NO TRACE LEFT BEHIND: STEREO OBSERVATION OF A CORONAL MASS EJECTION WITHOUT LOW CORONAL SIGNATURES

At the time of the eruption, the separation angle between the STEREO A and B spacecraft was 53 degrees. This separation allowed us to observe the CME simultaneously edge on (in A) and face on (in B). Figure 5 shows a schematic representation of the position of the CME relative to the STEREO spacecraft. We used several methods to derive the true propagation direction of the CME.

• 1.  
  The observation of a bright CME in the COR2A and an extremely faint halo in the COR2B images (see Figure 2) means that the CME propagated towards STEREO B and lies close to the POS of STEREO A.
• 2.  
  Polarization analysis of COR2 polarized brightness (pB) images suggests that the CME lay in the POS of STEREO A ±20°.
• 3.  
  Applying a forward modeling technique to the CME, Thernisien et al. (2009) estimate an angle of 26° ± 10° out of the POS of STEREO A (frontsided).
• 4.  
  The in situ observation of the arrival of a MC on June 6 in STEREO B (Yan Li 2008, private communication) confirms that the CME propagation path had a large component in the STEREO B direction.

Zoom In Zoom Out Reset image size

From the above, we estimate that the CME propagated at approximately 40° east from the Sun–Earth line. At the time of maximum acceleration (around 20:00 UT on 2008 June 1), this direction corresponds to a Carrington longitude of 65°. This also gives us an initial estimate of the position of the source region. The derived Carrington longitude corresponds to the position of the east limb as seen from STEREO A at 00:00 UT on 2008 May 31 (see Figure 4) and suggests indeed that this CME erupted from the pre-existing flux rope indicated with the white arrow.

During its evolution, the CME front is barely visible in COR1, but it can be clearly seen when it enters the COR2 FOV around 19:00 UT on 2008 June 1. The evolution of the event in the EUVI–COR1–COR2 combined FOV spans more than three days starting on 2008 May 31. The slow evolution of the event has allowed us to measure the kinematics of the event in great detail (Figures 6(a) and (b)). Because of the lack of a front in the COR1 images, we traced the back of the CME core, which is the best-observed feature in all three instruments. It is indicated with an "×" in Figure 1. Because of the CME expansion the core travels slower than the leading edge. The same feature can be seen off-limb in EUVI A (Figure 7) as it leaves the Sun. We only show the 171 Å  component where it is best observed. It is also visible in 195 Å, but not in the other 2 channels. We can identify three stages in the CME evolution: (1) a slow rise phase that starts around 20:36 UT on 2008 May 31, (2) a quasi-constant acceleration phase starting around 21:00 UT on 2008 June 1, and (3) a constant velocity phase. The asymptotic velocity of around 300 km s⁻¹ is only reached at around 20 R_☉ in the STEREO/HI (Heliospheric Imager) FOV. Here we focus on the first two phases.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Several hours before the actual eruption, EUVI A images show off-limb a helical structure rolling around its axis (see the top panel in Figure 7 and online movie). This activity may be interpreted as flux-rope activation or formation. Then a concave structure below the helical structure starts to rise very slowly in the EUVI FOV at around 10 km s⁻¹ (shown in the next two panels in Figure 7). This quasi-steady phase can be best approximated by a linear fit, until the feature reaches ∼ 2 R_☉. During this phase, the white light streamer brightens and swells. Once in the coronagraph FOV, the feature starts to accelerate gradually until it reaches a velocity of around 200 km s⁻¹ at 13 R_☉. At this stage, it has developed into the bright core of the CME. We find that this acceleration phase is well described by a second-order curve (red curve in Figures 6(a) and (b)).

Under certain conditions, different initiation models predict different height–time profiles (e.g., Schrijver et al. 2008). Motivated by this, we also fitted an exponential function (green, dash-dotted) and a power-law function (blue, dotted). Up to now, and to the best of our knowledge, researchers have fit only second-order curves to these gradual type events. From Figures 6(a) and (b), we can see that the three colored curves describe relatively well the acceleration phase. This implies that, at least for this event, it is impossible to distinguish between the existing models based on the kinematics alone.

