A&A 435, 1149-1157 (2005)
DOI: 10.1051/0004-6361:20042166
B. Vrsnak - D. Sudar - D. Ruzdjak
Hvar Observatory, Faculty of Geodesy, Kaciceva 26, 10000 Zagreb, Croatia
Received 13 October 2004 / Accepted 7 February 2005
Abstract
We present a statistical analysis of 545 flare-associated CMEs and 104 non-flare CMEs
observed in the heliocentric distance range 2-30 solar radii. We found that both data sets show
quite similar characteristics, contradicting the concept of two distinct (flare/non-flare) types of
CMEs. In both samples there is a significant fraction of CMEs showing a considerable acceleration
or deceleration and both samples include a comparable ratio of fast and slow CMEs. We present
kinematical curves of several fast non-flare CMEs moving at a constant speed or decelerating, i.e.,
behaving as expected for flare-associated CMEs. Analogously, we identified several slow flare-CMEs
showing the acceleration peak beyond a height of 3 solar radii.
On the other hand, it is true that CMEs associated with major flares are on average faster
and broader than non-flare CMEs and small-flare CMEs. There is a well-defined correlation between
the CME speed and the importance of the associated flare. In this respect, the non-flare CMEs show
characteristics similar to CMEs associated with flares of soft X-ray class B and C, which is
indicative of a "continuum'' of events rather than supporting the existence of two distinct CME
classes. Furthermore, we inferred that CMEs whose source region cannot be identified with either
flares or eruptive prominences are on average slowest. The results indicate that the magnetic
reconnection taking place in the current sheet beneath the CME significantly influences the CME
dynamics.
Key words: Sun: coronal mass ejections (CMEs) - Sun: flares - Sun: prominences - Sun: filaments - magnetohydrodynamiques
A comparison of the kinematics of 16 CMEs associated with (large) flares and 11 non-flare CMEs observed by the Skylab coronagraph led Gosling et al. (1976) to the conclusion that flare-associated CMEs are significantly faster than non-flare CMEs. Several years later, MacQueen & Fisher (1983) analyzed six flare-associated CMEs and six non-flare CMEs associated with filament eruptions observed by the MK3 coronagraph at the Mauna Loa Solar Observatory, and noted that the former ones were faster, moving at approximately constant velocities, whereas the latter were slower, showing considerable accelerations. Combining their results with the results by Gosling et al. (1976) and Hildner (1977), MacQueen & Fisher (1983) put forward the concept of two distinct classes of CMEs: the flare-associated ones, being accelerated impulsively at low heights, and the non-flare CMEs accelerated gradually over a large height range. Presently, the concept of two distinct CME classes is widely employed to interpret various aspects of CMEs (e.g., Sheeley et al. 1999; Andrews & Howard 2001; Low & Zhang 2002; Moon et al. 2002; Chen & Krall 2003).
While it is true that CMEs accompanied by major flares are usually very bright, fast and wide, the concept of two distinct classes of CMEs faces serious problems when confronted with observations. The first one appears in statistical analyses: for example, Fig. 12 of Moon et al. (2002) indicates that there is a correlation between the speed of CMEs and the time-integrated soft X-ray flux of the associated flares. Similarly, Fig. 6 of Burkepile et al. (2004) reveals a distinct correlation between the kinetic energy of CMEs and the SXR peak flux in a sample of 24 CME/flare limb-events. Finally, Moon et al. (2003) found a very good correlation between CME velocities and the associated SXR peak fluxes in two carefully selected sets of CME/flare events. The graphs show a "continuum'' of events (no distinct grouping), and furthermore, a considerable number of events is slow (much slower than usually considered as typical for the flare-associated CMEs). In this respect it should be noted that the samples used in the analyzes by Gosling et al. (1976), Hildner (1977), and MacQueen & Fisher (1983) are in fact not statistically reliable - the samples were small, preferably comprising large events. Finally, we emphasize that there is a certain controversy regarding the fraction of CMEs having "constant speed'' or "constant acceleration'' (for a discussion, see Sect. 5 of Andrews & Howard 2001). The controversy arises from the fact that the judgement on the type of kinematical curve is often highly subjective, and furthermore it depends on the performance of the instruments, the measurement technique, and the morphological characteristics of a CME.
