Learning objectives / 学习目标
完成本课后,你应能:
- 说明原子跃迁怎样校准 local oscillator,正确区分 accuracy、stability、resolution、quality factor 与 traceability。
- 计算 basic Allan deviation,并解释不同 averaging time 下白频率噪声、漂移和 dead time 对稳定度曲线的影响。
- 用 light-pulse atom interferometer 的 sensitivity phase \(\phi=\mathbf k_{\rm eff}\cdot\mathbf aT^2\) 计算加速度,并列出 laser phase、wavefront、Coriolis 与 vibration 等系统效应。
- 从二项分布推导 quantum projection noise 与 \(1/\sqrt N\) standard quantum limit scaling,明确单次、重复次数和总时间资源。
- 使用 Wineland parameter 判断自旋压缩是否真正改善相位灵敏度,并把 \(\xi_R^2\) 换算为 dB metrological gain。
Prerequisites / 先修知识
- Magnetic Resonance:Ramsey pulse、Bloch vector、detuning 与 decoherence。
- Laser Cooling and Trapping:冷原子制备、recoil、trap shift 与长 interrogation time。
- 二项分布、方差传播、Fourier-limited linewidth、log–log scaling。
- Measurement and Uncertainty 与 States, Bases, and Operators:projective measurement、state/operator、expectation value 与 variance。
Scope and conventions / 范围与约定
- 本课把“clock”视为 atomic reference + interrogation + local oscillator + feedback + readout 的闭环系统;单个原子跃迁本身不会自动输出连续时间。
- Fractional frequency 写作 \(y(t)=[\nu(t)-\nu_{\rm ref}]/\nu_{\rm ref}\)。Basic Allan deviation 使用相邻、等长、无重叠 averages;overlapping/modified Allan deviation 必须另行标明。
- Atom interferometer 的 \(\mathbf k_{\rm eff}\) 是两光子 Raman/Bragg 有效波矢;标量式 \(\phi=k_{\rm eff}aT^2\) 只适用于投影在 sensitivity axis 的 constant acceleration 和 ideal pulse timing。
- \(N\) 表示一次实验中 independently measured two-level atoms;\(m\) 表示 independent repeated shots。若总 averaging time 固定,cycle time、dead time and correlations 必须计入。
- SQL 在本课专指 independent-particle phase scaling \(\Delta\phi\sim1/\sqrt N\);SNL 常用于 equivalent shot-noise reference。Heisenberg scaling \(1/N\) 只是一种 ideal resource scaling,不是任何装置都能达到的普适数值下界。
Notation and modeling contract / 符号与建模契约
| Symbol | Meaning | SI unit |
|---|---|---|
| \(\nu_0\) | unperturbed atomic transition frequency | Hz |
| \(y\) | fractional frequency offset | 1 |
| \(\sigma_y(\tau)\) | Allan deviation at averaging time \(\tau\) | 1 |
| \(T\) | free-evolution interval between light pulses | s |
| \(\mathbf k_{\rm eff}\) | effective atom-optics wavevector | \(\mathrm{m^{-1}}\) |
| \(\phi\) | interferometric phase | rad |
| \(P\) | measured excited-state probability | 1 |
| \(N\) | atoms per independent shot | 1 |
| \(\mathbf J=\frac12\sum_i\boldsymbol\sigma_i\) | collective pseudospin | 1 in units of \(\hbar\) |
| \(\xi_R^2\) | Wineland metrological squeezing parameter | 1 |
1. Atomic clocks and frequency standards / 原子钟与频率标准
Knowledge ID: P08-U04-L03-K01
1.1 Atomic transition as a feedback discriminator
SI second is defined by fixing the unperturbed ground-state hyperfine transition frequency of \(^{133}\)Cs to exactly
\[\Delta\nu_{\rm Cs}=9\,192\,631\,770\ \mathrm{Hz}.\]
实际原子钟先以 quartz/microwave/laser local oscillator 产生可调频率,再用原子跃迁测出正负 detuning,最后由 servo 修正 oscillator。原子不是“摆动九十多亿次后直接数秒”的孤立器件,而是给出可重复、可校正的误差信号。
