132 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 20

and if e1,…,eM  are the roots of the equations

it follows from (4) that

   The right side of (7) vanishes only on the surface of an ellipsoid E, and is positive only in the interior of E. The surfaces of constant Q are concentric spheres. If S denotes a particular one of these spheres whose interior R contains the ellipsoid E, it is evident that each trajectory eventually becomes trapped within R.

3. The instability of nonperiodic flow

  In this section we shall establish one of the most important properties of deterministic nonperiodic flow, namely, its instability with respect to modifications of small amplitude. We shall find it convenient to do this by identifying the solutions of the governing equations with trajectories in phase space. We shall use such symbols as P(t) (variable argument) to denote trajec­tories, and such symbols as P or P(t) (no argument or constant argument) to denote points, the latter symbol denoting the specific point through which P(t) passes at time t0.
   We shall deal with a phase space Γ in which a unique trajectory passes through each point, and where the passage of time defines a continuous deformation of any region of Γ into another region, so that if the points P1(t0), P2(t0), … approach P0(t0) as a limit, the points P1(t0+τ), P2(t0+τ), must approach P0(t0+τ) as a limit. We shall furthermore require that the trajectories be uniformly bounded as t→ ∞; that is, there must be a bounded region R, such that every trajectory ultimately remains with R. Our procedure is influenced by the work of Birkhoff (1927) on dynamical systems, but differs in that Birkhoff was concerned mainly with con­servative systems. A rather detailed treatment of dynamical systems has been given by Nemytskii and Stepanov (1960), and rigorous proofs of some of the theorems which we shall present are to be found in that source.
   We shall first classify the trajectories in three different manners, namely, according to the absence or presence of transient properties, according to the stability or instability of the trajectories with respect to small modifications, and according to the presence or absence of periodic behavior. Since any trajectory P(t) is bounded, it must possess at least one limit point P0, a point which it approaches arbitrarily closely arbitrarily often. More precisely, P0 is a limit point of P(t) if for any ε > 0 and any time t1 there exists a time t2(ε,t1)>t1 such that |P(t2)-P0| < ε. Here

absolute-value signs denote distance in phase space. Because Γ is continuously deformed as t varies, every point on the trajectory through P0 is also a limit point of P(t), and the set of limit points of P(t) forms a tra­jectory, or a set of trajectories, called the limiting trajectories of P(t). A limiting trajectory is obviously con­tained within R in its entirety.
   If a trajectory is contained among its own limiting trajectories, it will be called central; otherwise it will be called noncentral. A central trajectory passes arbitrarily closely arbitrarily often to any point through which it has previously passed, and, in this sense at least, sepa­rate sufficiently long segments of a central trajectory are statistically similar. A noncentral trajectory remains a certain distance away from any point through which it has previously passed. It must approach its entire set of limit points asymptotically, although it need Dot approach any particular limiting trajectory asymptotically. Its instantaneous distance from its closest limit point is therefore a transient quantity, which becomes arbitrarily small as t→ ∞.
   A trajectory P(t) will be called stable at a point P(t1) if any other trajectory passing sufficiently dose to P(t1) at time t1 remains dose to P(t) as t→ ∞; i.e., P(t) is stable at P(t1) if for any ε > 0 there exists a δ(ε,t1) such that if |P1(t1) – P(t1)|<δ and t2>t1, |P1(t2) - P(t2)| < ε. Otherwise P(t) will be called unstable at P(t1). Because Γ is continuously deformed as t varies, a trajectory which is stable at one point is stable at every point, and will be called a stable trajectory. A trajectory unstable at one point is unstable at every point, and will be called an unstable trajectory. In the special case that P(t) is confined to one point, this definition of stability coincides with the familiar concept of stability of steady flow.
   A stable trajectory P(t) will be called uniformly stable if the distance within which a neighboring trajectory must approach a point P(t1), in order to be certain of remaining dose to P(t) as t→ ∞, itself possesses a positive lower bound as t1→ ∞; i.e., P(t) is uniformly stable if for any ε > 0 there exists a δ(ε) > 0 and a time t0(ε) such that if t1>t0 and |P1(t1) – P(t1)|<δ and t2>t1, |P1(t2) - P(t2)| < ε. A limiting trajectory P0(t) of a uniformly stable trajectory P(t) must be uniformly stable itself, since all trajectories passing sufficiently dose to Po(t) must pass arbitrarily dose to some point of P(t) and so must remain dose to P(t), and hence to P0(t), as t→ ∞.
   Since each point lies on a unique trajectory, any trajectory passing through a point through which it has previously passed must continue to repeat its past be­havior, and so must be periodic. A trajectory P(t) will be called quasi-periodic if for some arbitrarily large time interval τ, P(t+τ) ultimately remains arbitrarily close to P(t), i.e., P(t) is quasi-periodic if for any ε > 0 and for any time interval τ0, there exists a τ(ε,τ0) > τ0 and a time t1(ε,τ0) such that if t2>t1, |P1(t2+τ) - P(t2)|
 
     
1