The second problem is the various case-study counterexamples in the literature. For example, the
two flare-CMEs studied by Zhang et al. (2004) attained the acceleration maximum at
and
solar radii. Similarly, Moon et al. (2004) reported three
flare-associated CMEs, two of which clearly showed the acceleration phase in the range R>2. On
the other hand, the quiescent filament (fast) eruption of 18 August 1980 showed the acceleration
peak at a height of around H=0.5 solar radii above the solar surface (Illing &
Hundhausen 1986; Vrsnak et al. 1993, and references therein).
Furthermore, the rapid acceleration stage of CMEs is closely synchronized with the rising phase of the associated soft X-ray bursts (Zhang et al. 2001; see also Vrsnak et al. 2004a, and references therein). So, the duration of the CME acceleration depends on the impulsive/gradual character of the flare: CMEs accompanied by gradual flares should accelerate over a large distance range, contradicting the basic presumption of the two-class CME scheme.
In this paper we check the statistical aspect of the two-class CME concept, by comparing a large sample of flare-associated CMEs (hereafter F-CMEs) with a sample of non-flare CMEs whose source region is revealed by disappearing filaments (hereafter DSF-CMEs). After describing the data set and the sampling procedure in Sects. 2 and 3, we present in Sects. 4 and 5 the comparison of the two samples. The results indicate that the two-class scheme has to be reformulated since there is no distinct difference between DSF-CMEs and the majority of F-CMEs. Only CMEs associated with major flares show on average larger velocities. However, even this category of CMEs involves a significant fraction of slow events. On the other hand, the sample of DSF-CMEs contains a considerable number of fast events. The outcome is interpreted and discussed in Sect. 6, emphasizing the role of magnetic field reconnection.
The analysis is based on three data sets - the CME, the flare, and the disappearing filament data set. The initial CME sample consists of more than 5000 events observed in the range 2-30 solar radii by the Large Angle and Spectrometric Coronagraph (LASCO; Brueckner et al. 1995) in the period 1996-2001. The solar soft X-ray (SXR) flux measurements provided by the Geostationary Operational Enviromental Satellites (GOES; Donelly & Unzicker 1974), recorded in the same period 11 693 distinct flare events in the 1-8 Å GOES channel. The sample of DSF-CMEs is selected utilizing the disappearing filament data reported in the Solar Geophysical Data catalogues by various ground-based patrol observatories.
The analysis of the CME kinematics is based on the measurements of the plane-of-sky heliocentric
distances of the leading edge of CMEs, R(t), expressed in units of the solar radius,
. In the analysis we consider only the events
whose leading edge position was measured in at least four
instants. We employ the following CME
parameters:
In order to associate flares to CMEs we have applied temporal and spatial criteria. First, for each
CME we used the linear back-extrapolation of its trajectory to the surface of the Sun in order to
estimate roughly the lift-off time t0. Then, following the statistical procedure proposed by
Dougherty et al. (2002) we estimated that the flares that occurred out of the temporal
window
days are most likely not associated with a given CME
(0.04 days
1 h). In this way we established a set of 1321 CME-flare pairs whose
relative timing was indicative of a causal relationship. Finally, we required that the position
angle of the flare is within the position angle interval spanned by the CME:
.
In this way we established a sample of 545 CME-flare pairs that are
most likely physically related. The sample consists of 28, 153, 343 and 20 flares of the GOES
intensity-class X, M, C and B, respectively. The mean angular distance from the disc center amounts
to
,
which corresponds to a plane-of-sky heliocentric distance of
R=0.72.
In order to form the sample of CMEs that were associated with the filament eruptions, i.e., to
establish the sample of DSF-CMEs, an analogous procedure is applied to the leftover set of CMEs.
Due to ambiguities involved in determining the actual time of disappearance of a given filament,
the most appropriate time-window turned out to be larger: 0.08 days (
2 h). In
this way we established a sample of 545 CME-DSF time-associated pairs. After applying the spatial
criterion in the same manner as in the case of flares, we obtained a "raw'' sample of 109 DSF-CMEs.
Their mean angular distance from the disc center,
(R=0.69), is
very similar to that of F-CMEs.
