CORONAL SOURCES, ELEMENTAL FRACTIONATION, AND RELEASE MECHANISMS OF HEAVY ION DROPOUTS IN THE SOLAR WIND

In our previous study (Weberg et al. 2012) we identified 257 heavy ion dropouts between the years 2001 and 2009. Dropouts were identified primarily as time periods longer than 4 hr of low Fe/H more than one standard deviation (STD) below the local Carrington rotation (CR) average with some of the data points more than two STD below the CR average. This last criterion has no minimum time duration requirement (any number of very low data points is sufficient) and was imposed to exclude small events within normal solar wind variability. Special care was taken to exclude events with data gaps, sharp, small-scale enhancements in hydrogen density (which gives spurious low Fe/H ratios with nominally the same iron densities), and those which occurred during known ICMEs on the most up-to-date version of the Richardson & Cane (2010) list. Using the same methodology, we now extend our study to include five more years of data: 1998–2000 and 2010 and 2011. The resulting 14 yr of observations span more than a solar cycle, from early cycle 23 to early-mid cycle 24 (1998 February to 2011 August).

Figure 1 shows the annual number of dropouts. The red line corresponds to the smoothed monthly sunspot number available from the World Data Center (SILSO).⁴ For the sake of determining long-term trends, the bars for 1998 and 2011 have been normalized assuming the same observation rate for the missing months. The observed values are reported in Table 1. Altogether we have identified 379 total dropouts. The blue shaded regions in Figure 1 indicate the number of the dropouts which were found to be strongly mass fractionated (i.e., heavier ions more depleted that lighter ions) relative to the nearby slow type solar wind. The method used to determine and quantify mass fractionation is the same as described on Section 4.1 of our earlier paper with one notable difference: this time the X/H ratios inside the dropouts were only compared relative to the nearby (±5 days) slow wind rather than just all nearby solar wind as before. This change is motivated by the fact that heavy ion dropouts appear to originate from similar regions of the Sun as slow solar wind. Comparing to nearby slow wind allows for cleaner, less ambiguous fractionation trends and minimizes the impact of other fractionation processes such as the FIP effect which is different in the fast wind (we explore this process more in Section 2.2). Based on our analysis and classification scheme, 36.4% (138 events) of all dropouts exhibit mass-dependent fractionation.

Zoom In Zoom Out Reset image size

As we noted in our previous paper and can see in Figure 1, there does not appear to be an immediately obvious correlation between solar cycle and yearly dropout numbers. If we focus on 5 yr intervals centered on solar maximum and minimum, we find that there are on average 12 fewer dropouts per year around solar maximum (1999–2003; 22.4 ± 3.72 per year) than there are around solar minimum (2006–2010; 34.8 ± 3.92 per year). However, the number of mass fractionated dropouts remains fairly constant with an average of 9.9 ± 2.5 per year. Therefore the relative fraction of mass fractionated dropouts is lower during solar minimum, dropping from 43.4% ± 5.1% to 33.1% ± 6.3%. While it is possible that ICMEs, by disrupting the coronal conditions required for gravitational settling, could be partially responsible for the fewer dropouts seen around solar min, the steady observation of strongly mass fractionated dropouts suggests that there exists some fraction of source regions that remain relatively unperturbed by ICMEs. In the future we plan to more closely investigate dropouts found near ICMEs and provide a quantitative analysis of their interactions (if any).

Dropouts have been observed to have durations ranging from 6 to ∼82 hr. However these durations are not normally distributed with only 21.6% (82 events) having durations longer than 24 hr. The presence of extreme outliers means we must turn to the median and the median absolute deviation (MAD) about the median for a better quantification of typical duration and variability. We find that dropouts have a median duration of 12.02 hr and a MAD of 4.03 hr. Table 1 explores in further detail the yearly numbers and durations of dropouts. We find that the median duration is fairly uniform across the solar cycle. It is significant to note that, between 1998 February and 2011 August, heavy ion dropouts comprise 5.5% of all solar wind observed by ACE. This is on the same order of magnitude as the 7.6% of solar wind contained within ICMEs on the Richardson & Cane (2010) list.

