs temperature, humidity, seismic activity, and air pressure to support both real-time feedback and post-facto corrections.

5. Sensitivity Projections, Physics Reach, and Comparative Performance

MAGIS-100 and the evolving suite of MAGIS detectors enable world-leading sensitivity across multiple search channels:

• Ultralight Dark Matter: Reach scalar coupling strengths as low as ${}^{87}\mathrm{Sr}$0 at ${}^{87}\mathrm{Sr}$1, surpassing previous bounds by 1–2 orders of magnitude. B–L couplings can be constrained to ${}^{87}\mathrm{Sr}$2 for ${}^{87}\mathrm{Sr}$3 (Coleman, 2018, Zhou et al., 2024).
• Gravitational Waves: Achievable strain sensitivities in the mid-band (${}^{87}\mathrm{Sr}$4–${}^{87}\mathrm{Sr}$5 Hz) permit detection of sources such as binary neutron stars, massive black-hole mergers, inflationary relics, and white-dwarf binary inspirals inaccessible to both LIGO and LISA (Sala et al., 22 Oct 2025, Holgado et al., 2020, Balaz et al., 27 Mar 2025).
• Experimental Quantum Mechanics: Large-arm separations (${}^{87}\mathrm{Sr}$650 m) and long coherence times probe macroscopic superpositions and decoherence effects at unprecedented scales (Abe et al., 2021).

MAGIS-space proposals extend this capability to detect gravitational signatures of compact dark-matter clumps (${}^{87}\mathrm{Sr}$7–${}^{87}\mathrm{Sr}$8 kg) and to search gravitationally for ultralight DM with improved reach over laser interferometers due to the gradiometric Einstein redshift term (Badurina et al., 1 May 2025).

In astrophysics, the MAGIS infrared-spectroscopy initiative establishes abundance gradients for key elements (Fe, ${}^{87}\mathrm{Sr}$9-elements, s-process) in young populations with ${}^{87}\mathrm{Rb}$0 dex internal precision, validating RSGs as robust chemical tracers (Taniguchi et al., 17 Jan 2025).

For machine learning in code evolution, the MAGIS (“Multi-Agent GitHub Issue Solution”) framework surpasses monolithic LLMs (e.g., GPT-4) by an order of magnitude in GitHub issue resolution, leveraging explicit multi-agent planning, context filtration, code localization, and QA workflows (Tao et al., 2024).

6. Roadmap: Future Detectors and Networked Observatories

MAGIS-100 is explicitly a pathfinder for kilometer-scale and beyond deployments. The MAGIS-1000 and MAGIS-lambda proposals target ${}^{87}\mathrm{Rb}$1–2 km vertical baselines, exploiting improvements in atom flux, LMT splitting, and noise rejection to deepen gravitational-wave strain sensitivity below ${}^{87}\mathrm{Rb}$2 in the 1 Hz regime (Coleman, 2018, Abe et al., 2021, Balaz et al., 27 Mar 2025). Horizontal networked observatories (TVLBAI/AION) at underground sites (e.g., SURF, CERN PX46) are envisaged for triangulated cross-correlation, sky localization, and global ultralight dark-matter surveys.

Space-based configurations such as MAGIS-space, with 40,000-km baselines and long interrogation times (${}^{87}\mathrm{Rb}$3 s), open discovery potential for new classes of dark-matter substructure and mid-band gravitational-wave sources (including pre-merger eccentricity measurements for GW190521-like binaries) (Badurina et al., 1 May 2025, Sala et al., 22 Oct 2025, Holgado et al., 2020).

In all these designs, progressive refinement of environmental suppression, multi-interferometer "string of pearls" noise subtraction, and quantum-limited atom optics underpin continued increases in experimental reach.

7. Contextual Expansions: MAGIS in Astrophysics and AI

The “Measuring Abundances of red super Giants with Infrared Spectroscopy” (MAGIS) project marks a separate axis of innovation, applying infrared spectral synthesis to luminous red supergiants across the Milky Way and external galaxies (Taniguchi et al., 17 Jan 2025). The pipeline achieves internal abundance precisions (Fe, Mg: ${}^{87}\mathrm{Rb}$4–${}^{87}\mathrm{Rb}$5 dex) rivaling optical FGK analyses, with robust validation against Cepheid-based gradients and systematic error budgets evaluated via line-by-line corrections and NLTE modeling.

Within computational science, MAGIS: "LLM-Based Multi-Agent Framework for GitHub Issue Resolution" introduces a collaborative, role-based orchestration architecture for complex repository-level code evolution, achieving significant improvements in test-passing patch rates and localization accuracy over baseline LLMs through explicit plan–retrieve–develop–QA agent cascades (Tao et al., 2024).

---

Table: MAGIS Program Dimensions

Domain	Core Target/Function	Key Citation
Atom interferometry (MAGIS-100/1000)	Dark matter, GWs, quantum mechanics	(Coleman, 2018)
Astrophysical spectroscopy (MAGIS-RSG)	IR-based high-precision stellar abundances	(Taniguchi et al., 17 Jan 2025)
Multi-agent AI for code (MAGIS-ML)	Automated multi-agent code issue resolution	(Tao et al., 2024)
Space-based GW/DM detection	GW astronomy, gravitational DM signatures	(Badurina et al., 1 May 2025)

---

• (Coleman, 2018) MAGIS-100 at Fermilab
• (Abe et al., 2021) Matter-wave Atomic Gradiometer Interferometric Sensor (MAGIS-100)
• (Balaz et al., 27 Mar 2025) Long-Baseline Atom Interferometry
• (Sala et al., 22 Oct 2025) Detecting White Dwarf Binary Mergers with Gravitational Waves
• (Holgado et al., 2020) Dynamical Formation Scenarios for GW190521 and Prospects for Decihertz Gravitational-Wave Astronomy
• (Badurina et al., 1 May 2025) Detecting gravitational signatures of dark matter with atom gradiometers
• (Mitchell et al., 2022) MAGIS-100 Environmental Characterization and Noise Analysis
• (Zhou et al., 2024) Ytterbium atom interferometry for dark matter searches
• (Jachinowski et al., 2022) Beam Profiling with Noise Reduction for MAGIS-100
• (Glick et al., 2023) Coriolis Force Compensation and Laser Beam Delivery for 100-Meter Baseline Atom Interferometry
• (Tao et al., 2024) MAGIS: LLM-Based Multi-Agent Framework for GitHub Issue Resolution
• (Taniguchi et al., 17 Jan 2025) MAGIS (Measuring Abundances of red super Giants with Infrared Spectroscopy) project I