A comparison of two independent atomic time scales

Although the use of atomic beam devices clearly permits the meas~iremeiit of time i i i tervals of up to several years with a precision which is 100 times better than that available from astronomical ~iieasurements, the problem of preserving epoch exists if it is wished toreplace the traditional astronomical timekeeping methods by quantum electronic techniques. Two main approaches to this problem exist. The first involves securing reliable nonintermittent operation o f an atomic frequency standard with clock-driving quartz oscillators locked to the instrument for realizing the atomic transition. \Vith this approach, more than one atomic standard is necessary in each laboratory, siiice the expected lifetime for reliable operatioti of a single iiistrunicnt is far from iiilinitc. The serond, less expensive method requires intermittent use of the atomic standards to calibrate free-running quartz oscillators which drive rlocks. These calibrations are then used to convert indicated quartz time to atomic time. The accuracy of this method depends 011, among other things, how often the calibrations are made. One must also consider possible systematic errors arising from the possibility that the oscillators exhibit cyclic frequency variations having a frequency coiicidiiig with the frequency with which the calibrations were made. This type of error could arise, for instance, if during the nieasuring process the loading 011 thc oscillator changed and this i n turn induced a change ill the frequency a t which the oscillator was operating. This type of error could happen also if frequency meas~~reiiients were made a t iiii interval which is some integral multiple of 24 hours. Ambient temperatures and supply voltages ordinarily have 24-hour periods. 'Therefore, if the voltage and temperature regulating circuits of the oscillator were not functioning adequately because of design or deterioration of the componeiits, a systematic error might be expected. Such a mechanism is known to have affected the measurements reported below by about 1 part in 10". However, some quartz oscillators which were designed more recently than those used in obtaining data for this paper show smaller accumulated time errors due to diurnal frequency variations even though the frequency measurement schedule was maintained on a daily basis over a period of several months. The National Bureau of Standards, Boulder, Colo., has assigned atomic times to \T\ZT pulses on a daily basis for a period of over four years. These assignments are made using daily frequency measurements of \V\\'V based on the United States Frequency Standard assuming an atomic second as equivalent to 9,192,631,770.000. . . oscillations of a source resonant with the zero-held hyperfine level spacing of C S ' ~ ~ . 1 hey have been compared to similar assignments made by the U. S. Naval Observatory, \\'ashiiigton, I). C., according to their A.l scale.' The variations in the daily differences in times as assigned by the Naval Observatory and by NBS are plotted in Fig. 1. One would expect that over a period of 4 years measuring techniques would improve, causing a plot surh as Fig. 1 to have smaller variations; and systematic errors would be found and eliminated, causing an approach to a more nearly horizontal line as time progresses. The latter has not happened, as is shown in Fig. 1. Analysis of the data seems to indicate that, although the general agreement tends to be within one part in lo'", a persistent slope of about I .