We construct the graviton propagator on de Sitter background in the one parameter family of exact, de Sitter invariant gauges. Our result takes the form of a universal spin two part and a gauge dependent spin zero part. Scalar equations are derived for the structure functions of each part. There is no de Sitter invariant solution for either structure function, although the de Sitter breaking contribution to the spin zero part may drop out for certain choices of the gauge parameter. Our results imply that de Sitter breaking is universal for the graviton propagator, and hence that there is an error in the contrary results derived by analytic continuation of average gauge fixing techniques.

A nice paper by Morrison [1] demonstrates the recent convergence of opinion that has taken place concerning the graviton propagator on de Sitter background. We here discuss the few points which remain under dispute. First, the inevitable decay of tachyonic scalars really does result in their 2point functions breaking de Sitter invariance. This is obscured by analytic continuation techniques which produce formal solutions to the propagator equation that are not propagators. Second, Morrison's de Sitter invariant solution for the spin two sector of the graviton propagator involves derivatives of the scalar propagator at M 2 = 0, where it is not meromorphic unless de Sitter breaking is permitted. Third, de Sitter breaking does not require zero modes. Fourth, the ambiguity Morrison claims in the equation for the spin two structure function is fixed by requiring it to derive from a mode sum. Fifth, Morrison's spin two sector is not "physically equivalent" to ours because their coincidence limits differ. Finally, it is only the noninvariant propagator that gets the time independence and scale invariance of the tensor power spectrum correctly.1 The same de Sitter breaking occurs using the different field variables favored by Kitamoto and Kitazawa [15].2 Ignoring this problem in scalar quantum electrodynamics leads to on-shell singularities in the scalar self-mass-squared [23].3 The mathematical physics computation [29] had a number of significant errors that were discovered by comparison with the cosmological result [19] and then corrected [30].

We use the Hartree approximation to the Einstein equation on de Sitter background to solve for the one loop correction to the graviton mode function. This should give a reasonable approximation to how the ensemble of inflationary gravitons affects a single external graviton. At late times we find that the one loop correction to the plane wave mode function u(η, k) goes like GH 2 ln(a)/a 2 , where a is the inflationary scale factor. One consequence is that the one loop corrections to the "electric" components of the linearized Weyl tensor grow compared to the tree order result.

We compute the linearized Weyl-Weyl correlator using a new solution for the graviton propagator on de Sitter background in de Donder gauge. The result agrees exactly with a previous computation in a noncovariant gauge. We also use dimensional regularization to compute the one loop expectation value of the square of the Weyl tensor.

We use a de Sitter breaking graviton propagator [1,2] to compute the tree order correlator between noncoincident Weyl tensors on a locally de Sitter background. An explicit, and very simple result is obtained, for any spacetime dimension D, in terms of a de Sitter invariant length function and the tensor basis constructed from the metric and derivatives of this length function. Our answer does not agree with the one derived previously by Kouris [3], but that result must be incorrect because it not transverse and lacks some of the algebraic symmetries of the Weyl tensor. Taking the coincidence limit of our result (with dimensional regularization) and contracting the indices gives the expectation value of the square of the Weyl tensor at lowest order. We propose the next order computation of this as a true test of de Sitter invariance in quantum gravity.