Differential velocity between solar wind protons and alpha particles in pressure balance structures

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi:10.1029/2003ja010274, 2004 Differential velocity between solar wind protons and alpha particles in pressure balance structures Yohei Yamauchi and Steven T. Suess NASA Marshall Space Flight Center, Huntsville, Alabama, USA John T. Steinberg Los Alamos National Laboratory, Los Alamos, New Mexico, USA Takashi Sakurai National Astronomical Observatory of Japan, Tokyo, Japan Received 2 October 2003; revised 15 January 2004; accepted 23 January 2004; published 11 March 2004. [1] Pressure balance structures (PBSs) are a common high-plasma beta feature in highlatitude, high-speed solar wind. They have been proposed as remnants of coronal plumes. If true, they should reflect the observation that plumes are rooted in unipolar magnetic flux concentrations in the photosphere and are heated as oppositely directed flux is advected into and reconnects with the flux concentration. A minimum variance analysis (MVA) of magnetic discontinuities in PBSs showed there is a larger proportion of tangential discontinuities than in the surrounding high-speed wind, supporting the hypothesis that plasmoids or extended current sheets are formed during reconnection at the base of plumes. To further evaluate the character of magnetic field discontinuities in PBSs, differential streaming between alpha particles and protons is analyzed here for the same sample of PBSs used in the MVA. Alpha particles in high-speed wind generally have a higher radial flow speed than protons. However, if the magnetic field is folded back on itself, as in a large-amplitude Alfvén wave, alpha particles will locally have a radial flow speed less than protons. This characteristic is used here to distinguish between folded back magnetic fields (which would contain rotational discontinuities) and tangential discontinuities using Ulysses high-latitude, high-speed solar wind data. The analysis indicates that almost all reversals in the radial magnetic field in PBSs are folded back field lines. This is found to also be true outside PBSs, supporting existing results for typical high-speed, high-latitude wind. There remains a small number of cases that appear not to be folds in the magnetic field and which may be flux tubes with both ends rooted in the Sun. The distinct difference in MVA results inside and outside PBSs remains unexplained. INDEX TERMS: 2169 Interplanetary Physics: Sources of the solar wind; 2134 Interplanetary Physics: Interplanetary magnetic fields; 2109 Interplanetary Physics: Discontinuities; 7507 Solar Physics, Astrophysics, and Astronomy: Chromosphere; KEYWORDS: solar wind, Ulysses, plumes, pressure balance structures, Alfvén waves Citation: Yamauchi, Y., S. T. Suess, J. T. Steinberg, and T. Sakurai (2004), Differential velocity between solar wind protons and alpha particles in pressure balance structures, J. Geophys. Res., 109,, doi:10.1029/2003ja010274. 1. Introduction [2] Pressure balance structures (PBSs) are intervals in the solar wind in which changes in the plasma and magnetic pressures balance one another while total pressure remains approximately constant. In fact, the solar wind is generally in near pressure balance everywhere except where dynamic interactions dominate, as, for example, at the fronts of corotating interaction regions (CIRs). However, PBSs have been identified as a well-defined class of features in highspeed, high-latitude wind that is associated with high-plasma Copyright 2004 by the American Geophysical Union. 0148-0227/04/2003JA010274 beta (b) intervals lasting several hours. These intervals permeate the high-speed wind and are suggested to be the interplanetary remnants of coronal plumes [McComas et al., 1996]. This is because they have compatible geometric and statistical properties and plumes are a common feature in coronal holes, in particular at solar activity minimum [Thieme et al., 1988, 1990; Velli et al., 1994; Casalbuoni et al., 1999]. Support for the plume-pbs relationship comes from a correlation between fluctuations in b and in helium abundance across PBSs [Reisenfeld et al., 1999]. Since solar wind abundances are fixed in the chromosphere and transition region and since polar coronal holes are source regions of high-speed solar wind, Reisenfeld et al. [1999] concluded that the solar origin for PBSs may be plumes. 1of10