图 1 强调 atomic line、local oscillator 和 servo 都是 clock 的组成部分。任何一环的噪声或偏差都可能投射到最终频率。
闭环工作的物理含义是:本地振荡器负责在两个原子测量周期之间连续维持相位,原子只在离散时刻告诉它“偏高还是偏低”。若振荡器在原子不被探测的死时间内发生相位波动,这些波动会经采样混叠进入钟输出;若伺服增益太小,修正追不上漂移,增益太大又可能把测量噪声放大。于是原子跃迁的固有线宽只是性能的一部分,循环时间、探测效率、误差信号斜率和反馈带宽同样决定短期稳定度。
频率标准还必须区分“定义”和“实现”。定义给出理想、未受扰动的跃迁频率;实现则在有限温度、有限磁场和实际黑体环境中测量,并逐项修正可识别的位移。只有把修正值、残余不确定度和与其他标准的比对链写清楚,测量结果才具有可追溯性。单次读数恰好接近定义值,不能替代完整的不确定度预算。
1.2 Ramsey interrogation and discriminator slope
Ideal Ramsey sequence uses two \(\pi/2\) pulses separated by free evolution \(T\)。The first pulse creates a superposition;during \(T\),atomic phase and oscillator phase separate;the second pulse converts phase difference into population。For an ideal contrast \(C\) model,
\[P_e(\Delta)=\frac12\left[1+C\cos(\Delta T+\phi_L)\right].\]
Near a steep mid-fringe operating point,small phase changes produce approximately linear population changes。Longer \(T\) narrows the fringe scale as \(1/T\),but coherence、local-oscillator phase noise and atom loss limit useful interrogation time。
图 2 中 free-evolution phase is the signal carrier;两 pulse 的 phase noise 也会进入同一 observable。
拉姆齐方法把频率差转化为随时间累积的相位差。第一次脉冲只负责建立相干叠加,自由演化阶段才积累主要频率信息,第二次脉冲把相位重新投影到可计数的能级布居。延长自由演化时间会增加同一频率偏差产生的相位,因此提高斜率;但若相干在此期间衰减,条纹对比度下降,统计优势会被抵消。最佳时间不是越长越好,而是信号斜率、原子损失、振荡器相位噪声和循环速率之间的折衷。
图 3 表明 clock 通常锁在高 slope 而非 fringe maximum;在 maximum 上,一阶 frequency error 只产生二阶 population change,判别不灵敏。
实际伺服常在中心条纹两侧交替采样,取两次布居之差作为带符号误差信号。这样可消除部分共同的探测增益变化,并判断频率偏差方向。若条纹不对称、探测非线性或两次采样之间振荡器漂移,误差信号的零点可能不再等于真实线心,所以“锁住”只表示伺服误差为零,不自动表示系统误差为零。
1.3 Quality factor、accuracy and stability
Spectroscopic quality factor is
\[Q=\frac{\nu_0}{\Delta\nu}.\]
High \(Q\) provides a steep frequency discriminator but does not automatically remove systematic shifts。Blackbody radiation、Zeeman、Stark、collision、Doppler/recoil、gravitational redshift and servo errors must be evaluated in an uncertainty budget。
For equal-duration fractional-frequency averages \(\bar y_i(\tau)\),basic non-overlapping Allan variance is
\[\sigma_y^2(\tau)=\frac{1}{2(M-1)} \sum_{i=1}^{M-1}\left[\bar y_{i+1}(\tau)-\bar y_i(\tau)\right]^2.\]
Using adjacent differences suppresses a constant frequency offset,so Allan deviation is primarily a stability statistic;it is not an accuracy certificate。
阿伦偏差把测量时间尺度写进统计量。短平均时间处,白频率噪声通常随平均时间平方根下降;延长平均后,温度漂移、老化或环境周期变化可能形成平台甚至上升。曲线转折点告诉我们“继续平均是否仍有收益”,也帮助识别主导噪声。可是预先去趋势、重叠取样或存在缺测时,估计量的自由度会改变;报告数值时必须同时给出算法、采样间隔、置信区间和是否去除漂移。
准确度评估则需要另一张表:每一种外场和实验参数引起的频移、采用的修正、灵敏系数、测量不确定度以及相关性。稳定度很好可以帮助更快测出这些修正,却不能让未知偏差自动消失。将两类表格混在一起,是精密测量报告中最危险的概念错误之一。
图 4 是 noise diagnosis map,而不是任何具体 clock 的承诺曲线。不同 noise type、sampling and detrending produce different slopes。
2. Atom interferometry / 原子干涉
Knowledge ID: P08-U04-L03-K02
2.1 Matter-wave beam splitters
Two-photon Raman or Bragg pulses coherently couple momentum states separated by approximately \(\hbar\mathbf k_{\rm eff}\)。A \(\pi/2\) pulse acts as a beam splitter,a \(\pi\) pulse redirects the arms,and a final \(\pi/2\) recombines them。The measured population is determined by the phase difference between matter-wave paths and laser phases at pulse events。
For ideal pulses at \(0,T,2T\) under constant acceleration,the leading inertial phase is
\[\phi_a=\mathbf k_{\rm eff}\cdot\mathbf aT^2.\]
The sign depends on wavevector、acceleration and output-port convention;the scalar magnitude formula must not erase this geometry。
图 5 是 momentum-state path diagram,不是原子被分成两个经典小球的照片。Wave packets must overlap coherently at recombination for high contrast。
所谓两条“路径”首先是同一原子波函数中具有不同动量的两个分量。激光脉冲既改变动量,也把当时光场的相位写入原子态;中间脉冲交换两臂角色,最后脉冲令概率振幅相加。探测到的不是连续轨迹,而是两个输出内态或动量端口的粒子数。若波包在位置、速度、内部态或时间上不能充分重叠,哪怕经典几何图看似闭合,干涉对比度仍会降低。
加速度相位的平方时间因子可理解为:速度差随时间积分成位置差,而外力作用再把路径分离积累为时空面积。有效波矢决定每次分束写入的动量和光学相位尺度。反向翻转有效波矢会翻转惯性相位,却常使某些磁移或标量光移保持不变,因此正反向测量是区分信号和系统误差的重要手段。
2.2 Sensitivity function and systematics
The \(T^2\) scaling is the space–time area response of the pulse sequence。Increasing \(T\) increases acceleration phase but also magnifies vibration、wavefront aberration、rotation/Coriolis、gravity-gradient、magnetic/Stark shifts、laser phase noise and finite-temperature sampling。Differential interferometers reject common-mode noise only to the extent that both ensembles share the same transfer function。
图 6 用 sensitivity function 表示不同时间的 laser phase/noise 如何进入 output。Finite pulse duration 会 smooth the steps,不能在 high-bandwidth noise analysis 中忽略。
灵敏度函数还能说明为何振动会伪装成重力:反射镜位置变化改变光学相位面,原子在三个脉冲时刻采样这些相位,组合后与惯性相位进入同一输出。共同平台上的参考加速度计可以估计这项噪声,但传递函数、延迟和方向耦合必须匹配。两台原子干涉仪做差也只有在脉冲时序、原子温度和波前采样相同的频段内才真正共模抑制。
3. Quantum projection noise and standard quantum limit / 量子投影噪声与标准量子极限
Knowledge ID: P08-U04-L03-K03
3.1 From Bernoulli outcomes to phase uncertainty
Each two-level atom is detected in ground or excited state。For \(N\) independent atoms with excitation probability \(P\),the estimated fraction \(\hat P=n_e/N\) has
\[\operatorname{Var}(\hat P)=\frac{P(1-P)}{N}.\]
At a Ramsey mid-fringe with \(P=1/2\) and ideal slope \(\lvert dP/d\phi\rvert=1/2\),error propagation gives
\[\Delta\phi_{\rm QPN}=\frac{\sqrt{P(1-P)/N}}{\lvert dP/d\phi\rvert} =\frac{1}{\sqrt N}.\]
After \(m\) independent identical shots,\(\Delta\phi=1/\sqrt{Nm}\)。This is a resource scaling:if cycle time changes with \(N\)、interrogation or entangling protocol,total-time comparison must include it。
图 7 把 QPN 追溯到离散 projective outcomes。它不是 detector electronics noise,尽管理想实验中两者都出现在同一 histogram 中。
量子投影噪声不是“原子数真的在测量前已经随机分成两组”的经典抽样,而是同一量子叠加态在投影测量中产生离散结果。重复制备完全相同的独立原子集合,激发数仍按二项分布涨落。探测器的光子散粒噪声、电子读出噪声和原子数归一化误差会叠加在其上,实验必须通过改变原子数、独立校准探测链和测量空白样本来区分这些来源。
从布居噪声换算相位噪声时,分母是响应曲线的局部斜率。条纹对比度降低会同时减小斜率,所以即使布居方差不变,相位灵敏度也会变差。靠选择一个斜率很小的位置让布居波动看起来很小并没有意义;灵敏度比较必须在相同的被测参数估计问题上进行误差传播。
3.2 Collective spin picture and SQL
