An impression has sometimes arisen that ephemeris time was in use from 1900: this probably arose because ET, though proposed and adopted in the period 1948–1952, was defined in detail using formulae that made retrospective use of the epoch date of 1900 January 0 and of Newcomb's Tables of the Sun.
Ephemeris time (ET), adopted as standard in 1952, was originally designed as an approach to a uniform time scale, to be freed from the effects of irregularity in the rotation of the Earth, "for the convenience of astronomers and other scientists", for example for use in ephemerides of the Sun (as observed from the Earth), the Moon, and the planets.
But in the later nineteenth and early twentieth centuries, with increasing precision of astronomical measurements, it began to be suspected, and was eventually established, that the rotation of the Earth (i.e. the length of the day) showed irregularities on short time scales, and was slowing down on longer time scales.
Other astronomers of the period also made suggestions for obtaining uniform time, including A Danjon (1929), who suggested in effect that observed positions of the Moon, Sun and planets, when compared with their well-established gravitational ephemerides, could better and more uniformly define and determine time.
[10] Clemence (1948) made it clear that his proposal was intended "for the convenience of astronomers and other scientists only" and that it was "logical to continue the use of mean solar time for civil purposes".
Ephemeris time was defined in principle by the orbital motion of the Earth around the Sun[12] (but its practical implementation was usually achieved in another way, see below).
Its detailed definition was based on Simon Newcomb's Tables of the Sun (1895),[5] implemented in a new way to accommodate certain observed discrepancies: In the introduction to Tables of the Sun, the basis of the tables (p. 9) includes a formula for the Sun's mean longitude at a time, indicated by interval T (in units of Julian centuries of 36525 mean solar days[19]), reckoned from Greenwich Mean Noon on 0 January 1900: Spencer Jones' work of 1939[10] showed that differences between the observed positions of the Sun and the predicted positions given by Newcomb's formula demonstrated the need for the following correction to the formula: where "the times of observation are in Universal time, not corrected to Newtonian time," and 0.0748B represents an irregular fluctuation calculated from lunar observations.
Clemence's formula (today superseded by more modern estimations) was included in the original conference decision on ephemeris time.
In view of the fluctuation term, practical determination of the difference between ephemeris time and UT depended on observation.
Inspection of the formulae above shows that the (ideally constant) units of ephemeris time have been, for the whole of the twentieth century, very slightly shorter than the corresponding (but not precisely constant) units of mean solar time (which, besides their irregular fluctuations, tend to lengthen gradually).
[25] Reasons for the use of lunar measurements were practically based: the Moon moves against the background of stars about 13 times as fast as the Sun's corresponding rate of motion, and the accuracy of time determinations from lunar measurements is correspondingly greater.
[4] Partly in acknowledgement of the widespread use of Teph via the JPL ephemerides, IAU resolution 3 of 2006[29] (re-)defined Barycentric Dynamical Time (TDB) as a current standard.
Thus the new TDB, like Teph, is essentially a more refined continuation of the older ephemeris time ET and (apart from the < 2 ms periodic fluctuations) has the same mean rate as that established for ET in the 1950s.
[30] (But the ephemerides in the Nautical Almanac, by then a separate publication for the use of navigators, continued to be expressed in terms of UT.)
The relation with Newcomb's coefficient can be seen from: Caesium atomic clocks became operational in 1955, and quickly confirmed the evidence that the rotation of the Earth fluctuated irregularly.
After three years of comparisons with lunar observations, Markowitz et al. (1958) determined that the ephemeris second corresponded to 9 192 631 770 ± 20 cycles of the chosen cesium resonance.
[17] Following this, in 1967/68, the General Conference on Weights and Measures (CGPM) replaced the definition of the SI second by the following: The second is the duration of 9 192 631 770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom.
The difference between ET and UT is called ΔT; it changes irregularly, but the long-term trend is parabolic, decreasing from ancient times until the nineteenth century,[22] and increasing since then at a rate corresponding to an increase in the solar day length of 1.7 ms per century (see leap seconds).