Being aware of certain inconsistencies occasionally appearing in the SGD reports of active prominences and filaments, we have checked 36 fastest DSF events whose velocities v3 were larger than 600 km s-1, since these events are "incompatible'' with the two-class CME hypothesis. Utilizing the images gained by the Extreme-Ultraviolet Imaging Telescope (EIT) aboard the Solar and Heliospheric Observatory (Delaboudiniere et al. 1995) and the Soft X-Ray Telescope (SXT, Tsuneta et al. 1991) aboard the Yohkoh satellite, we eliminated 5 out of the 36 checked events (14%) since they were either associated with some weak flares (three cases) or with behind-the-limb flares (two cases). Nevertheless, 31 events turned to be faster than 600 km s-1 when entering the LASCO field-of-view.
So, in the following we present the data for 104 DSF-CMEs. We emphasize that in this sample there
were 73 events whose source regions were not checked directly, possibly containing a certain
fraction of CMEs erroneously attributed as DSF-CMEs. However, most of erroneously assigned DSF
associations in the checked DSF-subset tended to be related to the fastest CMEs, so the fraction of erroneous DSF associations in the unchecked subset is probably smaller
than 14%.
Finally, we also applied a more stringent spatial requirement to both samples, using
,
in order to check the influence of sampling. It turned out that
differences in the results are statistically insignificant. Furthermore, since the spatial
criterion becomes ambiguous for large values of
,
and moreover, for
becomes meaningless, we repeated all procedures using subsamples where events with
were excluded. Although some results changed, the outcome qualitatively remained the
same. In fact, the differences between F-CMEs and DSF-CMEs diminished.
Similarly, we re-inspected all distributions considering only CMEs with
and
.
Here we noticed the increase of velocities in limb events (as expected from the
projection effects - see, e.g., Burkepile et al. 2004) and a slight shift of accelerations
towards positive values. However, the difference between flare CMEs and DSF-CMEs remained
insignificant since both samples show similar trends towards the limb.
In Fig. 1 we compare statistical properties of the F-CMEs, DSF-CMEs and the complete CME sample. The corresponding overall statistical characteristics of the distributions (mean values, standard deviations and medians) are presented in first three columns of Table 1.
![]() |
Figure 1:
Distribution of: a) accelerations (![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
In Fig. 1a we show the distribution of accelerations a. The graph reveals that there
are no significant differences between accelerations of the F-CMEs and DSF-CMEs. Both samples show
a large scatter of accelerations, generally ranging from -50 to 50 m s-2 and having a peak at
,
consistent with the distributions obtained by Moon et al. (2002). The
distributions show that it is not true that the F-CMEs generally do not accelerate (
)
and that the DSF-CMEs do accelerate (a>0) in the LASCO C2/C3 field-of-view. For example, there is
approximately the same number of a>10 m s-2 and a<-10 m s-2 events in both samples.
However, there is a slight shift of distributions that is reflected also in the mean values
(Table 1). Such a shift is qualitatively similar to that found by Moon et al. (2002), but
our results show smaller differences between subsamples.
Figure 1b shows the distribution of velocities at R=3, i.e., the velocities at the time
when CMEs enter the LASCO-C2 field-of-view. The graph shows that the velocity distributions of
F-CMEs and DSF-CMEs do not differ significantly. Even the asymmetries of the distributions are very
similar (see the median values in Table 1). So, the velocities by which F-CMEs enter into the LASCO
field-of-view are quite similar to that of the DSF-CMEs. The distributions of the mean velocities
are also similar for both samples (shown in the inset in Fig. 1b).
The distributions of
are qualitatively similar to those obtained by Moon et
al. (2002) for CMEs associated with flares of SXR importance larger than C1 and non-flare
CMEs. However, our median values (Table 1) are somewhat larger, and the difference between F-CMEs
and DSF-CMEs is smaller. Such an outcome is due to the fact that our both samples (in DSF-CMEs, as
well as F-CMEs) include wider-than-average CMEs (see below), and wider CMEs are on average faster
(Vrsnak et al. 2004b).
Note that the mean values for DSF-CMEs that are presented in Table 1 represent a lower limit since
out of the 73 unchecked events (see Sect. 3) there are probably some cases of wrong source
identification - since these 73 events generally comprise slow events, the mean values in reality
might be somewhat larger. For example, reducing the influence of the non-checked events by 14%
would give the average value
km s-1.