[3] Plumes are rooted in concentrations of magnetic flux at the vertices of supergranules in the photosphere of the Sun [Suess et al., 1998]. Their geometric structure is defined by this flux concentration and other studies have shown that the formation of plumes is related to network activity (reconnection) as oppositely directed magnetic flux is advected up to the magnetic flux concentrations [Wang, 1998; DeForest et al., 1997]. Given this inherently magnetic structure of plumes at the Sun, it is logical to suggest that information on the magnetic structure of PBSs might be helpful in investigating the relationship between PBSs and plumes in more detail. Reconnection at the base of plumes could lead to PBSs containing magnetic features such as plasmoids or current sheets as a consequence of small-scale magnetic ejections. Yamauchi et al. [2002] looked for signs of such features inside PBSs by performing a minimum variance analysis (MVA) [Sonnerup and Cahill, 1967] on magnetic discontinuities in which the radial magnetic field changes sign. They found that tangential discontinuities dominate, in preference to rotational, so that magnetic discontinuities in PBSs have the character of current sheets or plasmoids. This is in contrast to structures in which the magnetic field reverses direction by folding back on itself in magnetic switchbacks, as in large-amplitude Alfvén waves, that would exhibit rotational discontinuities. A similar analysis was performed on discontinuous reversals in the radial magnetic field that did not lie inside high-b PBSs. The two samples displayed significantly different statistical properties, with the non-pbs discontinuities having a much greater probability of being rotational discontinuities such as in magnetic switchbacks. [4] Single-spacecraft MVA of discontinuities is subject to uncertainties [Horbury et al., 2001] so that an independent method of analyzing magnetic discontinuities that circumvents the uncertainties inherent in single spacecraft MVA studies would be an important step to validate the MVA results. One such method is to compare differential streaming of protons and alpha particles (He ++ ) on opposite sides of magnetic discontinuities. The radial flow speed of alpha particles in high-speed wind is typically larger than that of protons [Formisano et al., 1970; Robbins et al., 1970; Bollea et al., 1972; Ogilvie, 1975; Asbridge et al., 1976; Marsch et al., 1982; Neugebauer, 1981]. The difference is usually comparable to but less than the Alfvén speed. This property has been used by Steinberg et al. [1996] to show that the direction of differential flow between protons and alpha particles, V a V p (hereafter V ap ), is well correlated with the local magnetic field vector, B. This happens because any proton-alpha differential streaming must be aligned along the interplanetary magnetic field. In the context of a change in magnetic field direction at a discontinuity, this leads to three (possibly four) cases that might occur, as illustrated in Figure 1. The first of these (A) isa fold or Alfvénic fluctuation, the second (B) is an isolated current sheet that extends back to the Sun, and the third (C) is a loop which closes back to the Sun (or which could be closed back on itself, as in a plasmoid). In A, e r V ap reverses sign when e r B reverses sign, where e r is the unit vector in the radial direction. In B, only B reverses sign. In C, there will either be alpha particles streaming in both directions with respect to protons or there may be no net differential streaming, with the signature also depending Figure 1. (left) Cartoon of three possible configurations across discontinuities in the radial magnetic field. (right) Diagrams of the variation in radial magnetic field strength and alpha particle-proton differential streaming in the radial direction across the three classes of discontinuities. how the data are reduced from the original distribution function. Only in cases A and C does B have a normal component across the reversal. These three possibilities are diagrammed in the three panels on the right in Figure 1. This is discussed here in terms of e r V ap and e r B, even though the phenomenon is defined in terms of reversals in the vector quantity B. This is because we wish to compare our results with those of Yamauchi et al. [2002], where e r V ap and e r B were used. The reason this restriction seems successful here is discussed in section 2.2. [5] Differential streaming of alpha particles and protons thus presents a way of analyzing the magnetic topology in the vicinity of the short-duration magnetic field reversals observed inside PBSs. It is an alternative to 100 ev electron pitch angle distributions that have often been used to analyze the topology in magnetic clouds [Gosling et al., 1987]. Electron pitch angle distributions are less useful for PBSs because these data from Ulysses, the spacecraft that is used for this analysis, have a low time cadence due to the distance between Earth and Ulysses and the distance of Ulysses from the Sun. Ulysses is the only spacecraft sampling high heliographic latitude, high-speed wind. In particular, Ulysses passed over the south and north poles of the Sun in 1994 and 1995, respectively, when solar activity was minimum so that fast and nearly constant solar wind was coming from the polar regions [Phillips et al., 1995]. For these reasons, Ulysses has been used for detailed studies of PBSs and other high-latitude fine structures such as microstreams [Neugebauer et al., 1995]. The only other such studies are those using low-latitude Helios data. 2. Observations and Analysis 2.1. Identification of PBSs [6] The criteria for identifying PBSs are listed by Yamauchi et al. [2002]. For an isolated structure, they can be summarized as: (1) approximate local pressure balance across the structure, (2) duration of at least 6 hours, and (3) plasma b b avg + s b, where b avg is the average plasma b in highspeed wind, computed using the electron + proton + alpha particle pressure, normalized along with the magnetic field strength to 1 AU [McComas et al., 1996]. Here s b is the rms standard deviation of b avg. Yamauchi et al. [2002] 2of10