From our estimation of the true propagation direction (Section 2.1), we derive that EUVI B observes the source region face-on. Figure 4 shows a pair of images taken simultaneously by EUVI 171 Å  before the eruption. The white boxes extend from 30° to 100° in Carrington longitude and from −40° to 0° in latitude. The latitudinal extent of the box is derived from the edge-on view seen in EUVI A. In longitudinal direction, the box is centered at 65°, which is the direction of the CME derived in Section 2.1 roughly at the time when the CME detaches from the Sun. Close inspection of EUVI B data (all wavelengths) of the source region does not reveal any clear on-disk signature in any of the four EUVI wavelengths. There is no active region or sunspot group, no apparent filament eruption that caused the CME, no GOES X-ray emission (see Figure 6(c)), no post-eruption loop arcades, no EUV wave and no obvious EUV dimming. We strongly encourage the reader to inspect the movie that is available online (the movie runs from May 31 to June 2). Figure 7 shows five snapshots from EUVI A and B 171 Å  illustrating the low-coronal evolution during the event. As can be better seen in the accompanying movie, we only detect small-scale activity in the B images, which is ubiquitous in the quiet-Sun. Several small features are visible that may be related to the eruption. For example, we encircled in yellow (f3) a small bright patch that could be evidence of small-scale heating. Also, the small prominence indicated in orange (f2) that was seen off-limb in EUVI A 304 Å  cannot be readily identified in any of the EUVI B channels, which suggests that it is not very dense. The footpoint of it in the EUVI B FOV is indicated with a diamond. It is also interesting to note that we observed a small filament in Hα just after the CME erupted. Its shape is outlined in turquoise in the bottom frames of Figure 7. The filament forms fast, is not very conspicuous and disappears quickly between 21:00 and 22:00 UT (S. Martin 2008, private communication). While this filament is not related to the observed CME, it indicates the presence of a filament channel at the time of the eruption (e.g., Martin 1998). Further, a small dimming area is indicated (magenta box in f5). While these small features may have a relation to the observed CME, their scale is much smaller than the CME itself. Moreover, the movie illustrates that similar activity is observed outside the box, thus it is difficult to decide which features are connected to the observed CME.

The most surprising result is the lack of observation of an eruptive filament in the EUVI B images. We expected to see something given the clear helical structure seen in both EUVI A and COR2A. To make sure, we examined Hα images from varied sources, but no Hα filament was observed prior to the CME. Hα images are only available from Earth, which solar view is between that of STEREO A and B. On 2008 May 31 00:00 UT, the east limb of the Sun, as seen from the Earth, has Carrington longitude 395, which is 255 east of 65°, our initial estimate for the source region location from Section 2.1. If a filament had erupted from around that longitude, we should have observed it in the Hα images.

Because of the detailed coronagraph observations in STEREO A, we can assess the height from which the CME material originates by measuring the mass flux in the COR1 and COR2 FOVs (shown in Figure 6(d)). The CME mass is estimated by integrating over a region of interest (ROI), the excess brightness of a given image relative to a pre-event image. We defined the ROI as a sector bounded by the CME edges in position angle and by 1.8–3.7 R_☉ in COR1 and 3–14 R_☉ in COR2 in radial direction (the COR2 sector is drawn in Figure2), thus the curves in Figure 6(d) represent the mass evolution in these ROIs. The maximum mass measurement in the COR2 FOV is a representative number for the total CME mass, being ∼ 3.5 × 10¹⁵ g which is normal for such events (e.g., Vourlidas et al. 2002). The COR1 FOV is too small (in radial direction) to capture the whole CME at one instant of time, therefore the COR1 curve does not reach the full CME mass. To obtain a rough cross calibration between the two instruments, we compared the mass measurement in a common sector (from 2.4 to 3.7 R_☉) while the CME traveled through this ROI. The COR1 mass curve was on average 20% lower than the COR2 mass curve.

