After several individual laboratories with cesium atomic chronometers started constructing atomic time scales the CCDS meeting in 1961 suggested that BIH should construct an international version. BIH complied and produced tables of Temps Atomique Integre based on all of the available atomic chronometers.
Essen and Parry at UK NPL in Teddington along with Markowitz and Hall at USNO in Washington published papers in 1958 April with the results from several years of comparing the atomic frequency standard in Teddington with observations using the dual rate moon camera. This allowed a comparison of Ephemeris Time computed using the Improved Lunar Ephemeris and the frequency of the cesium resonance.
Before those publications there was no definition of an atomic second. After those publications it became possible to review the history of observations of cesium frequency standards and construct atomic time scales.
The first cesium atomic frequency standard at UK NPL in Teddington began operation in 1955; this was designated ET. In late 1956 USNO in Washington began observations of its cesium atomic frequency standard; this was designated WL. In early tabulations the BIH designated the average of all available atomic time scales as AM. The following plot shows the values of those time scales as tabulated by BIH.
From the beginning of 1957 BIH received observations of the cesium atomic frequency standard in Neuchâtel; this was designated N.
The BIH tabulations during 1957 November show that the value of AM was briefly constructed with double weight for the values of WL. Close inspection of plots made from the BIH tabulations of early atomic time show several other instances which are almost certainly arithmetic mistakes or typos. These were not visible at the time and no errata were published.
In 1970 the CCDS authorized the BIH to create TAI. BIH constructed TAI using the time scale they had been calling A3 since they first created it starting in 1961.
During the 1970s it became clear that the BIH constructions of TAI had always ticked at a rate that was running faster than the SI second at sea level. This was due to characteristics of the stability algorithm which TAI used when constructing TAI from the raw inputs. In 1976 the IAU and CIPM said that the TAI scale should be adjusted to match the SI second. On 1977-01-01 the BIH changed their definition of TAI and reduced its rate by 1e-12. The old TAI was continued under the new name EAL. The following plots are by Gerhard Becker of PTB.
During the 1970s USNO purchased more HP cesium chronometers than any other time laboratory. As a result the chronometers at USNO became the preponderance of statistical weight in the construction. Unfortunately the time laboratory at USNO did not have air conditioning. That lab experienced large seasonal variations in humidity. The chronometers suffered large seasonal variations in rate, and those variations became part of TAI.
At CCDS and CIPM meetings around 1970 Gerhard Becker of PTB requested a formal definition of TAI including a definition of its epoch. Those bodies never produced a formal definition other than the atomic time scale maintained by the BIH. I suspect that lacuna exists because they did not want to publicize these details of TAI.