Figure 2. Plasma and magnetic field parameters for intervals on (a) 6 8 June 1994 and (b) 9 11 August 1994. These intervals are at the time Ulysses was at (Figure 2a) 67 latitude and a heliocentric distance of 2.9 AU and (Figure 2b) 77 latitude and a heliocentric distance of 2.5 AU. The individual panels show the solar wind velocity, the radial magnetic field (B r ) in RTN coordinates, the pressures (total: solid, thermal: dashed, and magnetic: dotted), and the plasma b normalized to 1 AU, respectively. With the plasma b is also shown the average value, b avg of the normalized quantity over the time Ulysses was south of 50 as a dash-dotted line and b avg + s b /2 as a dotted line. The PBSs and non-pbss are defined by the shading and heading at the top of Figures 2a and 2b. identified 53 PBSs in the high-speed wind from the southern polar coronal hole in 1994 using these criteria, along with a comparison sample of 70 non-pbss that were selected under similar criteria, except for requirement 3 on b. All of these PBSs had the distinction of containing at least one interval of reversal in B r. This is not a necessary condition for the identification of a PBS, but associating PBSs with plumes suggests that every PBS will contain at 3of10

Figure 3. Top two panels show time variation of B r and V ap_r (solid lines) and the correlation between the two, computed as described in the text (dotted line). Bottom two panels show altitude (q) and azimuth (f) of the magnetic field B (solid line) and proton-alpha differential velocity V ap ( and +) in the PBS observed at 0022-0903 (UT) on 9 June 1994. The two dash-dotted vertical lines locate where the MVA was applied at a local reversal in B r by Yamauchi et al. [2002]. The times outside the PBS are lightly shaded in gray. least one pair of field reversals, even though a spacecraft passing through the PBS may not necessarily encounter it [Yamauchi et al., 2002]. For PBS identification, Ulysses 1-hour averaged plasma and magnetic field data from Solar Wind Observations Over the Poles of the Sun (SWOOPS) [Bame et al., 1992] and Vector Helium Magnetometer/ Fluxgate Magnetometer (VHM/FGM) [Balogh et al., 1992], respectively, were used. The data were taken in the south polar regions above 50 southern latitude from 18 January 1994 to 21 December 1994. The heliocentric distance of Ulysses in the period was 3.73 1.63 AU. There is no a priori reason to expect that PBSs differ in the north; thus no northern hemisphere results are described here. [7] Figure 2 shows examples of PBSs and non-pbss in the 1-hour average data. Two examples are contained in the time intervals (a) 6 8 June 1994 and (b) 9 11 August 1994. The PBS and non-pbs are indicated by the shaded areas. The individual panels show the solar wind velocity, the radial magnetic field strength (B r ), the pressures (total (solid), thermal (dashed), and magnetic (dotted)), and b normalized to 1AU, respectively. In the plot of b, the dashdotted and dotted lines are b avg and b avg + s b /2, respectively. From this figure, it is obvious that isolated intervals containing either high b or reversals in B r are not uncommon. However, intervals containing all of the characteristics of PBSs constitute a finite, well-defined category, with high b intervals being identifiable with little ambiguity. 2.2. Analysis [8] To analyze differential streaming in the vicinity of B r reversals, the highest resolution plasma data were used, which were normally 4-min cadence, although occasionally only 8-min cadence data were available. More than 90% of the plasma data were of 4-min cadence. One-min average magnetometer data were smoothed with a simple moving boxcar average onto the same temporal grid as the plasma data so that cross-correlations between the plasma and field could be computed. Alpha particle flow vectors were measured by the SWOOPS instrument at the same cadence as the other plasma data. This completes the data set necessary for the analysis. [9] A typical example of the high correlation that normally exists between the two vector quantities V ap and B is shown in the bottom two panels of Figure 3, which display the time variation of the altitude (q) and azimuth (f) ofb (solid lines) and V ap (+ and ). The example, on 9 June 4of10

