发表机构
Molecular Foundry, Lawrence Berkeley National Laboratory（分子发现中心，劳伦斯伯克利国家实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究利用进化强化学习优化欠阻尼双势阱中1比特存储器的擦除协议，发现平均功随协议持续时间呈现等温与绝热两种标度行为，并证明学习协议优于平衡边界条件最优协议。
AI 中文摘要
我们使用进化强化学习来确定欠阻尼悬臂在双势阱中运动的有效时间相关擦除协议，这是1比特存储器的实验实现。我们研究了擦除一个比特所需的平均功$\langle W \rangle$如何随协议持续时间$\tau$的变化而变化。我们发现了两个区域，取决于$\tau$与系统弛豫时间$t_r$的比较。对于$\tau \gg t_r$，即准静态等温区域，我们恢复了朗道尔极限加上一个随$1/\tau$变化的额外开销，这与过阻尼情况类似。相反，对于$\tau<t_r$，擦除变为绝热过程，$\langle W \rangle$的增长比等温情况更慢。这种增长的下界为$1/\tau$，我们通过一个思想实验的最优协议推导出这一结果。最后，与过阻尼擦除的比较表明，学习到的协议可以优于在平衡边界条件下最优的协议。
英文摘要
We use evolutionary reinforcement learning to determine efficient time-dependent erasure protocols for an underdamped cantilever moving in a double-well potential, an experimental realization of a 1-bit memory. We investigate how the mean work $\langle W \rangle $ needed to erase a bit scales as a function of the protocol duration $τ$. We find two regimes, depending on how $τ$ compares to the relaxation time of the system $t_r$. For $τ\gg t_r$, the quasistatic isothermal regime, we recover Landauer's bound plus an overhead that scales as $1/τ$, similar to the overdamped case. By contrast, for $τ<t_r$ erasure becomes adiabatic and $\langle W \rangle$ grows more slowly than in the isothermal case. This growth is bounded from below as $1/τ$, which we derive using a gedanken optimal protocol. Finally, comparison with overdamped erasure shows that learned protocols can outperform protocols that are optimal subject to equilibrium boundary conditions.
