Niels Bohr and MIT Researchers Devise Qubit Noise‑Mitigation Algorithm.

Dr. Fabrizio Berritta from the Center for Quantum Devices at the Niels Bohr Institute, University of Copenhagen, has demonstrated a real‑time noise‑mitigation algorithm for superconducting qubits, called Frequency Binary Search. Using a Quantum Machines controller equipped with a field‑programmable gate array, the method estimates qubit frequency fluctuations. It adjusts microwave control pulses within a single experimental cycle, requiring fewer than ten measurements to achieve exponential precision in calibration. The technique, validated across tens of qubits and scalable to future millions‑qubit systems, promises to reduce decoherence in quantum processors operating at cryogenic temperatures. The findings appear in PRX Quantum.

Quantum Noise Mitigation Algorithm Developed by Copenhagen, MIT, NTNU, Leiden Collaboration

The collaboration between the Niels Bohr Institute, MIT, the Norwegian University of Science and Technology (NTNU), and Leiden University has produced a new algorithm for the real-time management of noise in qubits. The method, named Frequency Binary Search, is described in a paper published in PRX Quantum and promises to mitigate decoherence by rapidly estimating and correcting qubit frequency shifts caused by magnetic and electrical disturbances. Noise, the “ghost in the machine,” remains the principal obstacle to scalable quantum computing, and this algorithm directly addresses that challenge.

Dr. Fabrizio Berritta, a PhD student supervised by Prof. Ferdinand Kümmeth at the Centre for Quantum Devices, led the development during his exchange at MIT. The algorithm is implemented on a Quantum Machines controller equipped with an integrated field‑programmable gate array (FPGA). The FPGA continuously estimates changes in the energy splitting (E) of a superconducting qubit threaded by magnetic flux, using a binary‑search routine, and adjusts the control microwave pulses in real time. “You can measure the actual noise, and once we know the noise, we can correct the control path to mitigate the decoherence,” Berritta explained.

The team experimentally validated the method with Lukas Pahl and Melvin Mathews at MIT. By keeping the data processing on the FPGA, the algorithm eliminates the latency that would otherwise allow noise to evolve before it can be corrected. The result is a quantum noise mitigation technique that can be applied to a wide range of qubits, even in large numbers. It is accessible to laboratories worldwide through commercial quantum controllers that can be programmed in a Python‑like language. During a brief workshop, Jacob Benestad and Jan Krzywda, under the guidance of their supervisors, explored efficient algorithmic options that led to the current design.

In current quantum processors, calibration typically requires thousands or tens of thousands of measurements per qubit. Frequency Binary Search reduces this requirement to fewer than ten measurements while achieving exponential precision with respect to the number of measurements. The method therefore scales favourably as the number of qubits grows from the present tens or hundreds toward the envisaged millions, offering a path to efficient decoherence suppression in future large‑scale devices. The quantum processing unit in the study is cooled to temperatures just above absolute zero (273 K) inside a cryostat, underscoring the need for rapid, real-time noise mitigation.

Real-Time Frequency Binary Search Enables Rapid Qubit Calibration

Dr. Fabrizio Berritta, supervised by Prof. Ferdinand Kuemmeth at the Centre for Quantum Devices, Niels Bohr Institute, University of Copenhagen, has devised a new algorithm called Frequency Binary Search that enables rapid qubit calibration. The work is a collaboration among MIT, the Norwegian University of Science and Technology (NTNU), and Leiden University, and has been published in the peer-reviewed journal PRX Quantum.

The algorithm is embedded in a Quantum Machines controller that houses an FPGA. By performing a binary‑search estimation of the qubit’s energy splitting E on the FPGA, the controller can adjust the microwave drive pulses within microseconds, eliminating the latency that would otherwise allow magnetic noise to alter the qubit state before correction. The qubit is a superconducting circuit threaded by a magnetic flux; magnetic fluctuations in the surrounding environment cause the energy splitting E to drift, and the FPGA continuously monitors these changes. The controller processes the data locally, so no information is sent to a desktop computer, ensuring that the correction is applied before the noise evolves further.

The team validated the method experimentally with Lukas Pahl and Melvin Mathews at MIT. Jacob Benestad and Jan Krzywda, guided by their supervisors, explored algorithmic options that led to the current design. Compared with conventional calibration routines that require thousands or tens of thousands of repetitions per qubit, Frequency Binary Search achieves exponential precision with fewer than ten measurements, allowing simultaneous calibration of all qubits in a system. The FPGA is programmed using a Python-like language, enabling researchers to implement the algorithm without specialised electrical engineering expertise and making the technique accessible to laboratories worldwide through commercial quantum controllers.

By providing rapid, real-time quantum noise mitigation, the method addresses the central obstacle of magnetic and electrical decoherence in devices cooled to just above absolute zero (273 K) within a cryostat. The technique scales favourably as quantum processors grow from the present tens or hundreds of qubits toward the envisaged millions, offering a path to efficient decoherence suppression in future large‑scale devices, according to a press release from Niels Bohr Institute, University of Copenhagen (press release).

Implications for Scaling Quantum Processors to Millions of Qubits

The prospect of quantum processors containing millions of qubits hinges on the ability to control decoherence while keeping calibration and control overhead manageable. In the current generation of superconducting devices, magnetic and electrical fluctuations in the cryogenic environment—typically maintained at just above absolute zero (273 K) inside a cryostat—cause the qubit energy splitting (E) to fluctuate, leading to rapid loss of coherence. Any strategy that can measure and counteract these fluctuations in real time is therefore essential for scaling up. The new Frequency Binary Search method, developed by Dr Fabrizio Berritta during his Ph D exchange at MIT under the supervision of Prof Ferdinand Kuemmeth of the Niels Bohr Institute, offers precisely such a strategy.

Implemented on a Quantum Machines controller equipped with an integrated field‑programmable gate array (FPGA), the algorithm performs a binary‑search estimation of the qubit frequency (E) without sending raw data to a desktop computer. By continuously adjusting the microwave control pulses in response to the FPGA’s real‑time estimate, the system keeps the qubit’s phase trajectory aligned with the intended trajectory. The method requires fewer than ten measurement repetitions per qubit, yet achieves exponential precision with respect to the number of measurements. Crucially, the same FPGA‑driven routine can be applied to all qubits in a device simultaneously, allowing a single calibration run to cover the entire array. This simultaneous, low‑measurement‑count calibration is a decisive advantage when the number of qubits grows from the present tens or hundreds toward the envisaged millions.

The implications for large‑scale quantum processors are significant. By reducing the calibration time from thousands of repetitions per qubit to a handful, the Frequency Binary Search method dramatically lowers the time‑overhead that would otherwise dominate the operation of a million‑qubit machine. Moreover, the algorithm’s reliance on a commercial quantum controller with a Python-like programming interface means that laboratories worldwide can adopt the technique without needing specialised electrical engineering expertise. The method’s applicability to a wide range of qubit modalities—beyond the superconducting flux‑tuned devices demonstrated in the study—suggests that it could become a standard tool for quantum noise mitigation across the industry. As quantum processors scale, the ability to perform rapid, real‑time noise mitigation will be a key factor in maintaining coherence and achieving fault‑tolerant operation.