Figure 2 compares the relative X/O ratios inside dropouts to those of the fast and slow solar wind. Ratios are given relative to the photospheric values and are ordered according to FIP. With the exception of iron, heavy ion dropouts appear to have X/O ratios similar to the slow solar wind and likewise exhibit a low FIP enhancement. This result is not entirely unexpected; within a typical dropout, all of the elements are depleted to some extent, including O. Therefore the relative X/O ratios remain relatively unchanged even though the absolute abundances of the input elements are lower. The one exception to this is iron—which being the heaviest element observed by SWICS—is significantly more depleted than O such that the Fe/O ratio is noticeably lower inside the dropout. Furthermore, current theories of the FIP effect place the driving mechanism in the chromosphere (Laming 2004) while gravitational settling, although occurring over all heights, would be most evident higher up in the corona. This means that solar plasmas are already FIP fractionated by the time they reach the corona and we would expect to still see some evidence of this even after gravitational settling has taken place.

Zoom In Zoom Out Reset image size

The effects of gravitational settling are more evident when looking at the X/H ratios. In Figure 3 we compare the [X/H]/[X/H]_photo ratios of dropouts to typical fast and slow solar wind (computed using averages from von Steiger et al. 2000 and the O/H ratio of von Steiger et al. 2010). Elements are arranged in order of mass. We also show the composition of a single, large helmet streamer spectroscopically observed by Raymond et al. (1997) which was found to be mass fractionated. As expected, the slow and fast wind values can be explained by FIP fractionation with the low FIP elements (Mg, Si, and Fe) being enhanced over those with high FIP (He, C, N, O, and S). Additionally, it can now be seen that all the elements inside the dropouts are, in fact, depleted relative to the photosphere; a fact that was not apparent from just looking at the X/O ratios. There are also some clear mass-dependent trends, particularly in the heaviest elements although they still exhibit increased ratios in accordance with FIP fractionation. It is worth noting here that the "error bars" shown in Figure 3 indicate the STD in the mean X/H ratios rather than the instrumental uncertainties in their values (which are much smaller).

Zoom In Zoom Out Reset image size

In contrast to Figure 2, the fast wind values in Figure 3 are actually higher than the slow wind due to the fact that the relative O/H ratio is lower in the slow wind (as found in von Steiger et al. 2010). The result is increased slow wind X/O ratios despite slightly lower relative X/H. This observation underlines the usefulness and extra insight provided by analyzing both the X/O and X/H ratios in tandem. Certain data trends that may be hidden in one measurement space can become apparent in the other.

Now that we have established a fairly complete list of dropouts between the years of 1998–2011, we now try to discover, or at least constrain, probable source region locations and release mechanisms in the corona. We do this by correlating the dropouts with plasma and magnetic structures embedded within the underlying solar wind. For example, if a particular dropout originated from a coronal streamer we would expect it to be well correlated with a HCS crossing and a stream interface since coronal streamers are known to lie below the HCS and emit slow velocity solar wind. In our analysis we focused on magnetic sector boundaries, true HCS crossings, local current sheets (CSs), and stream interaction regions (SIRs). Each structure was defined and identified as follows.

• 1.  
  Magnetic sector boundary (SB). Since suprathermal electrons (energies > 70 eV) are always flowing away from the sun along local magnetic field lines, their bulk streaming direction (as measured by their pitch angle distribution) is thought to indicate the absolute polarity of the magnetic field line at its footpoint on the Sun (Kahler et al. 1996). Parallel streaming electrons indicate outward (i.e., positive) magnetic polarity while antiparallel flows are evidence of inward (negative) directed fields. In our study, a SB was identified by a sudden shift in the 272 eV suprathermal electron pitch angle distribution from parallel to antiparallel (or vice versa). In order to filter out marginal or chaotic events, we require the pitch angles to be fairly steady 6 hr before and after the shift. Electron pitch angle distributions were acquired using level 3 data from ACE/SWEPAM-E (McComas et al. 1998) available on the ACE science center Web site.⁵
• 2.  
  Heliospheric current sheet. We identified HCS crossings by sharp reversals of the magnetic field azimuthal angle (in RTN coordinates) between the nominal angles predicted at Earth by the parker spiral (135^o and 315^o where 0^o is defined as pointing radially away from the Sun). Additionally, we required that reversals were located within 12 hr of a SB and that the magnetic field direction was moderately steady (i.e., < 45^o variation) 12 hr before and after the reversal. In all of this we used the 1 hr data from ACE/MAG data (Smith et al. 1998).
• 3.  
  Local current sheet (CS). We defined a local CS as a magnetic field reversal without the accompanying flip in suprathermal electron streaming as is found at the HCS. This may indicate either an incomplete crossing of the HCS or local folds and kinks in the initial mass function caused by interchange reconnection near the cusp of a helmet streamer (Crooker et al. 2004).
• 4.  
  Stream interaction region. SIRs occur wherever a fast velocity solar wind stream overtakes a slow stream. SIRs are found using a variety of plasma criteria and their identification is a large topic of research. For simplicity and reliability, we use a list made by Jian et al. (2011). It should be noted that this list, by definition, does not include the rarefaction regions where fast streams outrun slow.

Individual dropouts were required to start or end within 12 hr of the observation of a particular solar wind structure in order to be considered correlated with that structure. Unfortunately the Jian SIR list only goes up to the year 2009, thereby our statistics for 2010 and 2011 are incomplete. For the sake of accuracy, we give separate values for the years 1998–2009 (323 events) and 2010 and 2011 (56 events). Table 2 summarizes our findings and has a number of interesting features. First, there does not seem to be any significant difference in structure correlations between all dropouts and only those strongly mass fractionated. This suggests that all dropouts have similar distributions of source region types. Secondly, no single structure or combination thereof dominates in such a way as to indicate a very specific coronal source. For example, if a majority of dropouts were found well correlated with both the HCS and a SIR, we could reasonably conclude that dropouts mostly came from large helmet streamers at the base of the HCS and not pseudosteamers, which are often identified in-situ by a SIR without a CS of any kind (Wang et al. 2012). Finally, perhaps the most striking feature in Table 2 is the relatively large percentage of dropouts (37.7%) that do not appear to be correlated with any of the major solar wind structures.

Figure 4 shows a modified superposed epoch analysis plot of all 1998–2009 dropouts correlated with SIRs (37.1% of all events), regardless of whether or not HCS, CS, or SB correlated. The averaging for the period prior to the dropouts (left unshaded region) was performed with zero epoch centered on the start of the events. Likewise, the period after the dropouts (right unshaded region) was averaged centered on the event end times. Inside the shaded region we normalized and rescaled each dropout event to the median duration and then averaged the data at each time step. The result is a single plot that describes the average behavior of plasma properties before and after the dropouts while clearly separating the solar wind and dropout data.

Zoom In Zoom Out Reset image size

As we would expect for events correlated with SIRs, the dropouts averaged in Figure 4 exhibit plasma signatures consistent with slow wind found ahead of slow-to-fast solar wind transitions which suggests source regions are outside coronal holes, but near their leading edges. Primary features are the smooth rise in solar wind velocity accompanied by a peak in magnetic field magnitude, an increase (i.e., compression region) of proton density, and a transition from high O7/O6 signifying "slow" type wind) to low O7/O6 ("fast" type solar wind). Figure 5 shows a similar superposed epoch plot for all 1998–2009 dropouts uncorrelated with well-defined solar wind structures (last column of Table 2). In contrast to Figure 4, the events averaged in Figure 5 appear to occur after fast-to-slow solar wind transitions, but again in the slow wind. However, the trends, while qualitatively opposite to those described above, are somewhat less distinct. It appears then that these "uncorrelated" events in fact have sources at the trailing edges of coronal holes and may potentially help shed light on rarefaction regions in the solar wind which are notoriously difficult to define. When combined with the 37.1% of dropouts correlated with SIRs, we find that 74.8% of all dropouts can be associated with sources in the closed corona on either side of coronal holes. This is not surprising since it is at the boundary between open and closed fields that one would expect to find the reconnection processes able to transport closed-field plasma out into the heliosphere.

Zoom In Zoom Out Reset image size

When considering the correlation of dropouts with solar wind structures, it is vitally important to keep in mind the limitations inherent in single spacecraft measurements. In certain three-dimensional (3D) geometries it is possible for ACE/SWICS to sample plasma from the near vicinity of a given structure without actually crossing and observing the structure itself. Therefore the values in Table 2 should be considered as only approximate lower bounds. While SIR correlations are unavailable for 2010 and 2011, we note that the relative fraction of dropouts correlated with HCS crossings is almost the same as the sum of the "HCS with SIR" and "HCS without" fractions for the years 1998–2009. Similar comparisons hold for CS correlated events as well as the dropouts exhibiting no clear correlating structures. It is therefore reasonable to assume the actual correlations in 2010 and 2011 are nearly the same as 1998–2009.