The streamer swelling is represented by the initial slow rise in the mass curves. When the CME enters the COR2 FOV a sharper increase can be seen, until the leading edge of the CME reaches the outer edge of the COR2 ROI after which the mass decreases back to its pre-event level. As we can see from the dashed curve, the corona in the COR1 FOV does not recover its pre-event level showing an apparent "negative mass." This indicates that the COR1 corona was depleted. As a check, we did the same exercise for the 2008 April 26 CME, which was not a streamer blowout and had a clear source region and dimming. The COR1 mass curve for that event did return back to its pre-event level, implying that most of the CME mass originated from below the COR1 FOV. Thus, we can say with confidence that a large fraction of the CME mass in our event originated from the COR1 FOV.

Standard assumptions in mass measurements are (1) that all of the CME material is located in the plane of the sky (therefore, it gives a lower limit for the true CME mass) and (2) that the corona is completely ionized (Vourlidas et al. 2000). Since our CME lies close to the plane of the sky of STEREO A (where the mass measurements are made) these measurements are close to the real values.

Arguably, the most intriguing aspect of our observations is the lack of a well-defined source region of the event in EUVI B data (see Section 2.3). Using the information from the two viewpoints, we draw some estimates about the size of the source region along L_SR and across θ_SR the neutral line. Often the length of the erupting filament is used as a proxy for the length of the neutral line. Based on a large statistical survey of observations of erupting filaments and their associated CMEs, Cremades & Bothmer (2004) found that the observed lengths of the filaments ranged from a few degrees up to more than 40° and show a Poisson-type distribution peaking in the range 6°–12°. We do not directly observe a filament and therefore cannot measure its length, but from the edge-on orientation of the CME and the PFSS magnetic field extrapolation we estimate that the orientation of the neutral line in question is more or less in the longitudinal direction. Assuming that the two small features indicated in Figure 7 were both related to the CME, their longitudinal separation gives us a lower limit of L_SR ≈ 35°. A second estimation can be made using the three-dimensional information derived from a forward modeling technique (Thernisien et al. 2009). This technique models the CME as a hollow flux-tube (see schematic view in their Figure 1). The separation angle 2α between the legs of the flux-tube and the aspect ratio κ of the tube give an upper limit for the length of the source region. From the values given by Thernisien et al. (2009) for our source region (see their Table 1b), we find then $L_{\rm {{SR}}}\le 2\alpha +\arctan (\kappa)\approx 36^\circ \pm 10^\circ$.

Another interesting quantity is the size θ_SR of the source region across the neutral line, which should be commensurate to the footpoint-separation of the associated erupting arcade. Moore et al. (2007) determined a scaling-law relating the CME width θ_CME (which is what we observe edge-on in EUVI A) to the size of the associated erupting arcade. θ_CME corresponds to the CME angular span when the CME is fully developed in the COR2 FOV and has reached lateral pressure balance with the surrounding magnetic fields. This is thought to occur in the outer corona (i.e., >2 R_☉). The scaling-law (e.g., Equation (20) in Moore et al. 2007) can be written as

$$\begin{equation} \theta _{\rm{{SR}}} \approx \sqrt{\frac{1.4}{\langle B_r\rangle }} \theta _{\rm CME}\cdot \end{equation} \tag{ 1 }$$

〈B_r〉 is the absolute radial magnetic field averaged over the extended quiet-Sun area enclosed within the small box of Figure 4. A full disk photospheric magnetogram from SOLIS (Keller et al. 2003) was used for this calculation (registered from 20h55 to 21h42 on 2008 June 1). We found 〈B_r〉 ≈ 3 G, which indicates a very weak field in the ROI. Similar values were obtained from a Mt Wilson magnetogram (Yi-Ming Wang 2008, private communication). Further, we use the CME angular span θ_CME ≈ 54° as taken from the online CACTus catalog (Robbrecht & Berghmans 2004). Using these values, we find a source region size θ_SR ≈ 34°. This value is comparable to the latitudinal width of the box we derived at first sight from the edge-on view in EUVI A in Figure 4. This width is commensurate with the footpoint separation of the red magnetic field lines (see Figure 3) that erupted.

