1: \begin{abstract}

2: 

3: We investigate the possible accuracy that can be reached by analytical

4: models for the matter density power spectrum and correlation function.

5: Using a realistic description of

6: the power spectrum that combines perturbation theory with a halo model,

7: we study the convergence rate of several perturbative expansion schemes

8: and the impact of nonperturbative effects, as well as the sensitivity to

9: phenomenological halo parameters.

10: We check that the simple reorganization

11: of the standard perturbative expansion, with a Gaussian damping prefactor,

12: provides a well-ordered convergence and a finite correlation function that yields

13: a percent accuracy at the baryon acoustic oscillation peak (as soon as one goes to second order). 

14: Lagrangian-space expansions

15: are somewhat more efficient, when truncated at low orders, but may diverge at

16: high orders. We find that whereas the uncertainty on the halo-profile

17: mass-concentration relation is not a strong limitation, the uncertainty on the 

18: halo mass function can severely limit the accuracy of theoretical predictions for

19: $P(k)$ (this also applies to the power spectra measured in numerical simulations).

20: The real-space correlation function provides a better separation between

21: perturbative and nonperturbative effects, which are restricted to

22: $x \lesssim 10 h^{-1}$Mpc at all redshifts.

23: 

24: 

25: 

26: 

27: \keywords{Cosmology \and large scale structure of the Universe}

28: \end{abstract}

29: