利用重正化群的数学类比，发展了一种描述由两个膨胀场构成的膨胀的微扰策略。定义了两个小量$\alpha$和$\lambda$，它们对应于标准的慢滚参数，并且找到了所有膨胀量以$\alpha$和$\lambda$幂次展开的系统方法，无需其他慢滚参数。为了在多场背景下说明这种微扰方法，我们采用了一个简单的具有二次势的两场模型，在两个场对膨胀动力学贡献相似的参数范围内。尽管该模型并不是描述膨胀时期的可行替代方案，但它很好地展示了微扰方法的细微之处。特别是，这种方法允许我们推导出两种独立的规范不变标量扰动，在超视界极限下保持守恒，克服了多场膨胀中常见的问题。此外，还可以进行非微扰重整，以研究模型的真实多场区域。我们分别得到了张量和标量功率谱的次次领头阶和领头阶结果，以及它们的谱指数。两个标量扰动之间的层级关系使我们可以确定标量功率谱矩阵的主要项。由于第二个膨胀场的作用，修正已经在领头阶出现。最后，我们解释了为什么即使标量扰动确实出现了非平凡修正，二次两场模型仍然与当前实验限制不兼容。

A perturbative strategy for inflation described by two-inflaton fields is developed using a mathematical analogy with the renormalization-group. Two small quantities, $\alpha$ and $\lambda$, corresponding to standard slow-roll parameters are defined and systematic expansions of all inflationary quantities in terms of powers of $\alpha$ and $\lambda$ are found. No other slow-roll parameters are needed. To illustrate this perturbative method in the multi-field context, we adopt a simple two-inflaton model with quadratic potentials in the parameter range where both fields contribute similarly to the dynamics of inflation. The model, even though it is not a viable alternative for phenomenological description of the inflationary period, nicely illustrates subtleties of the perturbative approach. In particular, this method allows us to derive two independent gauge-invariant scalar perturbations that are conserved in the superhorizon limit, overcoming typical problems that emerge in multi-field inflation. Furthermore, it is possible to perform nonperturbative resummations that allow to study the model in a true multi-field regime. We derive tensor and scalar power spectra to the next-to-next-to-leading and next-to-leading orders, respectively, as well as their spectral indices. The hierarchy between the two scalar perturbations allows us to single out the dominant entry of the scalar power-spectrum matrix. Modifications due to the second inflaton occur already at the leading order. Finally, we explain why the quadratic two-inflaton model is not compatible with the present experimental constraints even though non-trivial corrections to scalar perturbations do emerge.