For \(N\) uncorrelated pseudospins,a coherent spin state has \(\lvert\langle\mathbf J\rangle\rvert=N/2\) and transverse variance \((\Delta J_\perp)^2=N/4\)。Small rotation moves the mean signal proportional to \(N\),noise grows as \(\sqrt N\),hence angle uncertainty scales as \(1/\sqrt N\)。
图 8 的 uncertainty disk 是 Bloch sphere 北极附近的 local tangent-plane picture;它不是 classical ignorance over predetermined spin directions。
独立粒子的均值随原子数成正比,而方差相加后只随原子数成正比,因此标准差随平方根增长;信号斜率与噪声相除便得到平方根标度。重复实验同理:彼此独立的次数只带来平方根平均。若各次实验共享同一个漂移或振荡器噪声,误差便相关,不能再把次数直接放进平方根。总测量时间固定时,还必须把制备和读出的死时间算入可完成的重复次数。
3.3 SQL is comparison-dependent
SQL comparison must state resources:atom number per shot、number of shots、total duration、contrast、bandwidth and loss。Technical noise above SQL does not change the quantum benchmark;postselection that discards most atoms cannot claim gain without counting success probability。Ideal entangled states may approach \(1/N\) scaling,but decoherence often changes optimum interrogation and can erase asymptotic advantage。
“超过标准量子极限”应理解为相对于一个明确定义的独立粒子基准更好,而不是超过一条脱离实验条件的永恒直线。若纠缠态制备时间更长、成功概率更低或可用原子更少,公平比较应按单位总时间和实际成功事件计入资源。技术噪声尚未压到基准以下时,实验可以证明产生了量子关联,却未必已经改善最终传感器性能;这两种结论应分别报告。
4. Squeezing and quantum-enhanced sensing / 压缩态与量子增强传感
Knowledge ID: P08-U04-L03-K04
4.1 Noise redistribution versus metrological gain
Spin squeezing redistributes collective quantum uncertainty:one transverse quadrature narrows while its conjugate broadens。But merely observing \((\Delta J_\perp)^2<N/4\) is not sufficient if the mean spin contrast has collapsed。Wineland parameter includes both noise and usable signal slope:
\[\xi_R^2=\frac{N(\Delta J_\perp)^2}{\lvert\langle\mathbf J\rangle\rvert^2}.\]
\(\xi_R^2<1\) means phase sensitivity below the coherent-state SQL under the corresponding linear readout assumptions。Metrological gain in dB is
\[G_{\rm dB}=-10\log_{10}\xi_R^2.\]
图 9 显示 squeezed ellipse 的 narrow axis 必须与待测 rotation 的 readout quadrature 对齐;若 orientation 错误,anti-squeezed noise 会降低灵敏度。
压缩并不是同时把所有方向的不确定度都变小。量子对易关系要求一个方向变窄时,共轭方向通常变宽;实验通过相互作用建立粒子间关联,使单粒子噪声不再独立相加。待测相位必须把平均自旋转向窄轴所对应的读出方向,若相位预估误差过大,反压缩方向的宽噪声会泄漏进结果。这也是压缩态常在小动态范围内最有优势的原因。
参数中平均自旋长度位于分母,专门惩罚制备过程造成的对比度损失。某方案可能把某一分量方差压得很低,却同时随机化大量原子,使平均信号几乎消失;这种状态在图上看起来“很窄”,但无法提供更陡的相位响应。只有噪声与可用信号长度合并后仍小于独立粒子基准,才可声称线性相位测量获得实际增益。
4.2 Entanglement, readout and decoherence
Useful squeezing may arise from one-axis twisting、cavity-mediated interaction、QND measurement or Rydberg interactions。Preparation itself consumes time and can introduce loss;finite detection resolution can hide or mimic sub-SQL noise;anti-squeezed quadrature is more sensitive to phase misalignment。完整验证需给出 calibration、contrast correction、technical-noise subtraction policy and independent SQL reference。
验证量子增强需要完整的噪声账本。首先用相同原子数和相同探测链建立相干自旋基准;其次测量压缩轴和反压缩轴,确认方差重新分配而非电子滤波造成;再校准平均自旋长度、探测分辨率和原子数涨落;最后把制备损失、额外时间和成功概率计入每单位时间灵敏度。若从数据中扣除技术噪声,必须同时报告未扣除结果和扣除方法,避免把模型假设当成实验增益。
图 10 说明 ideal Heisenberg scaling 不保证 large-\(N\) device advantage;随 \(N\) 增长的 decoherence 或 preparation overhead 可产生 optimum finite size。
