
to quantum gravity suggest that it is more natural, or may even be necessary [1–4] to have
more than one inflaton. This has generated renewed interest in multi-field cosmological
models, which had previously attracted only limited attention.
Multifield cosmological models have richer phenomenology than single field models
since they allow for solutions of the equations of motion whose field-space trajectories are
not (reparameterized) geodesics. Such trajectories are characterized by a non-zero turn
rate. In the past it was thought that phenomenological viability requires small turn rate,
by analogy with the slow roll approximation used in the single-field case. This assumption
leads to the celebrated slow-roll slow-turn (SRST) approximation of [5, 6]. However, in
recent years it was understood that rapid turn trajectories can also be (linearly) pertur-
batively stable [7, 8] and of phenomenological interest. For instance, a brief rapid turn
during slow-roll inflation can induce primordial black hole generation [9–12]; moreover,
trajectories with large and constant turn rate can correspond to solutions behaving as dark
energy [13, 14]. There is also a variety of proposals for full-fledged rapid-turn inflation
models, relying on large turn rates during the entire inflationary period [15–22].
Finding inflationary solutions in multifield models is much harder than in the single-
field case, because the background field equations form a complicated coupled system
of nonlinear ODEs. Thus usually such models are either studied numerically or solved
only approximately.3Mathematically, this complicated coupled system is encoded by
the so-called cosmological equation, a nonlinear second order geometric ODE defined on
the scalar field space of the model. The latter is a connected paracompact manifold,
usually called the scalar manifold. In turn, the cosmological equation is equivalent with a
dissipative geometric dynamical system defined on the tangent bundle of that manifold.
Little is known in general about this dynamical system, in particular because the scalar
manifold need not be simply-connected and – more importantly – because this manifold is
non-compact in most applications of physical interest and hence cosmological trajectories
can “escape to infinity”. The resulting dynamics can be surprisingly involved4and hard to
3A notable exception is provided by models with hidden symmetry, which greatly facilitates the search
for exact solutions [23–27].
4It is sometimes claimed that the complexity of this dynamics could be ignored, because in “phe-
nomenologically relevant models” one should “expect that” all directions orthogonal to the physically
relevant scalar field trajectory are heavy and hence can be integrated out, thus reducing the analysis
to that of a single-field model. This argument is incorrect for a number of reasons. First, current phe-
nomenological data does not rule out multifield dynamics. Second, such a reduction to a one field model
(even when possible) relies on knowledge of an appropriate cosmological trajectory, which itself must first
be found by analyzing the dynamics of the multifield model.
2