Magnetic pulsations are one of the most important phenomena in magnetospheric physics. Since Dungey  envisaged that the long but regular periods of magnetic field oscillations (continuous pulsations, or Pc) might be the result of standing waves being excited on geomagnetic field lines, it is generally believed that most Pc3-5 waves are field line resonances [Chen and Hasegawa, 1974; Southwood, 1974]. Wave activity in the foreshock region has been found to be a major energy source for Pc3-4 waves [Troitskaya et al., 1971]. Such a source provides a broad frequency band of energy which can excite multiple harmonics of field lines in the magnetosphere [Takahashi et al., 1984]. However, after many years of pulsation research, what we still eagerly seek from observations is a clear picture of how the pulsation energy propagates in the magnetosphere, since knowledge of this energy path should provide important constraints on and implications for the generation of the wave and the subsequent dynamics of the magnetosphere they cause.
In the following, we first discuss the phenomenon of ``phase skipping'' in pulsation signals that, as we shall see later, is related to the energy pathway in the magnetosphere. Historically, this phenomenon provided the first evidence for the dynamics of Pc3-4 paths. It will also lead naturally into the discussion of the physical process of importance, namely, the Poynting flux.
Phase skips are sudden changes in the phase of wave signals, while the frequency remains nearly constant. They are a common feature of the continuous pulsations in the magnetosphere. Herron  was perhaps the first to report that continuous pulsations consist of a succession of groups of oscillations, whose frequency remains relatively constant from group to group, but whose phase shifts from zero to almost one cycle from one group to the next.
Despite its long history, the topic of phase skipping is not often treated in the pulsation literature. One of the reasons for its omission is probably that such behavior is not well suited for analysis by conventional Fourier techniques. In addition, for the calculation of phase shifts, simplified models of wave signals that might not be realistic for pulsations are usually assumed a priori to avoid complicated fitting procedures of phase estimation.
An impulsive source of the wave energy and the phenomenon of wave beating are the two major reasons suggested in early work on the phase skipping phenomenon. Mier-Jedrzejowicz and Hughes  studied the pulsations observed simultaneously by multiple spacecraft, and also some events recorded by an array of ground magnetometers. They concluded that the impulsive nature of the energy source is the most likely reason for phase skipping. Lanzerotti et al.  discussed phase shifts when studying the polarization of pulsations at low latitudes. They found that the phase shifts seldom occur simultaneously at all latitudes and therefore concluded that the impulsive process suggested by Mier-Jedrzejowicz and Hughes  is not likely responsible for their observations. Later in this study we will discuss a possible reason that leads to these seemingly contradictory results.
For other contributions to this topic, Ansari and Fraser  studied Pc3 pulsations observed by low-latitude stations but they could not find any obvious pattern of phase jump occurrences between stations. Ostwald et al.  concluded that a typical pulsation signal exhibits packet structure and phase skipping mainly due to the superposition of interfering waves, although some phase skips that were observed simultaneously at spaced stations are most likely associated with global magnetospheric impulses. McDiarmid and Ziesolleck  devised a model of two overlapping responses that can be either two beating waves or a sequence of impulses. They applied their model to a real pulsation event and found that the result was consistent with the assumption of multiple impulsive excitations.
In addition to the impulsive source and wave beating, the rapid movement of a resonant region relative to the point of observation has also been postulated as a possible cause of phase skipping. However, it was found that the motion of the resonant region would be unrealistically fast [Mier-Jedrzejowicz and Hughes, 1980].
The phase skipping phenomenon is closely related to the Poynting flux of pulsations. Let and be the electric waves and magnetic waves, respectively. If we consider a simple condition that , not only does the Poynting vector depend on the amplitudes of and , it is also a function of the relative phase between and .When an impulse of new wave energy comes into the system, if it has a different relative phase or a different propagation direction from that of the preexisting wave, the phase must change in some components of or .Therefore it is critical to examine jointly these two properties of pulsations.
For the observation of the Poynting flux of pulsations, Cummings et al.  studied one Pc4 pulsation event observed by the ATS 6 spacecraft and found that the drift-aligned portion of the Poynting vector suggested a westward propagating wave. Junginger et al.  performed a statistical study of the Poynting vectors for the Pc5 pulsations observed by the geosynchronous satellite GEOS 2. Although the behavior of the Poynting vectors for the 150- to 300-s waves was not clear, the perpendicular component of the Poynting vectors associated with 400- to 600-s waves was directed inward and toward the local noon, which is consistent with the simulation study by Junginger  for the standing shear Alfvén waves driven by the surface waves on the outer boundaries of the magnetosphere. Cahill et al.  studied the azimuthal magnetic oscillations and their corresponding electric waves in the DE 1 data. By comparing the E and B amplitudes and the phase relations between E and B waves, they concluded that their events are consistent with the standing wave model. Takahashi et al.  reported a compressional Pc3 event observed by the Geotail satellite when it was located in the prenoon sector of the dayside magnetosphere. The calculation of the Poynting flux indicated that the wave energy flowed earthward, which supports the upstream-wave generation model for Pc3 pulsations [Troitskaya et al., 1971]. In this study, we will also discuss the propagation and possible generation mechanisms of Pc3-4 pulsations implied by the Poynting flux observations.