Non-Gaussianity in the Curvaton Scenario

Tomo Takahashi1 Department of Physics, Saga University, Saga 840-8502, Japan

Abstract
We discuss the signatures of non-Gaussianity in the curvaton model where the potential includes also a non-quadratic term. In such a case the non-linearity parameter fNL can become very small, and we show that non-Gaussianity is then encoded in the non-reducible non-linearity parameter gNL of the trispectrum, which can be very large. Thus the place to look for the non-Gaussianity in the curvaton model may be the trispectrum rather than the bispectrum. We also show that gNL measures directly the deviation of the curvaton potential from the purely quadratic form.

Introduction

The high precision of cosmological observations such as WMAP [1, 2] and the forthcoming Planck Surveyor Mission [3] will soon make it possible to probe the actual physics of the primordial perturbation rather than merely describing it. As is well known, the primordial scalar and tensor power spectra, characterized by spectral indices the and tensor-to-scalar ratio, can be used to test both models of ination and other mechanisms for the generation of the primordial perturbation such as the curvaton [4, 5, 6]. However, many models can imprint similar features on the primordial power spectrum. Thus a possible deviation from Gaussian uctuations, which may provide invaluable implications on the physics of the early universe, has been the focus of much attention recently. The simplest ination models generate an almost Gaussian uctuation. In contrast, in the curvaton scenario there can arise a large non-Gaussianity. However, in most studies on the curvaton so far, one simply assumes a quadratic curvaton potential. Since the curvaton cannot be completely non-interacting (it has to decay), it is of interest to consider the implications of the deviations from the exactly quadratic potential, which represent curvaton self-interactions. Such self-interactions would arise e.g. in curvaton models based on the at directions of the minimally supersymmetric standard model (MSSM). Even small deviations could be important for phenomenology, as was pointed out in [7] where it was shown that the non-Gaussianity predicted by the curvaton model can be sensitive to the shape of the potential. In particular, the nonlinearity parameter fNL which quanties the bispectrum of primordial uctuation can, in contrast to the case of the quadratic curvaton potential, be very small in some cases. However, the signatures of non-Gaussianity can be probed not only with the bispectrum but also with the trispectrum. In the following, we will show that even when fNL is negligibly small, gNL can be very large, indicating that the rst signature of the curvaton-induced primordial non-Gaussianity may not come from the bispectrum but rather from the trispectrum.

Non-linearity parameters

To discuss non-Gaussianity in the scenario, we make use of the non-linearity parameters fNL and gNL dened by the expansion 3 9 2 3 4 = 1 + fNL 1 + gNL 1 + O(1 ), (1) 5 25 where is the primordial curvature perturbation. Writing the power spectrum as
3 1 2 = (2) P (k1 )( 1 + 2 ), k k k k

(2)

1 E-mail:tomot@cc.saga-u.ac.jp

the bispectrum and trispectrum are given by 1 2 3 k k k 1 2 3 4 k k k k =

3 = (2) B (k1 , k2 , k3 )( 1 + 2 + 3 ). k k k

(3) (4)

3 (2) T (k1 , k2 , k3 , k4 )( 1 + 2 + 3 + 4 ), k k k k

where B and T are products of the power spectra and can be written as B (k1 , k2 , k3 ) T (k1 , k2 , k3 , k4 ) 6 fNL (P (k1 )P (k2 ) + P (k2 )P (k3 ) + P (k3 )P (k1 )) , 5 = NL (P (k13 )P (k3 )P (k4 ) + 11 perms.) 54 + gNL (P (k2 )P (k3 )P (k4 ) + 3 perms.) . 25 = (5)

(6)

Note the appearance of the independent non-linearity parameter NL . However, in the scenario we are considering in the following, NL is related to fNL by NL = 36 2 f . 25 NL (7)

To evaluate the primordial curvature uctuation in the curvaton model, we need to specify a potential for the curvaton. Here we go beyond the usual quadratic approximation and consider the following potential: n 1 V () = m2 2 + m4 , (8) 2 m which contains a higher polynomial term in addition to the quadratic term. For later discussion, we dene a parameter s which represents the size of the non-quadratic term relative to the quadratic one: s 2 m
n2

(9)

Thus the larger s is, the larger is the contribution from the non-quadratic term. When the potential is quadratic, the uctuation evolves exactly as the homogeneous mode. However, when the curvaton eld evolves under a non-quadratic potential, the uctuation of the curvaton evolves non-linearly on large scales. In that case the curvature uctuation can be written, up to the third order, as [8] = N =
osc 2 osc 1 osc osc 4r2 2r3 r + 3r 1 + 2 3 osc 9 osc osc 2 4 9r osc osc osc osc osc osc + +3 9r2 1 + 3 2 2 81 4 osc osc osc 2

( )2

r3 2

19

osc osc 2 osc

+ 10r4 + 3r5

osc osc

( )3 ,

(10)

where osc is the value of the curvaton at the onset of its oscillation and the prime represents the derivative with respect to . r roughly represents the ratio of the energy density of the curvaton to the total density at the time of the curvaton decay. The exact denition is given by r 3 4rad + 3 .
decay

(11)

Notice that osc /osc = 1/ for the case of the quadratic potential. With this expression, we can write down the non-linearity parameter fNL as

fNL =

5 4r

1+

osc osc 2 osc

5 5r . 3 6

(12)

200 100

0 -10000 -20000

0 -100

-30000 -40000 gNL -50000 -60000 -70000 -80000

s = 0.2 s = 0.1 s = 0.05 2 3 4 5 n 6 7 8

fNL -200 -300 -400 -500

-90000 -100000 -40 -20 0 20 fNL 40

s = 0.2 s = 0.1 s = 0.05 60 80 100

Figure 1: (Left) Plot of fNL as a function of n for several values of s. (Right) Plot of gNL as a function of fNL for several values of s. Notice that fNL and n have one-to-one correspondence. In both panels, r = 0.01.

