We consider next-to-leading-order (one-loop) nonlinear corrections to
the bispectrum and skewness of cosmological density fluctuations induc
ed by gravitation evolution, focusing on the case of Gaussian initial
conditions and scale-free initial power spectra, P(k) proportional to
k ''. As has been established by comparison with numerical simulations
, leading order (tree-level) perturbation theory describes these quant
ities at the largest scales. The one-loop perturbation theory provides
a tool to probe the transition to the nonlinear regime on smaller sca
les. In this work, we find that, as a function of spectral index n, th
e one-loop bispectrum follows a pattern analogous to that of the one-l
oop power spectrum, which shows a change in behavior at a ''critical i
ndex'' n(c) approximate to -1.4, where nonlinear corrections vanish. T
he tree-level perturbation theory predicts a characteristic dependence
of the bispectrum on the shape of the triangle defined by its argumen
ts. For n less than or similar to n(c), one-loop corrections increase
this configuration dependence of the leading order contribution; for n
greater than or similar to n(c), one-loop corrections tend to cancel
the configuration dependence of the tree-level bispectrum, in agreemen
t with known results from n = -1 numerical simulations. A similar situ
ation is shown to hold for the Zeldovich approximation, where n(c) app
roximate to -1.75. We obtain explicit analytic expressions for the one
-loop bispectrum for n = -2 initial power spectra, for both the exact
dynamics of gravitational instability and the Zeldovich approximation.
We also compute the skewness factor, including local averaging of the
density field, for n = -2: S-3(R) = 4.02 + 3.83 sigma(G)(2)(R) for Ga
ussian smoothing and S-3(R) = 3.86 + 3.18 sigma(TH)(2)(R) for top-hat
smoothing, where sigma(2)(R) is the variance of the density field fluc
tuations smoothed over a window of radius R. A comparison with fully n
onlinear numerical simulations implies that, for n < -1, the one-loop
perturbation theory can extend our understanding of nonlinear clusteri
ng down to scales where the transition to the stable clustering regime
begins.