Abstract
Kerr parametric oscillators (KPOs) have attracted increasing attention in terms of their application to quantum information processing and quantum simulations. The state preparation and measurement of KPOs are typical requirements when used as qubits. The methods previously proposed for state preparations of KPOs utilize modulation of external fields such as a pump and drive fields. We study the stochastic state preparation of stable coherent states of a KPO with homodyne detection, which does not require modulation of external fields, and thus can reduce experimental efforts and exclude unwanted effects of possible imperfection in control of external fields. We quantitatively show that the detection data, if averaged over an optimal averaging time to decrease the effect of measurement noise, has a strong correlation with the state of the KPO, and therefore can be used to estimate the state (stochastic state preparation). We examine the success probability of the state estimation taking into account the measurement noise and bit flips. Moreover, the proper range of the averaging time to realize a high success probability is obtained by developing a binomial-coherent-state model, which describes the stochastic dynamics of the KPO under homodyne detection.
Introduction
Kerr Parametric Oscillators (KPOs), often referred to as "Kerr-cat qubits," are a promising platform for quantum information due to their inherent resistance to phase-flip errors. However, initializing these systems into a specific starting state—a process known as state preparation—is traditionally challenging. Standard techniques require the precise modulation of external microwave fields (pumps and drives), which is experimentally demanding and susceptible to control noise.
This research proposes a stochastic state preparation method that bypasses complex field manipulation. Instead of forcing the KPO into a state, the researchers utilize continuous measurement to identify which of its two stable states ($|\alpha\rangle$ or $|-\alpha\rangle$) the system has naturally settled into. This approach simplifies hardware requirements and avoids imperfections associated with active field control.
Methods
The core of the proposed method is continuous homodyne detection. The KPO's output is mixed with a local oscillator at a 50/50 beam splitter, and the resulting signal—the difference in photon counts between two detectors—is monitored.
Because the raw measurement signal is dominated by noise, the data must be averaged over a specific time window ($T_a$). To determine the optimal window, the authors developed a binomial-coherent-state model. This theoretical framework treats the KPO as a system jumping between two states, allowing for the analytical calculation of a "Goldilocks zone": an averaging time long enough to filter Gaussian measurement noise but short enough to avoid the signal-smearing effects of random "bit-flips" (state jumps).
Key Results and Insights
The success of this measurement-based preparation depends critically on the choice of $T_a$. The researchers demonstrated that with an optimized averaging time, the state can be estimated with a success probability higher than 0.987 under realistic experimental conditions.
As illustrated in Figure 6, the state estimation is a balancing act:
- Too short ($T_a = 10^{-4} \mu s$): Gaussian measurement noise dominates, making it impossible to distinguish the states.
- Too long ($T_a = 10 \mu s$): The KPO's natural bit-flips occur within the averaging window, leading to an incorrect or blurred estimate.
- Optimal ($T_a = 10^{-1} \mu s$): The signal clearly distinguishes between the two states with high fidelity.
Notably, the method is robust to loss. Even with a detection efficiency as low as 10% ($\eta = 0.1$), high-fidelity state preparation is still achievable by adjusting the averaging time to compensate for the weaker signal.
Conclusion
This stochastic preparation method offers a path toward simpler and more robust KPO-based quantum computing architectures. By eliminating the need for pump amplitude modulation, it reduces the complexity of microwave control systems. While the process is stochastic, it can be integrated into deterministic algorithms using feed-forward control (measuring the state and then applying a flip gate if necessary). Future work will explore extending this model to multi-KPO systems, where ferromagnetic coupling might further stabilize the states against bit-flips, potentially increasing preparation efficiency.
Summary updated