1.National Time Service Center, Chinese Academy of Sciences, Xi’an 710600, China 2.Key Laboratory of Time and Frequency Primary Standards, National Time Service Center, Chinese Academy of Sciences, Xi’an 710600, China 3.University of Chinese Academy of Sciences, Beijing 100049, China
Abstract:Cold collision frequency shift is one of the major systematic effects which limit the frequency uncertainty of the cesium fountain atomic clock. It is proportional to the effective atomic density, which is defined as the average density over the initial spacial and velocity distribution. The measurement of the frequency shift is based on a differential method, in which the fountain clock is operated with two different atomic densities, i.e. high density and low density, in turn. The clock frequency without collision shift can be achieved by linear extrapolation with the frequencies and density ratios of two states. For the density ratio is estimated with the atom number, it plays a crucial role in generating atoms with same density distribution for reducing systematic uncertainty in cold collision frequency shift estimation. The rapid adiabatic passage method is used in Cesium fountain clock to realize homogeneous transition probability, which modulates the amplitude and frequency of microwave continuously to prepare atom sample. To investigate the precision of this method, theoretical analysis and experimental measurement are both used here. An equation of deviation is derived from the time evolution of Bloch vector. The vector rotates at angular speed Ω with the rotation axis processing at lower angular speed. The deviations in the two directions on the surface of Bloch sphere are determined by the equations which are similar to wave equations, and can be simplified into wave equations when the deviations are sufficiently small. It is shown in the equations that the deviations are stimulated by angular velocity and angular acceleration of the precession, and is inversely proportional to the square of Ω. Further calculation shows that the deviation becomes smaller when the amplitude of microwave frequency and Rabi frequency are close to each other. It is then confirmed experimentally. The effects of some other parameters, such as the pulse length and time delay, on transition probability are also measured, showing that the RAP method is insensitive to these parameters up to a large scope. The precision of RAP method is dominated by three factors. The first factor is the product of rotating angular speed Ω and pulse length T, i.e. ΩT: The increase of ΩT can reduce the uncertainty to a satisfactory degree. The second factor is the uncertainty of resonant frequency, so the measurement is required to be precise. The third factor is the unexpected atoms which are not selected by the microwave, and may be attributed to pulling light. After optimizing the parameters, the ratio of low density to high density can approach to 0.5 with 3 × 10–3 uncertainty, which leads to a systematic relative uncertainty of cold collision shift up to 1.6 × 10–16. Keywords:cesium fountain atomic clock/ cold collision frequency shift/ rapid adiabatic passage method
全文HTML
--> --> -->
2.二能级绝热跃迁理论分析对二能级系统, 原子态的演化可由Bloch方程来表示:
$\frac{{\rm{d}}}{{{\rm{d}}t}}\left( {\begin{aligned} x \\ y \\ z\end{aligned}} \right) = 2{\text{π}}\left( {\begin{array}{*{20}{c}} 0&{ - \delta }&0 \\ \delta &0&{ - b} \\ 0&b&0 \end{array}} \right)\left( {\begin{aligned} x \\ y \\ z \end{aligned}} \right),$
由于误差只与Δθ, Δφ的二次项成正比, 因此全脉冲对影响绝热跃迁演化过程的参数比较不敏感. 在绝热跃迁中, 微波功率和频率的调整幅度b0和δ0是最直接的参数, 对跃迁几率有直接的影响. 图3分别展示了固定b0调整δ0, 以及固定δ0调整b0的跃迁几率变化. 在脉冲形状确定后, 假定中心频率f0与原子谐振频率无失谐, 则b0与δ0的比值k完全决定误差源的变化. 当k接近1时, 误差源很小, 当k远离1时误差源迅速增大. 其中, 当k大于1时, 误差源主要出现在脉冲的两端, 反之则出现在脉冲的中间(图4). 图 3 脉冲长度8 ms, 脉冲起始点在进入腔后4 ms的跃迁几率 (a) δ0为5 kHz, 不同b0的跃迁几率; (b)功率幅度为10 kHz, 不同δ0的跃迁几率 Figure3. Transition probability as a function of b0 and δ0 with 8 ms pulse starts 4 ms after atoms entering the cavity: (a) δ0 = 5 kHz, with different b0; (b) b0 = 10 kHz, with different δ0.
图 4 误差源随时间的变化 Figure4. Time evolution of deviation excitation.
极端情况下, 在k过大或过小时, 绝热条件不能满足, 误差就会非常大. 在全脉冲情况下, 所有的偏差都会导致跃迁几率减小. 脉冲的时间参数是另一组重要的参数, 脉冲相对于原子团进入微波作用区的时间和作用时间, 都会对原子跃迁几率产生不同的影响. 原子团通过选态腔的时间约为12 ms, 通过原子的上抛高度和选态腔的位置, 计算得原子团到达腔中点的时间为上抛开始40 ms. 图5(a)测量了不同的脉冲长度造成的影响. 由于全脉冲误差很小, 在脉冲偏离8 ms不太远的范围, 跃迁几率近似都为1. 当脉冲长度接近0, 由于脉冲长度的缩短相当于Ω的缩小, 在脉冲长度过于小时, 绝热条件渐渐遭到破坏, 造成跃迁几率减小. 在图5(a)的另一端, 脉冲长度增大, 当脉冲的长度超过原子团过腔时间时, 脉冲起始时原子尚未进入选态腔, 原子团出腔时, 脉冲尚未结束. 因此, 原子团感受到的脉冲相比输入的脉冲会发生畸变, 在两端都会产生进动轴角速度的异常改变, 使原子的状态偏离进动轴, 跃迁几率下降. 在图5(b)中, 脉冲长度固定为8 ms, 改变起始时间, 跃迁几率的变化也会因为同样的原因而降低. 当起始点早于0 ms时, 原子感受到的脉冲会在开始的部分产生畸变; 当起始点大于4 ms时, 原子感受到的脉冲会在结束的部分产生畸变; 跃迁几率都会因此下降. 图5(b)中跃迁几率下降的时间点和计算所得相符合. 图 5 时间参数对跃迁比例的影响, 其中δ0 = 5 kHz, b0 = 10 kHz. 以原子到达选态腔下端面为时间0点 (a)固定脉冲以原子在腔中心的时间点为中心, 改变脉冲长度; (b)固定脉冲长度为8 ms, 改变脉冲起始时间 Figure5. Transition probability as a function of time parameters, δ0 = 5 kHz, b0 = 10 kHz, atoms enter selection cavity at time 0: (a) Pulse duration remaining symmetric about the central point of cavity; (b) start point of pulse with a fixed duration of 8 ms.