Figure 4. Two examples of a linear fit and correlation computed in a 1-hour window in the PBS displayed in Figure 3. (left) Here t = 0120 UT, showing a low correlation and poor linear fit. (right) Here t = 0231 UT, a high correlation and good linear fit. 1994, includes a PBS in the interval 0022 0930 (UT). V ap has been plotted because the magnetic field in the southern hemisphere was sunward at this time in the 22 year solar Hale magnetic cycle. The magnetic field vector and the differential alpha-proton streaming tracked each other extremely well, both during continuous changes and across discontinuities. The strong correspondence between these angles for the magnetic field and the differential streaming is remarkable in that these angles for the plasma are not always strictly well determined from the plasma moments; the formal uncertainty can be large. These results suggest that the angles may often be better determined than expected from statistics based on the plasma moments. [10] The MVA study used reversals in the scalar quantity B r. However, because the average Parker spiral angle during the southern polar pass of Ulysses above 50 S is30 ± 13, jb f j < jb r j. This means that fluctuations in the magnetic field will be less likely to destroy a correlation between B r and V ap_r e r V ap than when the spiral angle is >45. The dispersion in the magnetic field direction increases in the higher resolution data so this is an important consideration. In spite of this qualification, the results in Figure 3 imply limitation to B r is not a severe difficulty. Nevertheless, we examine the full vector relationship in any ambiguous cases. [11] The top two panels in Figure 3 show the time variation of the radial magnetic field, B r, V ap_r, and the correlation, C BV (dotted line) between the two quantities. The correlation was computed using a moving 1-hour window and has been utilized here as a quantitative measure of the overall relation between B r and V ap_r and as a tool for identifying how well the two quantities track each other in any given PBS or non-pbs. The plotted value was taken from the center of the moving window. The correlation is very close to 1 near the time when the MVA was applied at the discontinuity in this example (at the dash-dotted vertical lines). The MVA indicated that this was a tangential discontinuity (TD). However, the good correlation between B r and V ap_r across the discontinuity in B r and the interpretation shown in Figure 1 indicates that this must be a magnetic switchback or (equivalently) a rotational discontinuity (RD). Therefore this is an example of contradictory results between the MVA and the differential streaming of protons and a-particles. The f- and q-angles in the bottom panels confirm the interpretation made using only the scalars B r and V ap_r. Potential reasons for this disagreement are discussed in section 4. [12] Looking at the correlation at other times in Figure 3, there are intervals when it fell to very small values, such as near the time 0100 UT. Besides the statement that the correlation was simply poor at that time, there are at least two physical explanations for why this might have happened. The first is that if the radial magnetic field was nearly zero while the total field intensity remained finite, then the differential streaming might have been primarily nonradial and the correlation based on the radial components of the two vectors might have fallen to small values. That would be an example in which it is necessary to examine the vector relationship between V ap and B. The second reason, and the explanation for the low correlation at 0100 UT in Figure 3, has to do with the finite width of the window used in computing the correlation and a situation in which both V ap_r and B r were approximately constant for an interval at least as long as the width of the window. For a 1-hour window over which the variables are constant except for small, random fluctuations, the correlation will fall to zero. This is illustrated in Figure 4, which shows a linear fit to the data used for the correlations in Figure 3 for two 1-hour windows centered at 0120 UT and 0231 UT. The first is at the time of low correlation and the data had small, nearly random scatter. The resulting straight line fit in the left panel is an extremely poor fit, with a correlation of just 0.1. Conversely, when the variables change across the window, as they did in the data centered at 0231 UT in the right panel, the resulting straight line fit is excellent, with a correlation of 0.92. A wider window would avoid this problem, but would also limit the utility of the correlation for finding anomalous short intervals that potentially con- 5of10