The distributions of angular widths are compared in Fig. 1c. Again one finds that the
distributions of F-CMEs and DSF-CMEs are quite similar. The difference of the mean values (Table 1)
is primarily due to a considerably larger fraction of halo CMEs in the F-CMEs sample. If the halo
CMEs are excluded, the mean widths become 91,
85
and 61
for the
samples of the F-CMEs, DSF-CMEs and all CMEs.
In Fig. 1d we show the distribution of the distances up to which CMEs were traced (the
last measured distance ;
for details see Vrsnak et al. 2004b). One finds that the
distributions of the F-CMEs and the DSF-CMEs are very similar in this case too. Both samples show a
flat distribution, i.e., a large fraction of CMEs is followed up to the edge of the LASCO-C3
field-of-view. On the other hand, the sample of all CMEs peaks at
.
This implies that
our F-CMEs and DSF-CMEs represent a subset that comprises relatively bright CMEs remaining visible
even at large distances. This is consistent with their larger-than-average widths, since wider CMEs
tend to be brighter (Vrsnak et al. 2004b).
Table 1: Properties of the CME accelerations, velocities and widths. The number of CMEs in subsamples is denoted as N. The subscripts at the mean values represent the medians.
In Fig. 2 we show the distribution of F-CMEs and the DSF-CMEs in the acceleration-velocity
space. Both samples show a similar anti-correlation of the acceleration and the velocity a(v3),
consistent with that obtained by Vrsnak et al. (2004b). A very similar outcome is
also found if the velocity v3 is replaced by ,
only the correlation is of somewhat lower
statistical significance.
The
relationships based on our samples are consistent with those obtained by
Moon et al. (#moon02<#388), who compared different subsamples of flare-CMEs with non-flare CMEs.
The pattern exposed in Fig. 2 can be explained by the aerodynamic drag (Vrsnak et al. 2004b). The events faster than the solar wind generally decelerate, whereas the slower ones are still accelerating. The regression line of DSF-CMEs is somewhat steeper than that of F-CMEs, which is consistent with the smaller widths of DSF-CMEs (for details see Vrsnak et al. 2004b).
It should be noted that although the overall behavior is consistent with the aerodynamic drag
interpretation, there are several fast CMEs that still accelerate (see also
graphs in Moon
et al. 2002). These events can be explained by a prolonged action of the driving force
during the flux rope eruption (Vrsnak et al. 2004b). We emphasize that such exceptions
are present in both samples and are not caused by the errors in measurement. This implies that the
prolonged action of the driving force can be found in DSF-CMEs as well as in F-CMEs.
Although the distributions of accelerations, velocities and widths shown in Figs. 1a, b and c are quite similar for F-CMEs and DSF-CMEs, the distributions of of F-CMEs are slightly shifted to negative accelerations and higher velocities (see Table 1). This is primarily caused by the influence of large flares - in the last two columns of Table 1 we present separately the characteristics of CMEs associated with flares of the SXR importance B and C, and M and X. The data show that CMEs associated with more powerful flares are faster on average, and the deceleration is more prominent. A similar trend was also reported by Moon et al. (2002), however, our mean velocities are somewhat larger and our decelerations are weaker.
The differences of the mean values of v3,
and
,
between M and X-flare CMEs and
DSF-CMEs has a statistical (t-test) significance larger than 99.99%. The mean accelerations are
different only at a P=84% significance level. On the other hand, the difference between
DSF-CMEs and B and C-flare CMEs is statistically insignificant (P=43%, 83%, 54% and
73% for a, v3,
,
and
,
respectively).
Such a behaviour is a consequence of the distinct relationship between the flare importance and the
CME characteristics, which is illustrated by Fig. 3. In Fig. 3a we show the
relationship for the F-CMEs (for
see Moon et al. 2003). The sample
is divided into subsets according to the CME widths, and each subset is fitted by the power-law,
v3=aIb. Figure 3a reveals the distinct
correlation. The least squares
power-law fit for the complete data set reads
,
with the
correlation coefficient c=0.35 and the statistical (t-test) significance larger than 99.99%.
In addition, the graph also shows that wider CMEs are systematically shifted to higher velocities.
