We begin a systematic study of quantum energy inequalities (QEIs) in relation to local covariance. We define notions of locally covariant QEIs of both “absolute” and “difference” types and show that existing QEIs satisfy these conditions. Local covariance permits us to place constraints on the renormalized stress-energy tensor in one spacetime using QEIs derived in another, in subregions where the two spacetimes are isometric. This is of particular utility where one of the two spacetimes exhibits a high degree of symmetry and the QEIs are available in simple closed form. Various general applications are presented, including a priori constraints (depending only on geometric quantities) on the ground-state energy density in a static spacetime containing locally Minkowskian regions. In addition, we present a number of concrete calculations in both two and four dimensions that demonstrate the consistency of our bounds with various known ground- and thermal-state energy densities. Examples considered include the Rindler and Misner spacetimes, and spacetimes with toroidal spatial sections. In this paper we confine the discussion to globally hyperbolic spacetimes; subsequent papers will also discuss spacetimes with boundary and other related issues.
REFERENCES
By a stationary state, we mean one whose -point functions are invariant under translations along the Killing flow: , where is the group of isometries associated with the Killing field.
The proof employed in Refs. 18 and 16 proceeds by defining a function with , and therefore only applies in the first instance to the case where has connected support and no zeros in the interior thereof. We extend the result to more general by choosing a nonnegative with no zeros in the interior of its support, assumed to be connected, and which is equal to unity on the support of . Applying Flanagan’s result to , we may take the limit to obtain Eq. (25) (cf. Cor. A.2 in Ref. 20, where the notation is used).
We have , and want to show that for all . Differentiating and squaring yields from which it follows that except perhaps at zeros of . If vanishes in a neighborhood of then so must and the result holds trivially. For the remaining case, choose a sequence with ; since , we conclude the required result by continuity as .
To see this, we note that and that the maximum of this expression on occurs at the (unique) solution to , which is numerically . Now , so we find that on [) as claimed.
An operator is symmetric on a domain if for all , which shows that the adjoint agrees with on , but does not exclude the possibility that has a strictly larger domain of definition than .
The contour involved is the rectangle with “long” sides given by the interval of the real axis, and its translate . The contour encloses a single pole, of fourth order, at and the contribution of the “short” sides vanishes as . One also exploits the fact that the contributions from the two “long” sides are equal up to a factor of .
The Friedrichs extension has a domain contained in the closure of in the norm , which is equivalent to the norm of the Sobolev space since is bounded and bounded away from zero on . Accordingly, the closure of is precisely and the desired domain lies in this Sobolev space, all elements of which obey on .
For a scalar field with arbitrary curvature coupling constant , replaced the numerical coefficient in the stress tensor with .
To generalize the result to a scalar field with arbitrary curvature coupling constant , replace the 11 in the numerator with .