海森堡标度描述理想无损条件下对给定量子资源的最佳渐近趋势,并不保证任何有限装置都能达到。粒子损失会泄露“哪条路径”的信息,退相干会破坏多体关联,控制误差常随系统规模累积;高纠缠态因此可能比轻度压缩态更脆弱。工程上常见的最优方案是在适度关联、足够对比度和可重复循环之间取平衡,而不是单纯追求最陡的理论幂律。
Worked Examples / 完整例题
Worked Example 1 — Optical-clock quality factor and time error
已知: clock transition \(\nu_0=4.29\times10^{14}\ \mathrm{Hz}\),observed linewidth \(\Delta\nu=1.00\ \mathrm{Hz}\);corrected fractional frequency bias magnitude \(\lvert y\rvert=2.0\times10^{-18}\)。
研究对象: disciplined oscillator relative to the stated atomic reference。
系统边界: clock output and reference;assume constant residual bias over one day。
参考系/坐标: laboratory proper time;ignore relativistic height change during interval。
模型: constant fractional offset,linewidth defined as FWHM。
\[Q=\frac{\nu_0}{\Delta\nu}=4.29\times10^{14}.\]
For \(t=86400\ \mathrm s\),accumulated time offset magnitude is
\[\lvert\Delta t\rvert\simeq\lvert y\rvert t =1.73\times10^{-13}\ \mathrm s=0.173\ \mathrm{ps}.\]
检查: fractional offset times seconds gives seconds;\(Q^{-1}\) and \(y\) are distinct—linewidth does not equal systematic bias。解释: 这是假设 bias constant 的示例;真实 uncertainty budget 和 noise process determine confidence。
Worked Example 2 — Basic Allan deviation
已知: four adjacent equal-duration fractional-frequency averages are \([1,-1,2,0]\times10^{-13}\)。
研究对象: sampled oscillator stability at that averaging time。
system boundary/reference: oscillator minus fixed reference,no missing samples。
time coordinate: non-overlapping intervals of duration \(\tau\)。
模型: basic non-overlapping Allan deviation,no detrending。
Adjacent differences are \([-2,3,-2]\times10^{-13}\),so
\[\sigma_y(\tau)=\sqrt{\frac{4+9+4}{2(4-1)}}\times10^{-13} =1.68\times10^{-13}.\]
检查: result dimensionless and comparable to individual \(y_i\) scale。Adding the same constant to every sample leaves it unchanged。解释: This number estimates stability for this sampling;it does not reveal an unknown common bias。
Worked Example 3 — Atom-interferometric acceleration phase
已知: counterpropagating Raman beams near \(780\ \mathrm{nm}\),\(k_{\rm eff}\simeq4\pi/\lambda\);pulse separation \(T=0.100\ \mathrm s\);acceleration projection \(a=9.81\ \mathrm{m\,s^{-2}}\)。
研究对象: ideal \(\pi/2-\pi-\pi/2\) atom interferometer。
系统/reference: freely falling atoms and laser phase fronts in lab frame。
坐标: \(+z\) along \(\mathbf k_{\rm eff}\) and \(a\) taken positive along it。
模型: instantaneous pulses、constant acceleration、no chirp compensation。
\[k_{\rm eff}=\frac{4\pi}{780\times10^{-9}} =1.61\times10^7\ \mathrm{m^{-1}},\]
\[\phi=k_{\rm eff}aT^2 =1.58\times10^6\ \mathrm{rad}.\]
检查: \(k aT^2\) is dimensionless;large raw phase is expected。解释: Practical gravimeters chirp Raman frequency and infer residual phase modulo \(2\pi\);vibration isolation/reference accelerometers prevent fringe ambiguity。
Worked Example 4 — QPN and 6 dB spin squeezing