We now return to a more sophisticated analysis of dropout occurrence rates to see what information we might extract concerning the temporal or spatial distribution of source regions in the corona. Figure 6 shows a plot of dropouts within their respective CRs across the entire observation interval. The figure can be thought of as one long time series that has been sliced up by CR and stacked on end. The left Y-axis gives the CR number while the X-axis corresponds to the day within the CR. Colored bars represent all heavy ion dropouts we have identified so far. The widths of the bars indicate the actual duration of the events and the colors correspond to the solar wind structure correlations. A careful examination of Figure 6 leads to the observation that a large number of dropouts seem to belong to series which reappear over multiple solar rotations (highlighted by red ellipses). Altogether there are 62 potential dropout series comprising 40.6% of all dropouts and lasting up to 4 CRs in a row. This systematic reoccurrence suggests that all dropouts within a given series may originate from the same long-lived source region in the corona that is continuously, or semi-continuously, emitting heavy ion depleted plasma.

Zoom In Zoom Out Reset image size

Before we begin to analyze these "reoccurring dropouts," we calculate the time lag between dropouts to help determine whether or not reoccurrence could be simply explained by a random distribution of events. Figure 7 is a histogram of the time difference between each dropout and all subsequent dropouts that come after. The clear peaks centered on multiples of the CR period (red vertical lines) confirm that dropout reoccurrence is in fact a reality. If the events were random, we would not expect to see such periodicity. As a further test, we simulated a simple Poisson distribution of random events but were unable to reproduce anything near the reoccurrence rates observed. This indicates that the probability of reoccurrence being the result of a random process is exceedingly low.

Zoom In Zoom Out Reset image size

Reoccurring dropouts provide a unique opportunity to study the long-term evolution of coronal processes within the same source region. Furthermore, the time difference between observations of dropouts within a series can be used to estimate the source latitude based on the rotation rate of the source region. Assuming constant solar wind velocity and mostly radial propagation, we first determine the approximate times, t₁ and t₂, that two sequential dropouts in a series left the corona:

$$\begin{equation} t_i = t_{i,ACE} - \frac{{{\rm distance\; to\; Sun}}}{{\left\langle {V_{{\rm SW}} } \right\rangle }}.\end{equation} \tag{ 1 }$$

The angular velocity of the source region is then given by

$$\begin{equation} \omega = \frac{{2\pi + \lambda }}{{t_2 - t_1 }},\end{equation} \tag{ 2 }$$

where λ is the angle the Earth rotated around the Sun between observations of the dropouts. The angle λ was obtained using the PyEphem⁶ package for the Python programming language. PyEphem bases its ephemeris calculations on numerical routines from the scientific-grade XEphem astronomical software (Downey 2011). Due to differential rotation, higher solar latitudes are observed to rotate more slowly than near the equator (Snodgrass & Ulrich 1990). This rotation is described by the equation

$$\begin{equation} \omega = A + B\, {\rm sin}^2 \left(\varphi \right) + C\, {\rm sin}^4 \left(\varphi \right),\end{equation} \tag{ 3 }$$

where A, B, and C are constants with the values 14.713, −2.396, and −1.787 deg day⁻¹ and ϕ is the heliographic latitude. By comparing our calculated rotation rates to the standard model for the Sun's differential rotation, we are thereby able to determine the approximate latitude of the source region. It should be emphasized here that our method is only a simple first-order approximation; nevertheless it is driven directly by data and is free from many of the assumptions made in more complicated mapping techniques using magnetic field line tracing. The most significant assumption we make is that the structure of the corona near our source regions remains relatively stable on the order of a CR. While this is certainly true during most of solar minimum, the results at solar maximum should be viewed with a bit of caution since the corona is much more complex at that time.