Figure 5. B r and V ap_r (solid lines) and the correlation between the two (dotted line) for a non-pbs at 0327-0950 (UT) on 6 December 1994. The times outside the non-pbs are lightly shaded in gray. tain plasmoids, loops which close back to the Sun, and/or true TDs. [13] For comparison with the PBS in Figure 3, a non-pbs interval observed at 0327-0950 UT on 6 December 1994 is shown in Figure 5. A discontinuity at 0525 UT was identified as an RD in the MVA. The correlation is again almost 1 in the entire interval so the MVA result is in agreement with the differential streaming result for this case, in contrast to the example in Figure 3 where the MVA failed. Overall, the correlation here is similar to that shown in Figure 3 in a PBS and there is no reason to suggest a significant difference between PBSs and non-pbss based on these results. 3. Results [14] To evaluate, first, the overall relationship between the radial magnetic field and radial differential streaming in all the identified PBSs, the average correlation, hc VB i, was computed for intervals in the 53 PBSs lying between 1 hour before and 1 hour after the time at which the MVA was applied. The result is that hc VB i = 0.84 ± 0.17. This is the average of values in the 1-hour window which, at most, was computed with 15 data points or 13 degrees of freedom. For a sample of this size, any correlation above 0.5 is statistically significant at the 99% level. The rms deviation (±0.17) was computed from the variations within the set of correlation coefficients and does not depend on the number of degrees of freedom for each individual coefficient. The conclusion that can be derived from this is that local reversals in B r in PBSs are commonly due to kinks in the magnetic field, magnetic switchbacks or large amplitude Alfvénic fluctuations containing RDs and not plasmoids or current sheets. This is a statistical statement about the most common result and does not say anything about specific individual cases which might deviate from the norm. [15] In the comparison sample outside PBSs, the average correlation in the vicinity of radial magnetic field reversals where an MVA was carried out in 70 non-pbss, derived in the same manner as for PBSs, is hc VB i = 0.86 ± 0.13. This leads to the same conclusion as above, that reversals in B r in non-pbss where the MVA was applied are also commonly caused by magnetic switchbacks. It is entirely consistent with the earlier conclusions of Balogh et al. [1999], based on the propagation direction of waves during magnetic polarity inversions (reversals in B). The number of degrees of freedom is the same as for PBSs, and again this is a statistical statement about the most common case and does not say anything about specific individual cases which might deviate from the norm. [16] These overall statistical results confirm that the typical situation in high-latitude, high-speed solar wind when there is a reversal in the radial component of the magnetic field is a magnetic switchback. The magnetic field is locally folded back on itself but the overall field is continuous. In typical high-speed wind these are Alfvénic fluctuations [e.g., Balogh et al., 1999; Belcher and Davis, 1971], but we retain the switchback nomenclature since differential streaming is independent of Alfvénicity. This provides the basis to search for times which differ from this typical case. To do this, the entire set of PBSs and non-pbss was scanned for times when the correlation was distinctly positive, opposite to what is expected for the southern polar coronal hole at this time in the Hale magnetic cycle. The result of this search is that only at one time was there a significant local reversal in the sign of the correlation at the time of a reversal in B r. Furthermore, the correlation reversal, although in a PBS, took place at a different time, around a different local reversal in B r, than when the MVA was performed in Yamauchi et al. [2002]. [17] The time and PBS exhibiting a reversal in sign of the correlation is shown in Figure 6, where B r and V ap_r are plotted in a PBS that was observed at 1349-2230 UT on 23 September 1994. An MVA had been carried out on a local reversal in B r at 1930 UT by Yamauchi et al. [2002] and that analysis identified the discontinuity as an ED (that is, it could be either a TD or an RD). The correlation between B r and V ap_r was large and negative throughout that time, showing that it was actually an RD. However, within the same PBS, at a different and rather brief local 6of10

Figure 6. (top) B r and V ap_r (solid lines) and the correlation between the two (dotted line) in a PBS observed at 1350 2230 on 23 September 1994. (bottom) The f and q angles for B (solid line) and V ap in the PBS. The times outside the PBS are lightly shaded in gray. reversal in B r at 1500 UT, the correlation became positive for almost an hour and was greater than 0.5 for about 0.5 hours. This is a statistically significant result. The polar angles q and f of V ap and B, plotted in the lower two panels of Figure 6, confirm that there was a true positive correlation. What are shown here are two data points near 1450 UT for which the q and f angles of the magnetic field and plasma distinctly differed by 90. This appears to be a case where the correlation did reverse sign. The smoothness of the change in the correlation is an illusion caused by the width of the correlation window. This could be an instance of an isolated current sheet such as shown as case B in Figure 1. If so, this is a rare phenomenon. [18] A