Coronal mass ejections (CMEs) are huge, episodic eruptions of solar plasma and magnetic field which travel through the solar corona and out into the heliosphere. At Earth, they drive the most severe geomagnetic storms and thus are the primary focus of space-weather forecasting.
Using white-light imagers, individual CMEs can be tracked continuously from the low corona, through the solar wind, all the way to Earth  and beyond. Such observations show CMEs apparently bouncing off each other  and deflecting off other coronal and solar wind structures . Thus it is tempting to think of a CME as a coherent structure; a single – perhaps even quasi-solid — body, playing out a game of solar billiards.
Such structural coherence has two physical requirements. Firstly, coherence requires a restoring force which can (at least partially) resist deformation by external factors. E.g., A bubble is a coherent body as surface tension communicates external forces across the entirety of the structure, allowing it to respond as a single entity. A dust cloud, on the other hand, has negligible inter-speck forces, so does not does not behave in a coherent manner; nudge one speck and the others do not respond. Magnetic pressure and curvature forces within a CME can provide the restoring force required for coherence. But there is also the second requirement; that information be able to propagate across the structure.
CME expansion and propagation
Movie 1 – A coronal mass ejection starts life with a circular cross-section. It both expands and propagates radially away from the Sun, resulting in flattening of the cross section (or “pancaking”). This evolution causes the magnetic field intensity and plasma density in the CME to drop with distance from the Sun, lowering the Alfven speed. Two points on the CME leading edge, A and B, separate due to simple geometry. Past 0.65 AU, the speed of separation of A and B exceeds the local Alfven speed and thus information can no longer propagate between these two parts of the CME.
Close to the Sun, CMEs undergo rapid evolution and non-radial motion. Past a few solar radii, however, CMEs are observed to propagate nearly radially while also expanding. Thus even if a CME begins life with a circular cross-section, as shown at the start of Movie 1 above, it will quickly flatten in the non-radial direction, or “pancake” . Using a simple model for the change in the CME cross-sectional area with distance from the Sun, the magnetic field intensity and plasma density within a CME can be estimated by assuming constant magnetic flux and constant mass within the CME, respectively. The results are shown in the right-hand panels of the movie. With heliocentric distance, R, magnetic field intensity falls off as approximately R-2, while plasma density is closer to R-3. Thus the Alfven speed within a CME falls off as approximately R-½. This provides a reasonable approximation of the maximum information propagation speed within the CME.
CME expansion and propagation mean that parts of the CME are moving apart. As a simple consequence of spherical geometry, points on a CME front, A and B, move apart at a near-constant speed, VAB. When VAB exceeds the local Alfven speed, information is unable to propagate between A and B, and the CME ceases to be a coherent structure over such length scales. For the typical CME parameters shown in Movie 1, and for an angular separation of A and B with respect to the Sun of just 15°, this occurs at approximately 0.65 AU. Greater angular separation of A and B results in greater VAB and hence loss of coherence starts closer to the Sun.
Movie 2 – Left: A CME propagates from the Sun to the Earth, with Alfven wave fronts regularly released from a point on the flank (point B in Movie 1). Right: The view from the frame of reference of the CME centre (the black cross). Due to declining Alfven speed and increasing CME size, the wave fronts have increasingly limited ability to propagate across the CME as it moves further away from the Sun.
Movie 2 shows how spherical wave fronts initiated at point B on the CME front have increasingly limited reach within the CME structure with increasing distance from the Sun.
What does this mean for CME forecasting?
A recent study  shows that all observed CMEs have likely lost coherence over the half their angular extent (i.e., the East flank is effectively isolated from the West flank, etc) within 0.3 AU. No CMEs are expected to remain fully coherent structures by the time they reach 1 AU. Thus many of the current techniques used to track and forecast CMEs in the heliosphere may need revisiting in order to take account of this fragmentation. Ambient solar wind structure may have far more influence on the structure of CMEs than is currently assumed.
 Davis, C. J., Davies, J. A., Lockwood, M., Rouillard, A. P., Eyles, C. J. and Harrison, R. A. Stereoscopic imaging of an Earth-impacting solar coronal mass ejection: a major milestone for the STEREO mission. Geophys. Res. Lett., 36. L08102, doi: 10.1029/2009GL038021 (2009)
 Shen, C. et al. Super-elastic collision of large-scale magnetized plasmoids in the heliosphere. Nature Physics 8, 923-928, doi:10.1038/nphys2440 (2012)
 Byrne, J. P., Maloney, S. A., McAteer, R. T. J., Refojo, J. M. & Gallagher, P. T. Propagation of an Earth-directed coronal mass ejection in three dimensions. Nature Communications 1, doi:10.1038/ncomms1077 (2010)
 Owens, M. J., Merkin, V. G. & Riley, P. A kinematically distorted flux rope model for magnetic clouds. J. Geophys. Res. 111, A03104, doi:10.1029/2005JA011460 (2006)
 Owens, Lockwood & Barnard, Coronal mass ejections are not coherent magnetohydrodynamic structures, Scientific Reports, 7, 4152, doi:10.1038/s41598-017-04546-3 (2017)