Aalto University Team Develops Machine-Learning Optimized Pulses for Dark Matter Searches

A new microwave detection protocol improves the search for axions and dark photons. Yu-Han Chang and colleagues from Aalto University and University of Chicago and University of Zaragoza, employ a superconducting transmon qubit and a double-cavity system. The protocol addresses limitations in conventional qubit control by using machine-learning optimised pulses, achieving single-photon detection performance. It establishes new exclusion limits on the dark photon model with a kinetic mixing angle sensitivity of approximately 1times10^-14 at 5.051GHz. These advancements offer the potential for faster and more scalable microwave quantum sensors in the ongoing quest to understand the nature of dark matter.

Enhanced dark photon sensitivity via machine-learning optimised qubit control

Sensitivity to the kinetic mixing angle reached approximately 1times10^-14 at 5.051GHz, a substantial improvement over previous dark photon detection protocols. Earlier methods struggled to reliably identify signals below this level due to qubit control challenges and noise, and this threshold surpasses those limitations, enabling exploration of a previously inaccessible region of parameter space for dark matter candidates. The dark photon, a hypothetical particle proposed as a mediator between ordinary matter and dark matter, is often sought through its potential mixing with photons. This ‘kinetic mixing’ allows dark photons to interact weakly with electromagnetic fields, and the sensitivity of a detector to this mixing angle is a crucial figure of merit. Previous experiments, limited by qubit decoherence and cavity lifetimes, struggled to achieve the necessary precision to probe the parameter space below 1times10^-14. This new protocol overcomes these hurdles, opening up new avenues for investigation.

A superconducting qubit and double-cavity system now establishes a pathway for faster and more scalable microwave quantum sensors in the ongoing search for dark matter. Despite sharply reduced qubit coherence times and storage-cavity lifetimes of 10 microseconds, single-photon detection performance remained equivalent to previous implementations. This is a significant achievement, as shorter coherence times typically lead to increased noise and reduced signal fidelity. The double-cavity system is designed to store photons, allowing for increased interaction time with the qubit. The qubit, a superconducting transmon, acts as a sensitive probe of the photon number within the storage cavity. Machine-learning optimised pulses broadened the operational bandwidth to approximately 20MHz and suppressed gate errors by two orders of magnitude compared to standard rectangular pulses, as confirmed by Quantum Process Tomography, which demonstrated high-fidelity qubit control. Quantum Process Tomography is a technique used to characterise the performance of quantum gates, ensuring accurate and reliable operation. The broadened bandwidth allows for faster data acquisition, while the reduction in gate errors improves the overall signal-to-noise ratio.

The increased qubit-cavity coupling resulting from this approach shortens experimental times and improves sensitivity for future investigations. Stronger coupling means that the qubit and cavity interact more effectively, leading to a faster response and a more pronounced signal. A Hidden Markov Model, a statistical technique for modelling sequences of events, yielded background rates on the order of $\mathcal{O}$ Hz, enabling the establishment of exclusion limits on dark photon models. The Hidden Markov Model is used to distinguish between genuine signals and background noise, improving the accuracy of the measurements. Data processing of parity measurements, binary outcomes indicating qubit state, revealed clear distinctions between traces with and without injected photons, as visualised in representative traces. These parity measurements provide a direct indication of the photon number in the cavity, allowing researchers to identify the presence of a dark photon signal. However, the current detector still requires substantial shielding and cooling to operate, limiting its immediate deployment outside specialised laboratory settings. Maintaining the extremely low temperatures necessary for superconductivity and shielding against external electromagnetic interference are significant engineering challenges.

Dark photon sensitivity achieved despite limitations in axion detection

Researchers from Aalto University and the University of Chicago are continually refining methods to detect dark matter, utilising a superconducting qubit and double-cavity system to search for axions and dark photons. The protocol achieves impressive sensitivity, comparable to previous single-photon detectors despite reduced qubit performance, but currently focuses solely on the dark photon model. This presents a tension, as the initial scope included both dark photons and axions, and results for the latter remain absent, potentially limiting the breadth of the search and requiring dedicated future experiments. Axions, another leading dark matter candidate, are hypothetical particles predicted to interact very weakly with photons in the presence of a strong magnetic field. Detecting axions requires different experimental techniques and parameter ranges compared to dark photons, and the current setup appears optimised for the latter. The absence of axion detection results suggests that further modifications or a dedicated experiment may be necessary to explore this dark matter candidate effectively.

This work represents a major advance in microwave quantum sensing, validating machine-learning optimised control as a viable path towards faster and more scalable dark matter detectors. The team demonstrated a significant advancement in detecting potential dark matter signals, utilising a superconducting qubit and a double-cavity system, and overcoming challenges posed by signal degradation to achieve sensitivity comparable to previous detectors despite reduced performance of key components. The use of machine learning to optimise control pulses is particularly noteworthy, as it demonstrates the potential for automating and improving the performance of complex quantum experiments. This approach could be applied to other areas of quantum sensing and information processing. The strong performance allows for increased qubit-cavity coupling, shortening experimental times and improving sensitivity for future investigations encompassing a wider range of dark matter candidates. Future work could focus on extending the protocol to search for axions, increasing the detector’s bandwidth, and improving its scalability to enable the construction of larger and more sensitive dark matter detectors. The development of more robust and compact shielding and cooling systems would also be crucial for facilitating the deployment of these detectors in a wider range of environments.

The researchers demonstrated improved sensitivity in the search for dark matter utilising a superconducting qubit and a double-cavity system. This advancement enables faster measurements and enhances the potential to detect weakly interacting particles, specifically establishing exclusion limits on the dark photon model with a kinetic mixing angle of approximately 1times10^-14 at 5.051GHz. Machine-learning optimisation of control pulses proved crucial in maintaining performance despite reduced qubit coherence and cavity lifetimes. The authors suggest future work may extend this protocol to search for axions and further improve detector scalability.