much more common condition is that C VB 0for 1 hour, similar to the case illustrated in Figure 3 near 0100 UT on 9 June 1994. However, in contrast to that case, not all the time periods with C VB 0 can be dismissed by one of the two explanations given above. Such an anomalous case is shown in Figure 7, for a PBS on 2 October 1994 at 0535-1157 UT. Yamauchi et al. [2002] conducted an MVA on a local B r reversal at 0530 UT, identifying it as a TD. The correlation between B r and V ap_r remained very high at that time, and inspection of the f and q angles shown in the bottom two panels of the figure shows that the vector field and vector differential streaming tracked each other extremely well across the discontinuity and that it must therefore be an RD. However, elsewhere in the PBS at 0800-0900 UT the correlation became 0. Over most of this interval, the radial field, although changing sign four times, was not small. Inspection of the f and q angles in the bottom two panels shows that the f angles of V ap and B failed to follow each other over the first two-thirds of the interval. In particular, they seem to have been 90 briefly out of phase at 0820 UT. Over the interval 0800-0900 UT, V ap_r was mostly relatively small but there was a spike at the time of the phase reversal in the two f angles. [19] One possible explanation for this is that it could be a combination of a local current sheet (case B in Figure 1) and an advected loop (case C in Figure 1). However, it is also possible that the field changed direction in the middle of one or more plasma scans so that the plasma distributions in this case are not necessarily reliable and the poor correlations is thus an instrumental effect. Nevertheless, this example may contradict the paradigm that all reversals in B r in the highlatitude, high-speed solar wind are switchbacks. The number of such anomalies is relatively small and seems to be more-or-less evenly divided between PBSs and non-pbss. Therefore contrary to our initial expectation, it does not provide any support for the hypothesis that PBSs are the remnants of polar plumes. It is impossible to give a quantitative estimate of the rate these anomalous conditions occur. They are less common than PBSs since only a subset 7of10

Figure 7. (top) B r and V ap_r (solid lines) and the correlation between the two (dotted line) in a PBS observed at 0535 1157 (UT) on 2 October 1994. (bottom) The f and q angles for B (solid line) and V ap in the PBS. The times outside the PBS are lightly shaded in gray. of our PBSs contained them. Similarly, only a subset of our non-pbss contained them. There seems to be no consistent relationship between any given reversal in B r and whether one of these anomalies exists. 4. Summary and Discussion [20] The differential streaming velocity between solar wind protons and alpha particles has been examined to identify magnetic structures in 53 PBSs and 70 non-pbss. The data were taken from high-speed solar wind observed by Ulysses at high southern solar latitudes in 1994. The study was limited to the behavior of the differential velocity and magnetic field in the radial direction in order to compare with the results of Yamauchi et al. [2002]. The analysis has been successful with this limitation only because the Parker spiral angle is relatively close to radial. It was found that there is nearly always a strong correlation between the differential streaming in the radial direction, V ap_r, and the radial magnetic field, B r. The average correlation is 0.84 in PBSs and 0.86 in non-pbss, respectively. This correlation is found to be statistically significant when considering the small sample size in the moving window used to compute the correlation. [21] The conclusion is that nearly all intervals when B r reverses sign in the high-latitude, high-speed solar wind are magnetic switchbacks (i.e., case A in Figure 1), since the correlation is close to 1. This result has been reported previously for the overall high-speed wind by Balogh et al. [1999]. What is new here is that this conclusion applies equally well to the subset of high-speed wind identified by large values of b as being PBSs. [22] As Balogh et al. [1999] discussed in reference to the origin of switchbacks, outward propagating Alfvén waves are expected to play an important role. However, Alfvén waves generated simply by random motions of the footpoint of magnetic fields at the Sun do not seem able to lead directly to switchbacks. The duration of the B r reversals is typically tens of minutes to 2 hours, which corresponds to a heliographic longitude span 1 (= 10 4 km) on the surface of the Sun at midlatitudes. This is roughly the width of the magnetic lanes of the network in which the open magnetic field of the polar coronal hole and polar solar wind are rooted. This suggests that the magnetic switchbacks with B r reversal might be formed from network-scale magnetic loops that erupt into the corona and then undergo reconnection with the open field. Recently, Yamauchi et al. [2004] have found that many network-scale magnetic fields in polar coronal holes do erupt as loops, using Ha data from Big Bear Solar Observatory. [23] Minimum variance analysis of B r discontinuities in PBSs had identified the discontinuities as preferentially being 8of10

