Testing growth rate dependence in cosmological perturbation theory using scale-free models

We generalize previously derived analytic results for the one-loop power spectrum (PS) in scale-free models (with linear PS $P(k) \propto k^n$) to a broader class of such models in which part of the matterlike component driving the Einstein de Sitter expansion does not cluster. These models can be conveniently parametrized by $\alpha$, the constant logarithmic linear growth rate of fluctuations (with $\alpha=1$ in the usual case). For $-3< n<-1$, where the one-loop PS is both infrared and ultraviolet convergent and thus explicitly self-similar, it is characterized conveniently by a single numerical coefficient $c(n, \alpha)$. We compare the analytical predictions for $c(n=-2, \alpha)$ with results from a suite of $N$-body simulations with $\alpha \in [0.25, 1]$ performed with an appropriately modified version of the GADGET code. Although the simulations are of small ($256^3$) boxes, the constraint of self-similarity allows the identification of the converged PS at a level of accuracy sufficient to test the analytical predictions for the $\alpha$ dependence of the evolved PS. Good agreement for the predicted dependence on $\alpha$ of the PS is found. To treat the UV sensitivity of results which grows as one approaches $n =-1$, we derive exact results incorporating a regularization $k_c$ and obtain expressions for $c(n, \alpha, k_c/k)$. Assuming that this regularization is compatible with self-similarity allows us to infer a predicted functional form of the PS equivalent to that derived in effective field theory (EFT). The coefficient of the leading EFT correction at one loop has a strong dependence on $\alpha$, with a change in sign at $\alpha \approx 0.16$, providing a potentially stringent test of EFT.Comment: revised version published in Phys. Rev.