Summarizing the above, we deduced a source region size of about 36° × 34°. These values of the source region should be viewed only as estimates. They nevertheless point to a rather extensive source area. The main conclusion is that we expect a large source region, something which is in concert with the inferences of a large flux rope mentioned in Section 2. More refined calculations could pin down further the dimensions of the elusive source region. For example, we plan to determine the detailed energetics of our event as well as the magnetic properties (e.g., fluxes) of the associated MC. These values would allow some further estimates of the size of the source region, to be deduced from the conservation of the total CME energy and of the magnetic flux in MCs, respectively.

Flaring is usually expected for CMEs associated with active regions. As is obvious from Figure 7, no active region was present anywhere on the disk as seen from EUVI B. The CME originated in the quiet-Sun and the lack of flaring is not surprising. Besides energetic flares, polar crown filament eruptions are often accompanied by long post-eruptive arcades seen in EUV and soft X-rays. This type of flaring was not observed for our event, either on-disk (e.g., as in Hudson et al. 1995) or off-limb (e.g., as in Sheeley et al. 2007) which could imply that the field is closing at a large height (small densities) and that the energy release was small resulting in weak heating unobservable by EUVI.

The kinematics of our CME and its morphology characterize it as a gradual event usually linked to a prominence eruption. Why then is there no clear prominence visible in the EUVI B images, nor in Hα? EUVI A images provide an important clue: they show a very high flux rope (as derived in Section 2). The combined observation of a flux rope, only visible off-disk in the hotter EUVI channels and a small and tenuous filament, barely discernible in the EUVI A 304 Å  implies a hot flux rope with small amounts of cool material.

The presence of a cavity and the PFSS extrapolation confirm that the CME originated from a polarity inversion line and shows that the existence of an Hα filament is not a necessary condition for the eruption. Filament channels coincide with polarity inversion lines and when chromospheric mass loads into the channel, a filament is formed. As most CME initiation theories, models, and observations suggest, the important agent for an eruption is the existence of a filament channel only, not the chromospheric mass. For example, Lin (2004) pointed out that the magnetic configuration of the filament channel is more important than any mass loading for CME initiation. However, there is a widespread belief that Hα filament disappearances are reliable proxies for CMEs. We would like to caution observers and space weather forecasters against an over reliance on Hα data for disk signatures of eruptions. Our observation shows that CMEs can erupt without having a mass-loaded (and thus observable) filament.

Finally, we address the lack of a large-scale dimming signature in the EUV. Estimates of the mass associated with EUV dimmings showed it can represent a significant fraction of the total CME mass (e.g., Harrison et al. 2003; Zhukov & Auchère 2004). If the CME originated from deep in the corona carrying some part of it, we would have expected to see a large EUV dimming commensurate to the amount of mass carried off by the CME.

The electron density, n_e(r) drops dramatically with radial distance from the solar surface. The EUV intensities, proportional to n²_e(r), exhibit an even stronger fall-off. A small difference in initiation height of the ejected material can give a large difference in the signal of the coronal dimming. Conversely, the absence of a dimming suggests a large initiation height for our CME. To demonstrate this, we calculated proxies of the EUV intensities before, I_BG, and during, I_CME, the eruption of a hypothetical CME with its source region located on disk center. The background intensity per pixel is defined as

$$\begin{equation} I_{\rm BG} \equiv \int ^{\infty }_{1\,{{\rm R}}_\odot }{n^2_e(r) dr} \approx \int ^{10\,{{\rm R}}_\odot }_{1\,{{\rm R}}_\odot }{n^2_e(r) dr}, \end{equation} \tag{ 2 }$$