![]() |
Figure 2: The acceleration-velocity relationship, a(v3), of the F-CMEs (gray) and non-flare CMEs (black). The linear least squares fits (shown together with the correlation coefficient c) are almost the same for the two samples. |
Open with DEXTER |
![]() |
Figure 3:
a) Velocities of CMEs at R=3 plotted versus the SXR peak fluxes of the associated
flares. The relationship is shown separately for different classes of CME widths (the bins
0-60![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The velocity-width relationship, ,
is shown in Fig. 3b separately for large
flares of the SXR-classes M and X (mean importance M 6) and small flares of the classes B and C
(mean importance C4). This presentation provides a direct comparison of F-CMEs and DSF-CMEs.
Comparing the least squares fit lines one finds that the DSF-CMEs line (dashed) is very close to
the B and C line (bold-gray), whereas the M and X subset is shifted to larger velocities
(bold-black). A similar outcome would be obtained if the velocities v3 are replaced by the CME
mean velocities
.
The correlations for DSF-CMEs and B and C-flare CMEs become even closer if
the halo CMEs are excluded.
The relationship exposed by Figs. 3a and b indirectly implies that there is a relationship between the CME widths and the SXR importance of the associated flares. Indeed, as shown in Fig. 3c, wider CMEs tend to be associated with flares of a greater importance. The horizontal arrow drawn by the y-axis shows the mean width of DSF-CMEs. Again one finds that the mean widths of DSF-CMEs are very close to the mean widths of CMEs associated with the B and C class flares.
In Fig. 3d we show the dependence of the parameter
versus the
integrated SXR flux
.
(A very similar result is obtained if
is replaced by v23.)
Presumably, the parameter E* is a proxy for the kinetic energy of a CME since a larger
implies a larger volume, i.e., larger mass. It lacks the information on the density
,
but as
shown by Vrsnak et al. (2004b), wider CMEs can be followed to larger heights, i.e., they
are brighter, implying also denser.
What makes this dependence interesting is that it has a considerably higher correlation coefficient
(
)
than the correlations
and
themselves. Since the parameter
E* couples two parameters with lower correlation coefficients, the relatively high c is even
more significant.
Finally, we emphasize that we examined also the relationship between the CME acceleration and the
associated SXR peak flux. We found only a very weak anti-correlation (t-test significance of
95%) of the form
,
where a and
are expressed in
m s-2 and W m-2, respectively. (Note that all previously mentioned relationships have a
significance >99.99%.) The correlation indicates that CMEs associated with larger flares
tend to show deceleration. Obviously, such a trend is a consequence of the relationship shown in
Fig. 3a (stronger flare implies larger CME speed) and the tendency of fast CMEs to
decelerate (Fig. 2).
Bearing in mind that the LASCO-C2/C3 field-of-view covers a large range of distances, we are faced
with the possibility that under certain circumstances the quadratic fit gives false values of the
acceleration. For example, if a CME accelerates over some limited distance range (e.g., in the
range R<6 which was considered by Gosling et al. 1976; and MacQueen &
Fisher 1983), and then moves at a constant speed (or decelerates) untill R=30, its mean
acceleration would be considerably underestimated. Thus, such a CME would be erroneously considered
as an
event.
Consequently, there is a possibility that non-accelerating DSF-CMEs are artifacts caused by the
inappropriate use of the quadratic fit. In order to eliminate such a possibility we checked the 20
fastest non-accelerating DSF-CMEs (v3>750 km s-1; ), and found that all of them
really moved at an approximately constant speed (or were decelerating) in the LASCO-C2/C3 field of
view. In Fig. 4a we show kinematical curves of several CMEs from this subsample. All
of the presented events were associated with large quiescent filament eruptions in spotless
regions, so the identification of the source region is unambiguous. The eruptions were inspected in
the EIT-SOHO and SXT-Yohkoh data to confirm that they were not associated with flares (in some of
events there were weak SXR enhancements recorded by GOES, but according to H
,
EIT and
Yohkoh images all of these weak increases were caused by remote subflares). The mean radial
distance of the source region amounts to
.
A detailed analysis of these
events will be presented in a separate paper.
![]() |
Figure 4: a) Several examples of fast non-flare CMEs that show no acceleration, or even decelerate. b) Several examples of flare-associated CMEs showing a significant acceleration. The dates (dd.mm.yy) are attached by the curves. The slopes of 100, 500, and 1000 km s-1 are indicated in the upper-left corner of both graphs. Measurements are provided by the online LASCO CME catalogue prepared by Gopalswamy et al. (http://cdaw.gsfc.nasa.gov/CME_list/). |
Open with DEXTER |
The DSF-CMEs presented in Fig. 4a are clearly not compatible with the standard two-class CME concept, where such curves are expected to characterize flare-CMEs (see, e.g., kinematical curves in Andrews & Howard 2001).