Figure 8 shows the approximate latitudes calculated for all reoccurring dropouts. In general the uncertainties in latitude are very small and directly linked to the variability of solar wind velocity within the dropouts (faster or slower velocities will result in shorter or longer calculated rotation periods and thereby different final latitudes). Although our method is unable to distinguish between north and south solar hemispheres, we find that most reoccurring dropouts seem to originate from a band between 20^o and 40^o latitude, around the same latitudes as low latitude coronal holes and helmet streamer loops (Belik et al. 2004). It is unclear, however, why very few dropouts are observed closer to the equator. This could be due to the limitations and approximate nature of our calculation method, an observational bias of ACE itself, or an actual effect worthy of more study. One possible explanation is that even when the top of a large loop or streamer lies in the ecliptic plane, the legs are rooted at higher latitudes and thereby rotate at a slower rate than the equatorial.

Zoom In Zoom Out Reset image size

Table 3 gives the solar wind structure correlations for reoccurring dropouts. When compared to the values in Table 2, we find that reoccurring dropouts are slightly (∼3%) more strongly correlated with SIRs but the relative fractions are more or less the same and the difference could be due to uncertainty in structure identifications. The statistics for the later years (2010 and 2011) are too low to make definite conclusions (only 22 events total in those years); however, they do show similar fractions as we found in Table 2. The overall similarity in structure correlations between all dropouts and only the reoccurring events suggest that some of the non-reoccurring dropouts may also have their origins in long-lived coronal loops. In such an event the specific 3D geometry may be unfavorable for ACE to observe another dropout in the following CR. This would imply that the total amount of plasma coming from such structures may be significantly higher than the 5.5% we reported in Section 2.1. As a side note, we looked at the composition of reoccurring dropouts but found no discernable difference from non-reoccurring dropouts.

As we demonstrated in Section 2.2, heavy ion dropouts have composition most similar to plasma found in closed coronal structures such as coronal streamers. We also found that dropouts comprise 5.5% of all solar wind. However, it is unreasonable to assume that all plasma coming from large, closed coronal loops exhibit the same mass-fractionated signatures of heavy ion dropouts. Therefore our observations represent an absolute lower bound for the total amount of solar wind originating in those structures and being released into the heliosphere. The actual quantity is probably much higher than 5.5%. Nevertheless, the observations of heavy ion dropouts, reoccurring ones in particular, may indicate something about the occurrence rate of exceptionally large and quiet coronal loops.

Our evidence so far indicates that heavy ion dropouts originate from large coronal loops that are relatively long-lived and located predominately near the edges of low latitude coronal holes. The next question to consider then is how the plasmas were released from the loops to become part of the solar wind. Wang et al. (2000) describe three means by which closed-field plasmas may escape into the heliosphere.

• 1.  
  Stretching of the loop large distances out into the heliosphere while the footpoints remain rooted in the corona. This is often observed during ICMEs however it requires enormous plasma pressures to trigger the event.
• 2.  
  Pinching off of the loop tip (i.e., the so-called "LASCO blobs").
• 3.  
  Complex, 3D reconnection with adjacent open-field lines whereby plasma leaks out of the loop while leaving it relatively intact.

Given the reoccurrence rates observed, we conclude that (3) is most likely the primary release mechanism for heavy ion dropouts. While it is possible that some of our dropouts are in fact plasmoids or "LASCO blobs", the bursty nature of (2) makes the observation of reoccurring dropouts improbable over more than one or two CRs. Likewise, the extreme loop pressures required and careful filtering we did for ICMEs precludes significant contributions by mechanism (1). Furthermore, in the loop configuration caused by (1) we would expect to see counterstreaming suprathermal electrons which are not well correlated with the dropouts, although we do observe a few very short events in the data.

For the 3D reconnection case above, the height at which reconnection occurs may influence the elemental X/H ratios of a given dropout. Plasma escaping from lower down on the legs of a coronal loop would be expected to have smaller depletions of the heavy elements than plasma released from the top. However, smaller relative depletions could also be caused by the plasma being confined for a shorter time than required for gravitational settling. Additionally, the mechanism causing the FIP effect further complicates the picture. If the FIP effect is indeed caused by processes in the chromosphere as suggested by Laming (2004), then we would expect to see little to no difference in the FIP fractionation of the dropouts released from difference heights. But if the FIP effect higher up the in the corona, plasma from lower on a loop may be less (or perhaps even more) FIP fractionated depending on the process invoked. Differentiating between spatial and temporal effects will require careful theoretical considerations coupled with the modeling of X/H ratios in a coronal loop. Such a comparison is outside the scope of our present work, although we hope to address this topic in a future study.