已知: \(N=10^4\) atoms per shot,\(m=100\) independent shots,ideal mid-fringe contrast;squeezed protocol has \(\xi_R^2=0.25\) with same declared \(N,m,T\) and cycle time。
研究对象: phase estimate from binary atomic outcomes。
system/reference: atomic ensemble plus ideal detector;technical noise ignored。
coordinate: small phase around maximum-slope quadrature。
模型: independent CSS benchmark and Wineland linear-readout gain。
\[\Delta\phi_{\rm SQL}=\frac{1}{\sqrt{Nm}}=1.00\times10^{-3}\ \mathrm{rad}.\]
Squeezed phase uncertainty is
\[\Delta\phi_{\rm sq}=\sqrt{\xi_R^2}\,\Delta\phi_{\rm SQL} =5.00\times10^{-4}\ \mathrm{rad}.\]
\[G_{\rm dB}=-10\log_{10}(0.25)=6.02\ \mathrm{dB}.\]
检查: variance improves by factor 4, standard deviation by factor 2。解释: Claim is valid only because resources and contrast are already included in \(\xi_R^2\) and held equal;extra preparation time or loss would need inclusion。
Lesson synthesis / 全课综合
精密测量不是单独追求“小数点后更多位”。它把可重复 quantum phase 映射到 population,再用统计学把随机涨落和偏差分开。原子钟比较 atomic phase 与 oscillator phase;atom interferometer 比较两条 matter-wave paths 与 laser phase;QPN 给出 independent particles 的 sampling floor;squeezing 通过相关性重塑 noise ellipse。每一种改进都必须在明确的资源、带宽、对比度和系统误差预算下才有意义。
建模流程:
- 定义 measurand、reference、sign、coordinate and sampling interval。
- 写出 signal transfer function,例如 \(P(\phi)\) or \(\phi=k_{\rm eff}aT^2\)。
- 将 frequency/phase noise 传播到 output;再与 detector noise、QPN 分开。
- 用 Allan deviation 报 stability,用 uncertainty budget 报 corrected accuracy。
- 比较 quantum enhancement 时固定 \(N,m,T\)、cycle time、contrast and success probability。
Formula-and-decision table / 公式与决策表
| Task | Relation | Use only when | Boundary |
|---|---|---|---|
| clock quality | \(Q=\nu_0/\Delta\nu\) | linewidth convention stated | high \(Q\) does not prove accuracy |
| fractional offset | \(y=(\nu-\nu_{\rm ref})/\nu_{\rm ref}\) | reference and relativistic frame stated | drift/time dependence may matter |
| basic Allan deviation | \(\sigma_y^2=[2(M-1)]^{-1}\sum(\bar y_{i+1}-\bar y_i)^2\) | equal adjacent non-overlapping averages | not an accuracy metric |
| inertial phase | \(\phi=\mathbf k_{\rm eff}\cdot\mathbf aT^2\) | ideal three-pulse constant-\(a\) geometry | add laser phase, rotations, gradients, finite pulses |
| projection variance | \(\operatorname{Var}(\hat P)=P(1-P)/N\) | independent Bernoulli outcomes | detection noise/correlations add or alter variance |
| SQL phase | \(\Delta\phi=1/\sqrt{Nm}\) | ideal mid-fringe、independent shots | fixed-time comparison needs cycle time |
| Wineland squeezing | \(\xi_R^2=N(\Delta J_\perp)^2/\lvert\langle\mathbf J\rangle\rvert^2\) | linear collective-spin readout | axis, contrast and detection calibration required |
| dB gain | \(G_{\rm dB}=-10\log_{10}\xi_R^2\) | \(\xi_R^2\) is a variance ratio | amplitude ratios would use a different interpretation |
Model boundaries / 模型边界
- The SI second definition specifies an unperturbed Cs frequency exactly;realization requires corrections and an uncertainty statement。