On the other hand, in Fig. 4b we show some F-CME counterexamples, selected out of 30
F-CMEs with a>25 m s-2 (i.e., having accelerations larger than the non-flare events
considered by Andrews & Howard 2001). The chosen events are characterized by a relatively
low "initial'' velocity v3, making the effect of acceleration more prominent. We avoided events
showing a complex behavior (e.g., more than one flare and/or some additional ejection in the same
region), as well as events in which the predominant direction of motion departs from the position
angle of the flare. The mean radial distance of the source region is
,
very similar to DSF-CMEs shown in Fig. 4a. Obviously, all of the presented events show
characteristics that are expected for non-flare CMEs.
Note that further F-CME counterexamples can be found among the events presented in the case studies by Plunkett et al. (1997), Zhang et al. (2001), Zhang et al. (2004) and Moon et al. (2004).
Our results imply that the concept of two distinct classes of CMEs has to be reformulated. Figure 1 reveals that a significant fraction of F-CMEs have low velocities and there are many of them that show a considerable acceleration in the LASCO-C2 field-of-view. More specifically, around 30% of F-CMEs are slower than 400 km s-1 (i.e., slower than the slow solar wind), and around 8% have an acceleration larger than 20 m s-2 (consistent with the percentage estimated by Moon et al. 2004). A similar conclusion can be drawn from Figs. 3a and 3d which show that weak-flare CMEs can be fast and strong-flare CMEs can be slow.
Furthermore, there are many fast DSF-CMEs showing a deceleration, i.e., not behaving as presumed in the two-class concept. Specifically, more than 30% of DSF-CMEs in our sample attained velocities higher than 600 km s-1 and more than 40% showed negative accelerations.
These statistical arguments are strengthened by various case studies (e.g., Illing & Hundhausen 1986; Vrsnak et al. 1993; Moon et al. 2004; Zhang et al. 2004; see Introduction) and by the counterexamples shown in Fig. 4.
On the other hand, it is true that CMEs associated with large flares are on average faster and larger than non-flare CMEs, and that their distribution of accelerations is somewhat shifted to negative values. In fact, Figs. 3 and d show a well-defined correlation between the flare importance and the CME velocity, as well as the flare-released energy and the CME kinetic energy (see also Moon et al. 2002; Moon et al. 2003; and Burkepile et al. 2004). In these scalings, the CMEs associated with B and C class flares on average have characteristics very similar to the DSF-CMEs.
This implies that there is a continuum of events rather than two distinct categories, and that
DSF-CMEs overlap with the "lower-energy'' part of the
scaling. Since the complete set of
CMEs shows still smaller velocities and widths, one can speculate that the "weakest'' events
(without a readily recognizable source region) represent the lowest-speed category of CMEs. In this
respect it should be emphasized that Fig. 3b indicates that the CME width represents an
important parameter - generally speaking, broader CMEs are faster. Consistently, the mean width of
(slow) F-CMEs presented in Fig. 4b amounts to only
,
whereas the mean width of (fast) DSF-CMEs shown in
Fig. 4a is
.
Such a scaling provides important information on the nature of the process of eruption. Dynamical flares (or morphologically, "two-ribbon flares''), taking place beneath CMEs, are a consequence of reconnection of the field lines stretched by the arcade eruption (e.g., Forbes 2000). The correlation of the CME velocity and the flare importance implies that the energy release in a flare is in a certain way related to the CME dynamics. However, our results indicate that the energy release itself (e.g., the heating and the resulting pressure pulse) is not directly responsible for the CME acceleration - if this would be the case, the DSF-CMEs lacking hot plasma signature, would be much slower than F-CMEs.
