= Equal-time two-point function of a de Sitter scalar
{title2=$P_{\delta\phi}(k,\tau)=H^2(1+k^2\tau^2)/(2k^3)$}
For an exactly massless <minimally coupled scalar field> on flat-slicing <de Sitter spacetime> with constant $H$, use <natural units> and neglect its metric backreaction. With $a=-1/(H\tau)$ and the <Bunch-Davies vacuum>, its canonical mode is
$$
u_k(\tau)=\frac{e^{-ik\tau}}{\sqrt{2k}}\left(1-\frac{i}{k\tau}\right),\qquad
\delta\phi_k=u_k/a.
$$
With <power spectrum> convention $\langle\delta\phi_{\mathbf k}\delta\phi_{\mathbf k'}\rangle=(2\pi)^3\delta^3(\mathbf k+\mathbf k')P_{\delta\phi}(k,\tau)$, taking the squared mode amplitude gives
$$
P_{\delta\phi}(k,\tau)=\frac{H^2}{2k^3}(1+k^2\tau^2),\qquad
\Delta_{\delta\phi}^2=\frac{H^2}{4\pi^2}(1+k^2\tau^2).
$$
The constant part is the <frozen contribution to a de Sitter scalar spectrum>; the remaining part is the <subhorizon vacuum contribution to a de Sitter scalar spectrum>. A merely light but nonzero <mass> changes the exact mode equation, so the displayed massless solution is then an approximation rather than an exact identity. Nonminimal curvature coupling also changes it.
The massless infinite-volume <two-point correlation function> has an infrared divergence. A finite-volume or finite-duration infrared prescription is needed, and coincident products require ultraviolet regulation or smearing. A finite spectrum at each fixed nonzero $k$ does not by itself make the all-mode equal-position <integral> finite.
Back to article page