tangential discontinuities, while they are preferentially rotational discontinuities in non-pbss [Yamauchi et al., 2002]. The difference from the present results remains unexplained. However, in view of the present results, the MVA result cannot be used to conclude that there are plasmoids or expanding loops in PBSs. [24] In addition to computing the correlation between B r and V ap_r, each individual PBS and non-pbs was scanned for anomalous times when the correlation either reversed sign or went to zero. This search was motivated by the ideas illustrated in Figure 1, showing how differential streaming might be affected by the presence of a magnetic fold (A), an isolated current sheet (B), or an expanding loop (C). [25] There were no times when the correlation reversed sign in the vicinity of the same reversal in B r which had been analyzed in the MVA study of Yamauchi et al. [2002]. In scanning the PBSs and non-pbss, a single instance was identified, removed from one of the earlier selected reversals in B r, during which C VB did reverse sign. This case is illustrated in Figure 6 and evaluated in the associated discussion. Since this was an isolated instance, it is entirely possible that it was a statistical quirk. However, it could also be an example of an isolated pair of current sheets, case B in Figure 1. Without additional, independent data or more examples, it is impossible to determine more about this feature. [26] There were also several times when C VB! 0. There are two possible explanations for this (given in section 2.2) that do not invoke any of the phenomena illustrated in Figure 1. Some of the instances of C VB! 0 can be explained in this way. However, there were still several times when C VB!0 that could not be explained either by unusually small values of B r or by intervals with small overall variation, the situation illustrated in Figure 4. These remaining times are possibly magnetic loops being advected by the spacecraft. They do not have the nice reversal in correlation that would occur across a current sheet, and examination of the full vector variations shows that these instances are a true failure of the paradigm that V ap and B follow each other. One possible physical explanation is that there is counterstreaming, as might occur in case C in Figure 1. Another is a failure to accurately measure the plasma distribution in the presence of rapid, large amplitude magnetic field fluctuations. Further study of these instances, and possibly new measurements, will be required to make this determination. [27] The question still remains as to why the MVA results do not agree with those from analysis of proton-alpha differential streaming along the magnetic field. One possible scenario is that the folding of the magnetic field is more fully evolved in PBSs than in non-pbss by the time the solar wind has traveled to Ulysses. This would be the case if the solar wind speed was smaller in polar plumes, as has often been reported [Giordano et al., 2000; Teriaca et al., 2003]. In this case, an MVA might more commonly identify the discontinuities in PBSs defining the boundaries of the fold as TDs than RDs because the component of the magnetic vector normal to the discontinuity would be reduced as the thickness of a fold is reduced. Given the inherent uncertainties in MVAs that have been described by Horbury et al. [2001], another possibility is that some property of PBSs may give bias to the statistics of MVA results in PBSs. Such characteristics might include a spectrum of waves and turbulence that depends on the local value of b. Physically, it is entirely possible that local turbulence characteristics could depend on b. As a matter of fact, Horbury et al. [2001] attributed inaccuracies in MVA results to the presence of fluctuations on the surface defined by discontinuities; such a possibility may also exist. [28] Acknowledgments. This research has been supported by the Ulysses/SWOOPS solar wind plasma instrument team and we would like to thank the Ulysses/VHM/FGM magnetometer team for the use of their data. Y. Yamauchi is a NRC/NAS Resident Research Associate. [29] Shadia Rifai Habbal thanks Bruno Bavassano and Tim S. Horbury for their assistance in evaluating this paper. References Asbridge,J.R.,S.J.Bame,W.C.Feldman,andM.D.Montgomery (1976), Helium and hydrogen velocity difference in the solar wind, J. Geophys. Res., 81, 2719 2727. Balogh, A., T. J. Beek, R. J. Forsyth, P. C. Hedgecock, R. J. Marquedant, E. J. Smith, D. J. Southwood, and B. T. Tsurutani (1992), The magnetic field investigation on the Ulysses mission: Instrumentation and preliminary scientific results, Astron. Astrophys., Suppl. Ser., 92, 221 236. Balogh, A., R. J. Forsyth, E. A. Lucek, and T. S. Forbury (1999), Heliospheric