Also notice that, although the curvaton scenario generally generates large non-Gaussianity with fNL > O(1), the non-linearity parameter fNL can be very small in the presence of the non-linear evolution of 2 the curvaton eld which can render the term 1 + (osc osc )/osc 0 [7, 8]. In the curvaton model, the non-linearity parameter gNL can be written as gNL = 25 9 54 4r2
2 osc osc osc osc +3 3 2 osc osc

9 r

1+

osc osc 2 osc

1 2

19

osc osc 2 osc

+ 10r + 3r2 . (13)

As one can easily see, even if the non-linear evolution of cancels to give a very small fNL , such a cancellation does not necessarily occur for gNL . This indicates an interesting possibility where the nonGaussian signature of the curvaton may come from the trispectrum rather than from the bispectrum.

Signatures of non-Gaussianity

Let us now consider non-Gaussianity in curvaton models, paying particular attention to the non-linearity parameters fNL and gNL . We show the values of fNL and gNL as a function of the power n in Fig. 1. In the left panel of the gure, the value of fNL is plotted as a function of the power n for several values of s. There we have xed the value of r to r = 0.01. As can be read o from Eq. (12), when r is small, fNL can be well approximated as 5 osc osc fNL 1+ . (14) 2 4r osc
2 Notice that the combination osc osc /osc is zero when n = 2. As the value of n becomes larger, the above combination yields a negative contribution. Thus, fNL decreases to zero as the potential deviates away from a quadratic form and then becomes zero for some values of n and s. For small values of s, which correspond to the cases where the size of the non-quadratic term is relatively small compared to the quadratic one, the power n should be large to make fNL very small. It should also be mentioned that, for a xed s, if we take larger values of n beyond the cancellation point fNL = 0, fNL becomes negative. However, as already discussed, even if we obtain very small values for fNL , it does not necessarily indicate that non-Gaussianity is small in the model but may show up in the higher order statistics. Indeed, this appears to be a generic feature of the curvaton model: the trispectrum cannot be suppressed and gNL can be quite large and is always negative for small values of r. Interestingly, even if fNL is zero, the value of |gNL | can be very large, as can be seen Fig. 1. This is because a cancellation which can occur for fNL does not take place for gNL , which is a smooth function of n. When r is small, gNL is mainly

determined by the rst term in Eq. (13): gNL 25 9 54 4r2

2 osc osc osc osc +3 3 2 osc osc

(15)

2 2 3 In fact, one can show that the terms (osc osc )/osc and (osc osc )/osc give negative contributions for the cases being considered here, which indicates that the cancellation between these terms never occurs [9]. Therefore gNL is always negative for small values of r even when fNL is very small. Thus, it may turn out that the best place to look for non-Gaussianity in curvaton models is the trispectrum and gNL in particular.

Summary

We have discussed the signatures of non-Gaussianity in the curvaton model with the potential including a non-quadratic term in addition to the usual quadratic term. When the curvaton potential is not purely quadratic, uctuations of the curvaton eld evolve nonlinearly on superhorizon scales. This gives rise to predictions for the bispectrum, characterized by the non-linearity parameter fNL , which can deviate considerably from the quadratic case. However, by studying the trispectrum which is characterized by gNL along with fNL , we nd that even when fNL is negligibly small, the absolute value of gNL can be very large. Thus the signature of non-Gaussianity in the curvaton model may come from the trispectrum rather than from the bispectrum if its potential deviates from a purely quadratic form, as one would expect in realistic particle physics models. Here we discussed the case where a non-quadratic term is also taken into account in the curvaton potential, however the cases where uctuations from the inaton can also be responsible for cosmic density perturbations today would also be interesting since such a case can arise in general. Such mixed uctuation scenarios have been investigated in Refs. [10, 11, 12].

References
[1] E. Komatsu et al. [WMAP Collaboration], arXiv:0803.0547 [astro-ph]. [2] J. Dunkley et al. [WMAP Collaboration], arXiv:0803.0586 [astro-ph]. [3] [Planck Collaboration], arXiv:astro-ph/0604069; Planck webpage, http://www.rssd.esa.int/index.php?project=planck [4] K. Enqvist and M. S. Sloth, Nucl. Phys. B 626, 395 (2002) [arXiv:hep-ph/0109214]; [5] D. H. Lyth and D. Wands, Phys. Lett. B 524, 5 (2002) [arXiv:hep-ph/0110002]; [6] T. Moroi and T. Takahashi, Phys. Lett. B 522, 215 (2001) [Erratum-ibid. B 539, 303 (2002)] [arXiv:hep-ph/0110096]. [7] K. Enqvist and S. Nurmi, JCAP 0510, 013 (2005) [arXiv:astro-ph/0508573]. [8] M. Sasaki, J. Valiviita and D. Wands, Phys. Rev. D 74, 103003 (2006) [arXiv:astro-ph/0607627]. [9] K. Enqvist and T. Takahashi, JCAP 0809, 012 (2008) [arXiv:0807.3069 [astro-ph]]. [10] K. Ichikawa, T. Suyama, T. Takahashi and M. Yamaguchi, Phys. Rev. D 78, 023513 (2008) [arXiv:0802.4138 [astro-ph]]. [11] K. Ichikawa, T. Suyama, T. Takahashi and M. Yamaguchi, Phys. Rev. D 78, 063545 (2008) [arXiv:0807.3988 [astro-ph]]. [12] T. Moroi and T. Takahashi, arXiv:0810.0189 [hep-ph].