1.State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, Wuhan 430071, China 2.Center for Cold Atom Physics, Chinese Academy of Sciences, Wuhan 430071, China
Fund Project:Project supported by the National Key Research and Development Program of China (Grant Nos. 2016YFA0302800, 2017YFA0304501), the Strategic Priority Research Program of the Chinese Academy of Sciences, China (Grant No. XDB21010100), and the National Natural Science Foundation of China (Grant Nos. 11674361, 11774389).
Received Date:04 December 2018
Accepted Date:24 December 2018
Available Online:01 February 2019
Published Online:05 February 2019
Abstract:As an important candidate for quantum simulation and quantum computation, a microscopic array of single atoms confined in optical dipole traps is advantageous in controlled interaction, long coherence time, and scalability of providing thousands of qubits in a small footprint of less than 1 mm2. Recently, several breakthroughs have greatly advanced the applications of neutral atom system in quantum simulation and quantum computation, such as atom-by-atom assembling of defect-free arbitrary atomic arrays, single qubit addressing and manipulating in two-dimensional and three-dimensional arrays, extending coherence time of atomic qubits, controlled-NOT (C-NOT) gate based on Rydberg interactions, high fidelity readout, etc.In this paper, the experimental progress of quantum computation based on trapped single neutral atoms is reviewed, along with two contributions done by single atom group in Wuhan Institute of Physics and Mathematics of Chinese Academy of Sciences. First, a magic-intensity trapping technique is developed and used to mitigate the detrimental decoherence effects which are induced by light shift and substantially enhance the coherence time to 225 ms which is 100 times as large as our previous coherence time thus amplifying the ratio between coherence time and single qubit operation time to 105. Second, the difference in resonant frequency between the two atoms of different isotopes is used to avoid crosstalking between individually addressing and manipulating nearby atoms. Based on this heteronuclear single atom system, the heteronuclear C-NOT quantum gate and entanglement of an Rb-85 atom and an Rb-87 atom are demonstrated via Rydberg blockade for the first time. These results will trigger the quests for new protocols and schemes to use the double species for quantum computation with neutral atoms. In the end, the challenge and outlook for further developing the neutral atom system in quantum simulation and quantum computation are also reviewed. Keywords:Rydberg state/ single neutral atom/ quantum entanglement/ coherence time
全文HTML
--> --> --> -->
2.1.实现了可扩展的量子比特的系统
中性原子体系通常采用一个碱金属原子(如铷原子和铯原子)基态超精细能级的磁子能级作为一个量子比特的0态和1态. 这样的量子态具备纯净和易操控的特点, 是理想的量子比特. 单个原子捕获和囚禁则一般用光阱来实现, 主要有两种方法, 一是在光晶格中, 利用Mott绝缘态在每个格点制备一个原子[24], 但由于冷原子云和光场的空间不均匀性, 只有光晶格中心的部分会均匀地装载; 二是利用碰撞阻塞效应, 当一个光镊型光偶极阱足够小时, 阱中两个以上的原子在共振光的作用下会很快损失掉, 只有一个原子能保存在阱中, 从而获得单个原子[25], 如图2所示. 但该装载是随机的, 扩展到多个原子阵列时, 无法实现确定性的制备. 2016年, 法国Barredo等[13]发展了一种用可移动光阱实现单原子逐个装配的技术. 他们在2D包含100个光偶极阱的阵列中采用碰撞阻塞的原理随机装载单原子, 随后对光阱阵列进行成像, 判断出哪些阱中有原子, 然后用一个可移动的光阱将单原子逐个转移到所需要的光阱中, 从而确定性地制备包含了50个单原子的不同构型的单原子阵列, 随后他们还演示了包含72个单原子的任意3D构型的确定性制备[12]. 类似的工作还包括美国Endres等[14]在一维(1D)包含50个单原子的阵列中制备演示, 韩国Kim等[26]在2D阵列中通过对转移算法的优化和对格点中单原子的实时反馈装载来提高制备效率, Kumar等[27]展示了3D光晶格中利用态依赖的光晶格重排原子得到了$ 5\times5\times2 $和$ 4\times4\times3 $的单原子阵列. 采用单原子装载后重新排列的方案理论上可以扩展到包含更多单原子的阵列的确定性制备, 从而基本解决了中性原子体系扩展性的问题. 图 2 远红失谐光聚焦形成的偶极阱和偶极阱中收集的单原子荧光信号 Figure2. The optical dipole trap formed by strongly focusing far red-detuned laser. The fluorescence of single atoms trapped by the dipole trap.
22.2.实现了高精度的态初始化 -->
2.2.实现了高精度的态初始化
利用成熟的光泵技术将原子制备到量子比特的$ |0\rangle $态或$ |1\rangle $态, 效率可以达到99.9%以上[28]. 例如, 将铷原子制备到编码量子比特的$ |1\rangle $态的$ F=2, $$ m_F=0 $态时(如图3所示), 一般采用高斯量级的磁场将不同的磁子能级区分开, 然后用${\text{π}}$偏振的$ F=2 $到$ F'=2 $的共振光配合$ F=1 $到$ F'=2 $的回泵光, 由于跃迁选择定则, $ F=2 $, $ m_F=0 $是暗态, 经过一段时间的激光作用后, 原子会全部布居到暗态上. 而且由于相同种类的原子能级结构都是一致的, 因此采用同样的光泵光可以同时实现阵列中所有原子的态初始化. 图 3$ ^{87}{\rm Rb} $原子能级和相关的冷却光$ I_{\rm cool} $、回泵光$ I_{\rm rep} $、态制备光$ I_{\rm pum} $和态探测光$ I_{\rm prob} $对应的跃迁(量子比特的$ |0\rangle $态和$ |1\rangle $态编码在 $ F=1 $,$ m_F $ = 0和$ F=2 $, $ m_F=0 $上) Figure3. The energy levels and lasers used for cooling, repumpiup ng, optical pumpiup ng, and state detection of $ ^{87}{\rm Rb} $. The ground hyperfine states of $ F=1 $, $ m_F=0 $ and $ F=2 $, $ m_F=0 $ are used for encoding the qubit.
2010年, Derevianko[64]以及Lundblad等[65]提出和演示了可以利用矢量光频移抵消微分光频移的方案. 为了诱导矢量极化率, 即等效磁场, 基本的实验方案是, 将囚禁光场从原来的线偏振改为圆偏振, 同时将光的偏振矢量的方向与量子化轴的磁场方向设为一致. 由此得到微分光频移为$ \text{δ}\nu(B, U_{\rm a})=\beta_1 U_{\rm a}+\beta_2 B U_{\rm a}+\beta_4 U_{\rm a}^2 $, 其中B为量子化轴磁场, $ \beta_1 $为标量光频移系数, $ \beta_2 $为矢量极化率与塞曼相互作用导致的3阶交叉项的系数, $ \beta_4 $为基态超极化率系数, $ U_{\rm a} $为偶极阱的阱深. 他们发现, 在弱场极限条件下, 光场诱导的原子的4阶超极化率$ \beta_4 $可以忽略, 从而可以得到一个“魔幻磁场”囚禁条件, 即在一个特定的磁场下(通常需要几个高斯), 微分光频移能够被矢量极化率与塞曼相互作用导致的3阶交叉项所抵消. 但是实验上为了囚禁温度为十几微开的单原子量子比特, 光镊所需的光场较强, 这时光场诱导的等效磁场达到高斯量级, 与外加磁场可比拟, 因而弱场近似不成立, 高阶项即超极化率不可忽略. 而且由于超极化率的贡献, 光频移是势阱深度的二次函数. 实验上, 考虑到$ \beta_2 $和$ \beta_4 $也依赖于偶极光的圆偏振度, 为简单起见, 我们采用完全相同的圆偏振光, 测量了不同阱深和磁场下原子所感受到的微分光频移, 结果如图4(a)所示. 采用圆偏振阱中$ \beta_{1}\approx3.47\times$$ 10^{-4} $去拟合, 得到$ \beta_2 $为$-0.99(3) \times 10^{-4}\ {\rm G}^{-1} $, 并且首次测量了$ ^{87}{\rm Rb} $原子的超极化率$ \beta_4 $ = 4.6(2) $ \times $$ 10^{-12} $$ {\rm Hz}^{-1} $. 该结果与理论计算结果$\beta_{2}=$$ -1.03\times 10^{-4}\;{\rm G}^{-1} $和$ \beta_{4}=4.64\times10^{-12}\;{\rm Hz}^{-1} $相符合. 