The Atomic Wristwatch

Modern society relies heavily on precise timing. The most accurate timing devices — the primary atomic clocks in national metrology laboratories — each take up as much space as a large wardrobe and keep time with an accuracy of 1 part in 10¹⁵, or 1 second in 30 million years. Not everyone needs this level of performance, and stability to about 1 part in 10¹² is possible in shoebox-sized clocks. Several thousand such devices are in use worldwide, for instance in telecommunications, to synchronize high-rate data transmission in multi-user networks. A much wider range of application would open up if the clocks could be miniaturized further and produced in an affordable way.

Solutions have now been found to some of the physical and technical problems encountered on the road to tiny atomic clocks. Liew et al describe the miniaturization of a vapour-cell atomic-frequency standard to a volume of just a few cubic millimetres, using a sequence of steps that are fully compatible with mass production. In Physical Review Letters, Jau et al. propose a novel mode of operation that could overcome some of the inherent disadvantages of small vapour-cell clocks. The result could be atomic clocks costing less than €100 (US$120) each, whose size and power consumption would be suitable for hand-held devices, perhaps even for high-tech wristwatches.

In a traditional atomic clock, atoms are illuminated by microwaves, in the gigahertz frequency range. The microwave frequency can be regulated such that its absorption by an atom induces transitions between two specific internal atomic energy states at the highest rate possible. In this way, the regulated frequency of the microwave, the primary output signal of the atomic clock, becomes a macroscopic representation of the internal structure of the atom. In a caesium clock, the optimum frequency corresponds to 9,192,631,770 oscillations of the microwave per second (just over 9 GHz in frequency), so dividing the output electronically by this number provides one-second ‘ticks’ of the atomic clock.

Miniaturization of such a microwave-driven clock cannot go below a centimetre or so because the microwave cannot fit into guide structures much smaller than its 3.2-cm wavelength. But this barrier can be broken by using laser light. And Liew et al. have now carried this work a crucial step further, by developing a new manufacturing technique for tiny vapour cells.

Their latest results include the integration of the complete optical set-up (the ‘physics package’) in a device that takes up only a few cubic millimetres, not counting its electronics and power supply. Furthermore, all of the processing steps are fully compatible with existing technology for wafer-scale integration and processing. In principle, hundreds or thousands of physics packages could be manufactured simultaneously, keeping the manufacturing cost per unit low. The overall power consumption of the complete clock would be a few tens of milliwatts, so battery operation becomes a possibility.

But the miniaturization does come at the price of reduced stability: 2 parts in 10¹⁰ because in a miniature clock, the vapour cell is so short that the vapour density must be increased. To do this, the caesium atoms are heated to a temperature of more than 80 °C, so that the same total number of atoms is producing a signal. At those high densities, the atoms frequently collide with each other. Each collision disturbs the atom’s interaction with the laser light or the microwave in a process called spin exchange, which causes a broadening of the resonance line. Collisions can also transfer atoms away from the two energy levels involved in the clock transition, with the result that the effective number of atoms taking part in signal generation, and hence the SNR, is reduced.

Jau et al. propose to counter these effects by choosing a different pair of states, at the outer edge of the ground-state multiplet and the clock's stability might be well improved by a factor of ten to a hundred, taking the small clocks into the performance range — better than 1 in 10¹² — of existing devices that cost about €50,000.

It will still be a while before a miniature atomic clock like that of Liew et al. becomes a commercial product. Would people really want an atomic wristwatch? An accuracy of 1 part in 10¹¹ corresponds to a microsecond per day — perhaps a bit overblown for normal everyday use. But its small size and estimated low price will make it an important part of other products, such as improved GPS receivers or timing-dependent devices designed for autonomous operation. If the performance gains predicted by Jau et al. can be realized, we will then have affordable atomic time for the mass market.

原子表

现代社会高度依赖于精确的计时。最精确的计时装置是美国国家度量衡实验室里的一级原子钟，每台有大衣柜那么大，计时精度达到1/10¹⁵，即每3000万年差一秒。但是并不是人人都需要这么高的性能，现在已经有鞋盒大小的钟，稳定性可达大约1/10¹²。数百个这种装置已在全球各地使用，如用于电讯，以便在多用户网络上同步高速度的数据传输。如果这些钟能够做到体积更小，价格合理，那么它的应用将更加广泛。

现在已经发现了一些方法，可以解决在微型化原子钟时所碰到的一些物理和技术问题。Liew等人描述了一个蒸汽元件原子频率标准的小型化，能够把体积缩小到仅仅几个立方毫米，而这些步骤完全符合大规模生产的要求。Jau等人在《物理评论快报》上提出了一种新的操作模式能够克服小型蒸汽元件钟的一些内在缺陷。其结果是每个原子钟的价格能够降到不到100英镑 (120美元)，体积和能耗完全适于手持式设备，甚至有可能用于高技术的原子表。

在传统的原子钟里，原子受到吉赫波段的微波照射。微波的频率能够调节，以便原子吸收微波，引起在两个特定的原子内部能级上的尽可能高比率的跃迁。这样，微波的调节频率，原子钟的原始输出信号就成为表示原子内部结构的宏观量。对于铯原子钟，其最佳频率是微波每秒振动9,192,631,770次(频率略高于9 GHz)，而用电路把输出的信号除以这个数字就提供了原子钟的一声“滴答”。

这种微波驱动的钟不能缩小到厘米见方，因为在比它3.2厘米波长小得多的导向结构里，微波便不能产生。但使用激光就可以打破这个障碍。现在Liew等人开发出一项新的微型蒸汽元件制造技术，推动这项工作了迈出了关键性的一步。

他们的最新结果包括在一个仅有几个立方毫米大小的装置上(不计电路和电源)集成整个光学装置(物理模块)。而且所有的处理步骤都完全符合现有的晶片级集成和处理技术。在理论上可以同时生产成百上千个物理模块，使得每个模块的生产成本很低。整台钟的总能耗是数十毫瓦，这样就可以用电池来供电。

但是小型化的确是以牺牲稳定性为代价的，精度只有2/10¹⁰。这是因为小型钟里的蒸汽元件太小了，必须提高蒸汽密度。为了达到这一目标，铯原子被加热到摄氏八十多度，这样产生信号的总原子数还是保持不变。在那么高的密度下，原子之间经常相互碰撞。每一次的碰撞都会产生自旋交换过程，该过程会干扰原子和激光或者微波的相互作用，导致共振谱线的变宽。碰撞也使得原子离开和原子钟里的跃迁相关的那两个能级，导致参与发生信号的有效原子数减少，信噪比也随之下降。

Jau等人提出了可以通过在基态多重态的外缘另外选择一对能级来消除这些影响。钟的稳定性可能会提高成百十倍，使这种小型钟能够进入高性能领域，精度优于1/10¹²，而现存的同类设备要价大约5万英镑。

Liew等人提出的小型原子钟成为商品还要过一段时间。人们是否真的想要一块原子表？1/10¹¹的精度相当于每天一个毫秒，也许对于一般的日常使用有点夸张。但它体积小巧，价格低廉，将成为别的产品的重要零件，如改进的GPS接收机或者设计用于独立操作的时控装置。如果Jau等人预测的性能提升能够实现，那么我们将能在大众市场上买到价格合理的原子时钟。