In this respect note that in DSF-events (especially the large ones) a growing system of
post-eruption EUV loops is often observed (the loops were also observed in all of DSF-CMEs
presented in Fig. 4a). Such a loop system behaves in a similar manner as the flare
loop system, although it is cooler. This implies that the reconnection, analogous to that in
dynamical flares, also takes place beneath the DSF-CMEs (Forbes 2000). The reason why such
a reconnection does not heat the plasma to flare temperatures is
probably related to a weaker magnetic field involved in DSF eruptions. Generally, these events are
associated with the quiescent filament eruptions taking place outside active regions and are thus
launched from a relatively large height (all DSF-CMEs presented in Fig. 4a are
quiescent filament eruptions from spotless regions). In such an environment the plasma-to-magnetic
pressure ratio
is higher than in active regions, which directly implies that the
reconnection associated with such events results in lower temperatures (Skender et al. 2003; Vrsnak & Skender 2004).
So, the results indicate that the reconnection itself influences the CME acceleration. There are
two effects of reconnection process which may play an important role in the CME dynamics. Firstly,
the reconnection reduces the net tension of the overlaying arcade field and increases the magnetic
pressure below the erupting flux-rope (e.g., Van Tend & Kuperus 1978; Anzer &
Pneuman 1982; Forbes 1990; Cheng et al. 2003; Lin et al. 2004).
Secondly, the upward-directed reconnection jet carries the reconnected field lines to the erupting
flux-rope, supplying it with a "fresh'' poloidal flux. This effect enhances the "hoop-force''
(e.g., Mouschovias & Poland 1978; Chen 1989; Vrsnak 1990; Chen &
Krall 2003), thus raising and
prolonging the flux-rope acceleration (Vrsnak 1990).
The feed-back effect of the reconnection is also consistent with the case studies showing that the CME acceleration is closely related to the energy release in the associated flares (e.g., Kahler et al. 1988; Zhang et al. 2001; Gallagheret al. 2003; Neupert et al. 2001; Shanmugaraju et al. 2003; Subramanian et al. 2003; Zhang et al. 2004; Vrsnak et al. 2004a). The existence of the relationship between the reconnection rate and the acceleration of the ejection was presented even more directly by Wang et al. (2003), Fletcher et al. (2004) and Qiu et al. (2004), who showed that the CME acceleration is correlated with the product of the ribbon expansion-velocity and the underlying photospheric magnetic field strength. A similar conclusion can be drawn from observations presented by Zhang & Golub (2003) who found that fast CMEs are associated with flares that show a prominent flare-ribbon expansion, whereas in flares associated with slow-CMEs the expansion is less exposed (however, note that the difference is attributed therein to a different pre-eruption magnetic field structure, i.e., normal or inverse polarity configuration).
Consequently, the acceleration lasts longer in CMEs associated with gradual-flares than in the
impulsive-flare CMEs, so there is a greater chance of observing the acceleration in the LASCO-C2
field-of-view (the three events denoted in Fig. 4b by boldface-labels are associated
with very gradual flares), especially if the source region is close to the limb (as in the case of
F-CMEs from Fig. 4b). Note that in this respect also the dimensions of eruption play a
significant role (Feynman & Ruzmaikin 2004): large features, such as quiescent filaments,
starting the eruption from a relatively large height are more likely to show the acceleration phase in the
LASCO-C2 field-of-view, especially if their source region is close to the limb. Note that the
examples presented in Fig. 4a were quiescent filament eruptions taking place close to
the limb, but nevertheless finished the acceleration phase below the LASCO-C2 field-of-view.
Finally, we emphasize that our analysis concerns only the dynamics of CMEs in the
2-30
range. So, we do not exclude the possibility of different initiation mechanisms
(see, e.g., Vrsnak 2003, and references therein), nor possibly different origin of CMEs -
e.g., the ``ordinary'' CMEs that are the consequence of the eruption of the pre-existing coronal
structure, and the so-called ``flare-sprays'' or ``ejections'' (cf., Tandberg-Hanssen 1974),
apparently being launched from the chromospheric layers (Bruzek & Durrant 1977). Similarly,
the analysis also does not exclude differences presumably arising due to the inverse or normal
magnetic field configuration of the pre-eruption structure (e.g., Low & Zhang 2002; Zhang &
Golub 2003).
Acknowledgements
We are grateful to the LASCO-SOHO and GOES teams for operating the instruments and performing the basic data reduction. We are especially grateful to Nat Gopalswamy, Seiji Yashiro, Grzegorz Michalek, and other colleagues who compiled the online LASCO CME Catalog. The authors thank the anonymous referee for constructive comments and suggestions.