图 4 (a)超极化率不可忽略情况下, 原子量子比特的微分光频移在不同磁场下随偶极阱势深的变化; (b)原子量子比特相干时间在不同偶极阱势深下的实验值, 蓝色实线为理论值; 内插图显示了阱深为$ U_{\rm M} $时, 通过拟合Ramsey条纹的对比度得到相干时间为$ \tau= (225 \pm 21) $ ms[17] Figure4. (a) In the presence of hyperpolarizability, the differential light shift (DLS) of a qubit in the circularly polarized trap is measured as a function of trap depths at various magnetic field strengths; (b) coherence time $ \tau $ and its dependence on normalized ratios $ U/U_{\rm M} $ obtained from experiment. The solid blue line is the theoretical curve. A coherence time is extracted from a decay time of the envelope of Ramsey visibility, as shown as in the inset. At $ U_a= U_{\rm M} $, $ \tau= (225 \pm 21) $ ms[17].
在大规模的单原子阵列里, 受限于中性原子间微弱的相互作用, 要实现任意两个原子间的量子算法进而实现量子计算和量子模拟, 需要将单原子相干地转移到相互作用区. 我们提出了一种简单易行的深阱转移方案, 实现了单原子在光阱阵列中的高效转移(约95%), 但在转移过程中, 即使采用动态退耦的方法也无法保持量子比特的相干性, 主要原因在于偶极光引起的微分光频移[63]. 运用魔幻光强偶极阱技术(图5所示), 首先在其中一个魔幻光强偶极阱中(trap 2)制备一个量子比特, 通过微波制备到相干叠加态上, 之后通过一个移动魔幻阱(trap 1)将目标量子比特提取出来转移到5 $ {\text{μ}}{\rm m} $的地方后再送回原来的偶极阱, 而后分析量子比特的相干性. 结果发现无转移和被转移的量子比特的Ramsey相干时间几乎一致, 因此在实验的测量精度内, 没有观测到原子的相干性在进行转移操作后有明显的损失, 即使转移后原子的温度从8 ${\text{μ}}{\rm K}$升到16 ${\text{μ}}{\rm K}$. 因此, 实验上证明了用魔幻光强偶极阱转移原子量子比特的过程中, 指向涨落、加热等退相因素变得很小. 然而线偏振阱中这些退相机制是占主导的. 由此解决了中性单原子大规模阵列中单原子相干转移的问题, 极大地提高了原子量子比特间的互联性, 该方法与原子量子比特阵列灵活的构型互相结合, 将有效地降低中性原子量子计算算法的复杂性. 图 5 (a)原子量子比特相干转移的实验装置示意图(Trap 1是可移动阱, 其在焦平面上的位置由2D声光偏转器控制; Trap 2是静止阱; 两阱的偏振可以通过液晶相位片(Thorlabs LCR-1-NIR)实时控制); (b) 原子量子比特在两阱中不转移(黑色方块)和转移(红色圆点)时的Ramsey条纹(实验数据中每个点是100多次实验的平均值; 通过衰减的正弦函数拟合(实线部分), 可以得到静止量子比特和转移量子比特的相干时间分别是(206 $ \pm $ 69) ms和(205 $ \pm $ 74) ms[17] Figure5. (a) Experimental setup for coherent transfer of atomic qubit. Trap 1 is a movable trap which can be shiftted in two orthogonal diretions by an AOD. Trap 2 is a static one. Both of their polarizations can be actively controlled by a liquid crystal retarder (LCR). (b) Measured Ramsey signals for single static qubits (black squares) and single mobile qubits (red dots) at $ B = 3.115 $ G. Every point is an average over 100 experimental runs. The solid curves are fits to the damped sinusoidal function, with coherence times of static qubits and mobile qubits are (206 $ \pm $ 69) ms and (205 $ \pm $ 74) ms, respectively[17].