Aalto University team develops machine learning-optimized pulse to search for dark matter

A new microwave detection protocol improves the search for axions and dark photons. Yu-Han Chang and colleagues at Aalto University, the University of Chicago, and the University of Zaragoza employ a superconducting transmon qubit and a dual-cavity system. This protocol addresses the limitations of traditional qubit control by using machine learning-optimized pulses to achieve single-photon detection performance. We establish a new exclusion limit for the dark photon model with a dynamic mixing angle sensitivity of approximately 1×10.^-14 At 5.051GHz. These advances offer the potential for faster and more scalable microwave quantum sensors in the continuing quest to understand the nature of dark matter.

Improving dark photon sensitivity through optimized qubit control using machine learning

Sensitivity to dynamic mixing angle reaches approximately 1×10^-14 At 5.051GHz, it is a significant improvement over previous dark photon detection protocols. Previous methods have had difficulty reliably identifying signals below this level due to qubit control challenges and noise, but this threshold exceeds those limits and allows exploration of previously inaccessible regions of parameter space for dark matter candidates. Dark photons are hypothetical particles that have been proposed as intermediaries between ordinary matter and dark matter, and are often explored through potential mixing with photons. This “dynamic mixing” allows the dark photons to weakly interact with the electromagnetic field, and the detector’s sensitivity to this mixing angle is an important figure of merit. Previous experiments were limited by qubit decoherence and cavity lifetime, and struggled to achieve the precision needed to explore parameter spaces smaller than 1×10.^-14. This new protocol overcomes these hurdles and opens new avenues of investigation.

Superconducting qubits and dual-cavity systems establish a path to faster and more scalable microwave quantum sensors in the ongoing search for dark matter. Even though the qubit coherence time was significantly reduced and the storage cavity lifetime was reduced to 10 microseconds, the single-photon detection performance remained comparable to previous implementations. This is an important achievement because shorter coherence times typically increase noise and reduce signal fidelity. The dual-cavity system is designed to conserve photons, allowing them to interact longer with the qubit. Qubits, which are superconducting transmons, serve as sensitive probes of the number of photons within the storage cavity. As confirmed by quantum process tomography demonstrating high-fidelity qubit control, the machine learning-optimized pulses extended the operating bandwidth to approximately 20MHz and suppressed gate errors by two orders of magnitude compared to standard square pulses. Quantum process tomography is a technique used to characterize the performance of quantum gates, ensuring accurate and reliable operation. The increased bandwidth allows for faster data acquisition and reduces gating errors, improving the overall signal-to-noise ratio.

The increased qubit-cavity coupling provided by this approach reduces experimental time and improves the sensitivity of future studies. Stronger coupling means the qubit and cavity interact more effectively, resulting in a faster response and more pronounced signal. Hidden Markov models, a statistical method for modeling sequences of events, produced background rates on the order of $\mathcal{O}$ Hz, allowing the establishment of exclusion limits for the dark photon model. Hidden Markov models are used to distinguish between real signals and background noise, improving measurement accuracy. Data processing of the parity measurements, binary results showing the qubit states revealed a clear distinction between traces with and without injected photons, as visualized in a representative trace. These parity measurements provide a direct indication of the number of photons within the cavity, allowing researchers to identify the presence of dim photon signals. However, current detector operation still requires sufficient shielding and cooling, limiting immediate deployment outside of specialized laboratory settings. Maintaining the extremely low temperatures necessary for superconductivity and preventing external electromagnetic interference are important engineering challenges.

Achieving dark photon sensitivity despite limitations in accident detection

Researchers at Aalto University and the University of Chicago are continually improving methods to detect dark matter by using superconducting qubits and dual-cavity systems to search for axions and dark photons. This protocol achieves excellent sensitivity comparable to previous single-photon detectors despite the reduced qubit performance, but currently focuses only on the dark photon model. This introduces tensions, as the initial range includes both dark photons and axions, and results regarding the latter still do not exist, potentially limiting the search range and requiring future dedicated experiments. Another promising dark matter candidate, axions, are hypothetical particles that are predicted to interact very weakly with photons in the presence of strong magnetic fields. Detection of axions requires different experimental techniques and parameter ranges than dark photons, and current settings appear to be optimized for the latter. The lack of axion detection results suggests that further modifications or dedicated experiments may be required to effectively explore this dark matter candidate.

This work represents a major advance in microwave quantum sensing, validating machine learning-optimized control as a viable path toward faster and more scalable dark matter detectors. The research team demonstrated significant advances in the detection of potential dark matter signals by leveraging superconducting qubits and a dual-cavity system, overcoming the challenges posed by signal degradation and achieving sensitivity comparable to previous detectors despite the reduced performance of key components. The use of machine learning to optimize control pulses is particularly noteworthy as it demonstrates the potential for automation and performance enhancement of complex quantum experiments. This approach could also be applied to other areas of quantum sensing and information processing. The strong performance increases qubit-cavity coupling, reduces experimental time, and improves sensitivity for future studies involving a wider range of dark matter candidates. Future research may focus on extending the protocol to search for axions, increasing detector bandwidth, and improving scalability to enable the construction of larger and more sensitive dark matter detectors. The development of more robust and compact shielding and cooling systems is also important to facilitate the deployment of these detectors in a wider range of environments.

Researchers have demonstrated that superconducting qubits and a dual-cavity system can improve the sensitivity of dark matter searches. This advance enables faster measurements, increases the possibility of detecting weakly interacting particles, and establishes exclusion limits for dark photon models, especially with velocity mixing angles of about 1×10.^-14 At 5.051GHz. Machine learning optimization of the control pulses was found to be critical to maintaining performance despite reduced qubit coherence and cavity lifetime. The authors suggest that future studies could extend this protocol to search for axions and further improve the scalability of the detector.