Back-Reaction of Clocks and Limitations on Observability in Closed Systems

Measurements are ordinarily described with respect to absolute "Newtonian" time. In reality however, the switching-on of the measuring device at the instance of the measurement requires a timing device. Hence the classical time $t$ must be replaced by a suitable quantum time variable $\tau$ of a physical clock. The main issue raised in this article is that while doing so, we can no longer neglect the {\em back-reaction} due to the measurement on the clock. This back-reaction yields a bound on the accuracy of the measurement. When this bound is violated the result of a measurement is generally not an eigenvalue of the observable, and furthermore, the state of the system after the measurement is generally not a pure state. We argue that as a consequence, a sub-class of observables in a closed system cannot be realized by a measurement.