另一个重要参数是中心频率f0. 理想情况下, 频率调整的范围应以谐振频率为中心. 但由于测量偏差, 可能会有所偏离. 此处测量了这项误差对于全脉冲跃迁几率的影响, 结果如图6所示. 当中心频率改变时, 转动轴的角速度在中点两侧会产生不对称的变化, 改变偏差激励. 但实验结果显示, 对于一个相当大的频率范围(±4.5 kHz), 在观测精度以内没有发现跃迁几率的变化. 只有在两端偏离值接近5 kHz时, 跃迁几率才有显著下降. 由于8 ms的脉冲产生的Rabi跃迁峰, 半高宽就约为125 Hz左右, 因此在中心频率偏差接近5 kHz时, 在脉冲的开始或结束时, 绝热条件都遭到了显著破坏, 因此才会出现跃迁几率下降. 图 6 不同频率中心值的跃迁几率 Figure6. Transition probabilty as a function of center frequency detuning.
误差正比于Δξ和Δφ, 因此, 半脉冲的误差相比全脉冲更容易受到影响. 除了原子状态因为激励源偏离转轴造成的误差, 由于中心频率的误差造成半脉冲结束时轴的角度偏差也能造成跃迁几率的偏差. 对半脉冲的测量, 脉冲结束时间点的频率会影响此时转动轴的角度, 从而影响跃迁几率. 由于磁场不均匀, 若脉冲结束时原子的位置不同, 则中心频率的相对移动就会带来额外的误差. 为对比起见, 图7测量不同脉冲长度时, 脉冲结束点固定在进入腔后10 ms处, 通过改变脉冲起始点来改变脉冲长度. 如前所述, 脉冲时间缩短, 相当于转动速度减小, 因此对不同的脉冲长度, 低功率下跃迁几率都减小, 但脉冲越短, 降低得越多. 而以b0T为横坐标时, 则不同脉冲长度的曲线近似于完全重合. 图 7δ0 = 5 kHz, 脉冲开始于不同时刻, 结束于入腔后10 ms. 5个不同脉冲长度下, 不同b0的跃迁几率 (a)横坐标为b0; (b)横坐标为b0与脉冲长度T的乘积 Figure7. Transition probability as a function of b0 for a 5 kHz δ0 pulse start at 5 different points and end at 10 ms after entering cavity: (a) b0 as the abscissa; (b) b0T as the abscissa.
图8(a)显示了不同的δ0对跃迁的影响. 在b0较低时, 跃迁几率会显著小于1, 这同之前的结果相类似. 而δ0此时对跃迁几率就有明显的影响, δ0越大, 跃迁几率越小. 在功率幅度小于δ0时, 越大的δ0, 造成误差激励源的增大越明显, 即使Ω的增大也不能补偿由此产生的偏差. 只有在功率幅度和δ0大小接近时, 角加速度和Ω变化率带来的偏差减小, Ω的增大才会有助于减小误差. 在图8(b)中, 脉冲长度的改变相当于Ω的变化, 但脉冲幅度和频率幅度的比保持不变, 于是在短到1 ms的脉冲时, 跃迁几率也没有像图8(a)中那样出现明显的下降. 图 8 (a)在5个δ0下, 不同b0的跃迁几率; (b)脉冲频率幅度为5 kHz, b0为10 kHz, 不同脉冲长度的跃迁几率 Figure8. Transition probability as a function of (a) b0 for five different δ0, (b) pulse duration with δ0 = 5 kHz, and b0 = 10 kHz.
当半脉冲结束时, 由于失谐为0, 所以不论微波功率多大, 转动轴都正好指向π/2角度, 对应于1/2跃迁几率. 这正是半脉冲对功率相对不敏感的原因所在. 但是, 如果中心频率有误差Δf, 则跃迁几率为0.5 + Δf/Ω. 由于角度反比于Rabi频率Ω, 因此随着功率幅度的增高, 误差逐渐缩小. 图9(a)展示了在3个中心频率下跃迁几率的变化, 明显可以看到, 在中心频率大于0时, 随着脉冲幅度的增加, 跃迁几率会先增大到接近0.7, 再逐渐缩小. 图9(b)中, 在脉冲长度趋于0的一端, 跃迁几率的上升也同样意味着中心频率偏大. 在图8(b)中, 不同中心频率下的跃迁几率展示出非常好的线性度. 这一结果同样很明显地显示出中心频率的偏差. 图 9 (a)中心频率取100, 0, –100 Hz, 不同b0时的跃迁几率; (b)δ0 = 5 kHz, b0 = 10 kHz时, 不同中心频率的跃迁几率 Figure9. (a) Transition probability as a function of b0 for 100, 0, –100 Hz center frequency detuning; (b) transition probability as a function of center frequency detuning for δ0 = 5 kHz and b0 = 10 kHz.