From the 17th century to the late 19th century, planetary ephemerides were calculated using time scales based on the Earth's rotation: usually the
mean solar time of one of the principal observatories, such as Paris or Greenwich. After 1884, mean solar time at Greenwich became a standard, later named
Universal Time (UT). But in the later 19th and early 20th centuries, with the 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.
Ephemeris time was consequently developed as a standard that was free from the irregularities of Earth rotation, by defining the time "as the independent variable of the equations of celestial mechanics", and it was at first measured astronomically, relying on the existing gravitational theories of the motions of the Earth about the Sun and of the Moon about the Earth. After the caesium
atomic clock was invented, such clocks were used increasingly from the late 1950s as
secondary realizations of ephemeris time (ET). These secondary realizations improved on the original ET standard by the improved uniformity of the atomic clocks, and (e.g. in the late 1960s) they were used to provide standard time for planetary ephemeris calculations and in astrodynamics. But ET in principle did not yet take account of relativity theory. The size of the periodic part of the variations due to
time dilation between earth-based atomic clocks and the
coordinate time of the Solar-System barycentric reference frame had been estimated at under 2 milliseconds, to replace ET (in the ephemerides for 1984 and afterwards) to take account of relativity. ET's direct successor for measuring time on a geocentric basis was
Terrestrial Dynamical Time (TDT). The new time scale to supersede ET for planetary ephemerides was to be Barycentric Dynamical Time (TDB). TDB was to tick uniformly in a reference frame comoving with the
barycenter of the Solar System. (As with any
coordinate time, a corresponding clock, to coincide in rate, would need not only to be at rest in that reference frame, but also (an unattainable hypothetical condition) to be located outside all of the relevant
gravity wells.) In addition, TDB was to have (as observed/evaluated at the Earth's surface), over the long term average, the same rate as TDT (now
TT). TDT and TDB were defined in a series of resolutions at the same 1976 meeting of the
International Astronomical Union. It was eventually realized that TDB was not well defined because it was not accompanied by a general relativistic metric and because the exact relationship between TDB and TDT had not been specified. (It was also later criticized as being not physically possible in exact accordance with its original definition: among other things the 1976 definition excluded a necessary small offset for the initial epoch of 1977.) After the difficulties were appreciated, in 1991 the IAU refined the official definitions of timescales by creating additional new time scales:
Barycentric Coordinate Time (TCB) and
Geocentric Coordinate Time (TCG). TCB was intended as a replacement for TDB, and TCG was its equivalent for use in near-Earth space. TDT was also renamed to
Terrestrial Time (TT), because of doubts raised about the appropriateness of the word "dynamical" in that connection. In 2006 TDB was redefined by IAU 2006 resolution 3; the 'new' TDB was expressly acknowledged as equivalent for practical purposes to
JPL ephemeris time argument Teph; the difference between TDB according to the 2006 standard and
TT (both as observed from the surface of the Earth), remains under 2 ms for several millennia around the present epoch. ==Use of TDB==