- Ramsey linewidth depends on pulse shape、contrast and chosen width definition;only the \(1/T\) scale is universal in the ideal limit。
- Basic Allan deviation assumes an explicit sampling scheme。Overlapping、modified、Hadamard deviations answer related but different noise questions。
- Simple \(k_{\rm eff}aT^2\) ignores Earth rotation、gravity gradient、finite speed of light、wavefront curvature and pulse duration。
- QPN formula assumes independent atoms and binary projective readout;technical detection noise can be below or above it without changing the benchmark。
- Spin squeezing below SQL requires correlation/entanglement under appropriate conditions,but sub-Poissonian count noise alone may result from classical preparation constraints。
- Ideal Heisenberg scaling is fragile under loss/decoherence;a fair sensor comparison uses total Fisher information per unit time or equivalent declared resources。
Misconception clinic / 误区诊所
- “Atomic transition itself is a clock.” A usable clock needs oscillator、interrogation、detection、servo and an output timescale。
- “Allan deviation 越小就越准确。” It quantifies stability versus averaging time;a constant systematic offset cancels in adjacent differences。
- “Higher \(Q\) means zero systematic error.” High \(Q\) narrows discriminator;environmental shifts can still bias the center。
- “Atom interferometer measures a classical path separation only.” Output comes from coherent phase including laser phases at pulse events。
- “QPN is bad detector electronics.” It remains with ideal detection because \(N\) quantum outcomes are probabilistic。
- “\(1/\sqrt N\) improvement and \(1/N\) improvement can be compared without conditions.” Resource count、total time、contrast、loss and bandwidth must match。
- “Any squeezed variance proves metrological gain.” Reduced contrast or wrong quadrature can eliminate phase advantage;Wineland parameter checks both noise and signal length。
Practice and retrieval / 练习与提取
- [K01 定义] 写出 SI second 对 \(^{133}\)Cs transition frequency 的 exact numerical definition。
- [K01 概念] 一个 oscillator 的 Allan deviation 很小但 frequency bias 为 \(10^{-9}\);它稳定还是准确?
- [K01 定量] \(\nu_0=5.0\times10^{14}\ \mathrm{Hz}\)、\(\Delta\nu=2.0\ \mathrm{Hz}\),求 \(Q\)。
- [K02 定量] \(k_{\rm eff}=1.0\times10^7\ \mathrm{m^{-1}}\)、\(T=50\ \mathrm{ms}\)、\(a=1.0\times10^{-6}\ \mathrm{m\,s^{-2}}\),求 inertial phase magnitude。
- [K02 边界] 列出两个会伪装成 acceleration phase 的系统效应。
- [K03 推导] 从 binomial variance 推导 \(P=1/2\) 时 \(\Delta P=1/(2\sqrt N)\)。
- [K03 定量] \(N=2500\)、\(m=16\),求 ideal SQL phase uncertainty。
- [K03 判断] 将 atom number 增 4 倍,在其他资源相同且 independent 时 phase standard deviation 改变多少?
- [K04 定量] \(\xi_R^2=0.10\) 对应多少 dB metrological gain?phase standard deviation 相对 SQL 是多少?
- [K04 综合] 一个 experiment reports transverse variance 低于 CSS 8 dB,但 mean spin contrast 只剩 20%。为什么还不能直接声称 8 dB metrological gain?