magnetic field polarity inversions at high heliographic latitudes, Geophys. Res. Lett., 26, 631 634. Bame, S. J., D. J. McComas, B. L. Barraclough, J. L. Phillips, K. J. Sofaly, J. C. Chavez, B. E. Goldstein, and R. K. Sakurai (1992), The Ulysses solar wind plasma experiment, Astron. Astrophys., Suppl. Ser., 92, 237 265. Belcher, J. W., and L. Davis Jr. (1971), Large-amplitude Alfvén waves in the interplanetary medium: 2, J. Geophys. Res., 76, 3534 3563. Bollea, D., V. Formisano, P. C. Hedgecock, G. Moreno, and F. Palimotto (1972), Helios 1 helium observations in the solar wind, NASA Spec. Publ., 308, 588. Casalbuoni, S., L. Del Zanna, S. R. Habbal, and M. Velli (1999), Coronal plumes and the expansion of pressure-balanced structures in the fast solar wind, J. Geophys. Res., 104, 9947 9961. DeForest, C. E., J. T. Hoeksema, J. B. Gurman, B. J. Thompson, S. P. Plunkett, R. Howard, R. C. Harrison, and D. M. Hasslerz (1997), Polar plume anatomy: Results of a coordinated observation, Solar Phys., 175, 393 410. Formisano, V., G. Moreno, and F. Palmiotto (1970), a-particle observations in the solar wind, Solar Phys., 15, 479. Giordano, S., E. Antonucci, G. Noci, M. Romoli, and J. L. Kohl (2000), Identification of the coronal sources of the fast solar wind, Astrophys J., 531, L79 L82. Gosling, J. T., D. N. Baker, S. J. Bame, W. C. Feldman, R. D. Zwickl, and E. J. Smith (1987), Bidirectional solar wind electron heat flux events, J. Geophys. Res., 92, 8519. Horbury, T. S., D. Burgess, and M. Fränz (2001), Three spacecraft observations of solar wind discontinuities, Geophys. Res. Lett., 28, 677 680. Marsch, E., K.-H. Muhlhauser, H. Rosenbauer, R. Schwenn, and F. M. Neubauer (1982), Solar wind helium ions: Observations of the Helios solar probes between 0.3 and 1 AU, J. Geophys. Res., 87, 35 51. McComas, D. J., G. W. Hoogeveen, J. T. Gosling, J. L. Phillips, M. Neugebauer, A. Balogh, and R. Forsyth (1996), Ulysses observations of pressure-balance structures in the polar solar wind, Astron. Astrophys., 316, 368 373. Neugebauer, M. (1981), Observations of solar-wind helium, Fund. Cosmic Phys., 7, 131. Neugebauer, M., B. E. Goldstein, D. J. McComas, S. T. Suess, and A. Balogh (1995), Ulysses observations of microstreams in the solar wind from coronal holes, J. Geophys. Res., 100, 23,389 23,395. Ogilvie, K. W. (1975), Differences between the bulk speeds of hydrogen and helium in the solar wind, J. Geophys. Res., 80, 1335 1338. Phillips, J. L., et al. (1995), Ulysses solar wind plasma observations from pole to pole, Geophys. Res. Lett., 21, 3301 3304. Reisenfeld, D. B., D. J. McComas, and J. T. Steinberg (1999), Evidence of a solar origin for pressure balance structures in the high-latitude solar wind, Geophys. Res. Lett., 26, 1805 1808. Robbins, D. E., A. J. Hundhausen, and S. J. Bame (1970), Helium in the solar wind, J. Geophys. Res., 75, 1178 1187. Sonnerup, B. U. Ö., and L. J. Cahill (1967), Magnetopause structure and attitude from Explorer 12 observations, J. Geophys. Res., 72, 171 183. Steinberg, J. T., A. J. Lazarus, K. W. Ogilvie, R. Lepping, and J. Byrnes (1996), Differential flow between solar wind protons and alpha particles: First WIND observations, Geophys. Res. Lett., 23, 1183 1186. 9of10

Suess, S. T., A.-H. Wang, I. Cuseri, G. Poletto, and S. T. Wu (1998), The geometric spreading of coronal plumes and coronal holes, Solar Phys., 180, 231 246. Teriaca, L., G. Poletto, M. Romoli, and D. A. Biesecker (2003), The nascent solar wind: Origin and acceleration, Astrophys J., 588, 566 577. Thieme, K. M., E. Marsch, and R. Schwenn (1988), Relationship between structures in the solar wind and their source regions in the corona, in Proceedings of Sixth International Solar Wind Conference, Tech. Note NCAR/TN-306 + Proc., edited by V. J. Pizzo, T. E. Holzer, and D. G. Sime, pp. 317 321, Natl. Cent. for Atmos. Res., Boulder, Colo. Thieme, K. M., E. Marsch, and R. Schwenn (1990), Special structures in high speed streams as signatures of fine structures in coronal holes, Ann. Geophys., 8, 713 724. Velli, M., S. R. Habbal, and R. Esser (1994), Coronal plumes and fine scale structure in high speed solar wind streams, Space Sci. Rev., 70, 391. Wang, Y.-M. (1998), Network activity and the evaporative formation of polar plumes, Astrophys, 501, L145 L150. Yamauchi, Y., S. T. Suess, and T. Sakurai (2002), Relation between pressure balance structures and polar plumes from Ulysses high latitude observations, Geophys. Res. Lett., 29(10), 1383, doi:10.1029/ 2001GL013820. Yamauchi, Y., R. L. Moore, S. T. Suess, H. Wang, and T. Sakurai (2004), The magnetic structure of Ha macrospicules in solar coronal holes, Astrophys J., in press. T. Sakurai, National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan. (sakurai@corona.mtk.nao.ac.jp) J. T. Steinberg, Los Alamos National Laboratory, Los Alamos, NM 87545, USA. (jsteinberg@lanl.gov) S. T. Suess and Y. Yamauchi, NASA Marshall Space Flight Center, SD 50, Huntsville, AL 35812, USA. (steven.t.suess@nasa.gov; yohei.yamauchi@ msfc.nasa.gov) 10 of 10