where r is the distance from Sun center. Similarly, the EUV intensity corresponding to the portion of the corona that is removed by the CME corresponds to

$$\begin{equation} I_{\rm CME} \approx \int ^{10\,{{\rm R}}_\odot }_{r_0}{n^2_e(r) dr}, \end{equation} \tag{ 3 }$$

assuming a starting height of r₀ for the CME. A ratio I_CME/I_BG significantly below 1 implies an intensity depletion caused by the removal of part of the corona by the CME. If this depletion occurs simultaneously over a large set of continuous observational pixels then we observe a dimming.

Figure 8 shows this ratio for starting heights r₀ in the range 1–2 R_☉. For n_e(r), we used the density profiles of Baumbach–Allen (Aschwanden 2005; Gibson et al. 1999; Guhathakurta et al. 1996). The first profile (solid line) corresponds to an "average" corona, whereas the other two correspond to coronal streamers. As expected, the dimming ratio I_CME/I_BG is appreciably smaller than 1 only for starting heights below 1.4 R_☉. For CMEs starting higher up in the corona, this ratio is approximately 1, meaning that for those CMEs no observable coronal dimming can be expected. Conversely, an absence of dimming implies that most of the plasma that contributed to the CME mass originated from above 1.4 R_☉. For reference, we note that this height corresponds to the inner edge of the COR1 FOV. This limit is quite robust given that the dimming curves have little dependence on the employed density profile as can be seen in Figure 8.

Zoom In Zoom Out Reset image size

Including temperature effects (through temperature response functions R(T)) into the determination of the EUV intensities would further compress the radial distance range of sensitivity of I_CME/I_BG. This is because for a multi-thermal line of sight only points with temperatures around the peak of the R(T) of the employed channel will contribute to the observed intensities. Moreover, EUV and soft X-ray observations show small or little variations of the electron temperature in streamers (e.g., Foley et al. 2002; Warren & Warshall 2002). Finally note that a large-scale dimming, if present, would have manifested at least in one or more of the four EUVI channels. The response functions of these channels peak at 0.08, 0.90, 1.50, and 2.00 MK, which spans the bulk of the quiet-Sun temperature domain (e.g., Brosius et al. 1996). We found no evidence of significant flaring, which means that no large amount of plasma was heated off quiet-Sun conditions.

To summarize, we offer the following scenario based on the above discussion. The event occurs over a largely empty filament channel, hence no or weak emission in 304 Å  or Hα, on the quiet-Sun. The flux rope, that eventually forms the CME core, is visible for at least two days prior to the eruption in emission in the coronal lines (171 and 195 Å). We deduce that the flux rope mostly consists of hot material of about 1 MK and the material is concentrated at the bottom of the feature. The flux rope is also situated at 0.15 R_☉ (bottom) above the surface, which is unusually high. We illustrated the large height of the overlying loop system, by comparing the cavity beneath it with the smaller cavity that is visible in EUVI B, and found that our cavity is 2.2 times larger. This is further corroborated by the magnetic field properties of the postulated source region. It shows that most of the overlying field lines have widely separated footpoints, comparable to the widths derived in Section 2.5. We believe that this is the first EUV observation of such a high-lying flux rope. This observation was possible thanks to the large EUVI FOV and the wavelet processing. The EUVI A time series show activity at the flux rope such as structures rising up and into the feature causing it to rotate. This is a process very suggestive of tether-cutting and explains the existence of hot plasma in the flux rope. We think that the flux rope is likely to have formed over several days via successive "tether-cutting"-like events, storing free magnetic energy that was later used to drive the CME (e.g., Gibson et al. 2006). The last and most severe one must have occurred on May 31 after which the flux rope is clearly seen rising and the CME is in process. The scenario of the large flux rope is consistent with the lack of heating or dimming signatures and the lack of any appreciable eruptive filament but is also consistent with the standard reconnection model of CMEs (e.g., Forbes 2000).