Answers / 答案
- Unperturbed \(^{133}\)Cs ground-state hyperfine transition frequency is exactly \(9\,192\,631\,770\ \mathrm{Hz}\)。
- It may be highly stable but inaccurate relative to reference;constant bias is not revealed by Allan deviation。
- \(Q=\nu_0/\Delta\nu=2.5\times10^{14}\)。
- \(\phi=k_{\rm eff}aT^2=(10^7)(10^{-6})(0.050)^2=0.025\ \mathrm{rad}\)。
- 例如 mirror vibration/laser phase noise、wavefront curvature、Coriolis rotation、gravity gradient、magnetic/Stark shift,任答二。
- \(n_e\sim\operatorname{Binomial}(N,1/2)\) has variance \(N/4\);divide by \(N^2\) for fraction variance \(1/(4N)\),so \(\Delta P=1/(2\sqrt N)\)。
- \(\Delta\phi=1/\sqrt{Nm}=1/\sqrt{40000}=5.00\times10^{-3}\ \mathrm{rad}\)。
- Standard deviation improves by \(\sqrt4=2\),becoming half。
- \(G_{\rm dB}=10.0\ \mathrm{dB}\);standard deviation ratio is \(\sqrt{0.10}=0.316\)。
- Phase sensitivity depends on variance divided by signal slope;contrast loss reduces \(\lvert\langle J\rangle\rvert\)。Need compute \(\xi_R^2=N(\Delta J_\perp)^2/\lvert\langle J\rangle\rvert^2\) and include preparation/detection resources。
Exit ticket / 离堂检验
- 为一个“比 SQL 好 3 dB”的 atomic sensor claim 写出最小验收清单:至少包含 benchmark、atom number、shots/total time、contrast、detection noise and confidence interval。
- 用一张因果链说明 clock 中 local-oscillator phase noise 怎样经过 Ramsey interrogation、population readout and servo 进入 Allan deviation。
References and source notes / 参考资料与来源说明
以下权威直接来源均于 2026-08-30 实际打开核验;所有统计/相位公式重新推导,所有 TikZ 为原创。
- BIPM, SI base unit: second:current SI definition and exact \(^{133}\)Cs frequency。https://www.bipm.org/en/si-base-units/second
- BIPM, Mise en pratique for the definition of the second in the SI:primary/secondary frequency standards and realization context。https://www.bipm.org/documents/20126/41489667/SI-App2-second.pdf
- NIST, How NIST-F1 Works / Realization of the Official U.S. Time:atomic fountain/time-scale feedback and Allan-deviation context。https://www.nist.gov/pml/time-and-frequency-division/time-services/utcnist-time-scale/how-utcnist-works
- Ludlow et al., Optical Atomic Clocks, Reviews of Modern Physics, NIST-hosted direct PDF:optical-clock interrogation、systematics and stability。https://tf.nist.gov/general/pdf/2740.pdf
- Cronin, Schmiedmayer, and Pritchard, Atom Interferometers author preprint:atom-optics beam splitters、interferometer phase and applications。https://arxiv.org/abs/0712.3703
- NIST, Turning Atoms into Waves to Measure Gravity and Time:light-pulse atom interferometry and inertial sensing。https://www.nist.gov/quantum-information-science/quantum-sensing-explained/turning-atoms-waves-measure-gravity-and
- NIST, New Spin-Squeezing Techniques Let Atoms Work Together Better for Quantum Measurements:collective spin squeezing、correlations and metrological motivation。https://www.nist.gov/news-events/news/2023/09/new-spin-squeezing-techniques-let-atoms-work-together-better-quantum
- NIST-hosted paper, Spin squeezing with Rydberg interactions:direct experimental squeezing methods and limitations。https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=936513
Source-to-content map
| Source | Used for | Knowledge ID / Section | Model boundary or convention |
|---|---|---|---|
| BIPM second page | exact SI definition | K01 | definition distinguished from realization |
| BIPM mise en pratique | frequency-standard realization and traceability | K01 | primary/secondary standards, current document context |
| NIST time-scale page | clock feedback、time scale、stability | K01 | operational description; basic Allan formula stated explicitly here |
| NIST-hosted Optical Clocks RMP | Ramsey interrogation、clock systematics、stability | K01, K03 | optical and microwave clocks share framework but differ transitions |
| Cronin et al. author preprint | matter-wave splitting and inertial phase | K02 | scalar \(k_{\rm eff}aT^2\) labeled ideal geometry |
| NIST Atom Waves | gravity/time sensing application | K02 | explanatory application cross-check, not sole equation source |
| NIST Spin Squeezing pages/paper | QPN、collective correlations、squeezing | K03, K04 | Wineland metric includes contrast and readout axis |
复习与拓展
- 与 Magnetic Resonance 相连:Ramsey sequence and collective spin geometry are the same control language。
- 与 Laser Cooling and Trapping 相连:temperature、trap shifts、wavefront sampling and long coherence limit clock/interferometer accuracy。
- 与 planned Interatomic Bonding and Material Classes 跨 Part 衔接(后续课程尚未制作):精密谱学给出原子/分子参数,而下一 Part 转向许多原子结合形成材料后的 emergent structure。
- 可扩展:Dick effect、servo transfer functions、quantum Fisher information、continuous measurement、networked clocks、gravitational redshift geodesy。