Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 We have obtained direct evidence for local magnetic reconnection in the solar wind using solar wind plasma and magnetic field data obtained by the Advanced Composition Explorer (ACE). The prime evidence consists of accelerated ion flow observed within magnetic field reversal regions in the solar wind. Here we report such observations obtained in the interior of an interplanetary coronal mass ejection (ICME) or at the interface between two ICMEs on 23 November 1997 at a time when the magnetic field was stronger than usual. The observed plasma acceleration was consistent with the Walen relationship, which relates changes in flow velocity to density-weighted changes in the magnetic field vector. Pairs of proton beams having comparable densities and counterstreaming relative to one another along the magnetic field at a speed of ∼1.4VA, where VA was the local Alfven speed, were observed near the center of the accelerated flow event. We infer from the observations that quasi-stationary reconnection occurred sunward of the spacecraft and that the accelerated flow occurred within a Petschek-type reconnection exhaust region bounded by Alfven waves and having a cross section width of ∼4 × 105 km as it swept over ACE. The counterstreaming ion beams resulted from solar wind plasma entering the exhaust region from opposite directions along the reconnected magnetic field lines. We have identified a limited number (five) of other accelerated flow events in the ACE data that are remarkably similar to the 23 November 1997 event. All such events identified occurred at thin current sheets associated with moderate to large changes in magnetic field orientation (98°–162°) in plasmas characterized by low proton beta (0.01–0.15) and high Alfven speed (51–204 km/s). They also were all associated with ICMEs.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Magnetic reconnection describes a process in which the frozen-in field condition of magnetohydrodynamics is violated in such a manner that pairs of magnetic field lines merge to produce topological changes in the field. This process generally favors oppositely directed magnetic fields at thin current sheets and plays a central role in many interpretations and models of space, solar, astrophysical, and laboratory plasma phenomena (see, for example, the collections of articles in the works of Hones  and Russell et al. ). One characteristic feature of the reconnection process is the acceleration of plasma away from the reconnection site in a pair of oppositely directed exhaust regions. The relatively frequent observation of such accelerated plasma flows within the Earth's magnetopause current layer, the speeds of which are often quantitatively consistent with reconnection model predictions, have provided perhaps the strongest direct evidence for the reconnection process in space plasmas [e.g., Paschmann et al., 1979, 1986; Sonnerup et al., 1981; Gosling et al., 1982, 1990a, 1991; Phan et al., 1996, 2000]. This direct evidence has been bolstered by the simultaneous observation of a variety of other phenomena, such as time-of-flight effects in the low-latitude boundary layer [e.g., Gosling et al., 1990b] and cusps [e.g., Phillips et al., 1993; Fuselier et al., 2000; Wing et al., 2001; Trattner et al., 2002a, 2002b], convective drift of cold plasma of ionospheric origin in the magnetopause current layer [e.g., Gosling et al., 1990c], reflected plasma particle beams upstream from the magnetopause [e.g., Sonnerup et al., 1981; Fuselier et al., 1991; Gosling et al., 1991; Smith and Rodgers, 1991], and the mixing of magnetosheath and magnetospheric plasma [e.g., Sonnerup et al., 1981; Gosling et al., 1990b], that are consistent with reconnection of solar wind and magnetospheric field lines at Earth's magnetopause. Somewhat similar phenomena to the above have also been observed in the plasma sheet and plasma sheet boundary layer [e.g., Forbes et al., 1981; Onsager et al., 1990; Oieroset et al., 2000, 2004] in the geomagnetic tail and likewise provide strong evidence that field lines on opposite sides of the tail current sheet often reconnect with one another far down the tail. Indeed, such reconnection of field lines in the geomagnetic tail is needed to balance the reconnection of magnetospheric and solar wind field lines at the magnetopause.
 Prolonged (∼1 day) solar wind speed increases are occasionally observed behind some interplanetary coronal mass ejections (ICMEs) in the solar wind near 1 AU. Riley et al.  have suggested that those speed increases may be signatures of reconnection that occurs in the corona at current sheets that presumably form behind coronal mass ejections (CMEs) lifting off from the Sun. In contrast, Moldwin et al. [1995, 2000] have suggested that reconnection might commonly occur far from the Sun in the solar wind at the quasi-stationary heliospheric current sheet that normally encircles the Sun and have presented observations of small-scale magnetic flux ropes in the solar wind to support that suggestion. In addition, McComas et al. [1988, 1994] and McComas  have suggested that reconnection should commonly occur in the solar wind at the interfaces between fast ICMEs and the ambient solar wind ahead in much the same manner as occurs at the Earth's dayside magnetopause. They cite observations of counterstreaming suprathermal electrons and a suprathermal electron dropout ahead of an ICME as evidence supporting that suggestion.
 Despite the above suggestions, to the best of our knowledge, direct evidence for reconnection in the solar wind in the form of accelerated plasma flows that are quantitatively consistent with predictions of quasi-stationary reconnection models have not been reported. Our purpose here is to present and discuss plasma and magnetic field observations of accelerated plasma flow obtained within a field reversal region that was either inside an ICME or at the interface between two ICMEs encountered by the Advanced Composition Explorer (ACE) on 23 November 1997. This accelerated plasma flow had the clear signature of a reconnection exhaust [e.g., Petschek, 1964]. Remarkably similar accelerated flow events have been observed by ACE in association with at least five other ICMEs but not elsewhere in the solar wind and specifically not at the leading edges of ICMEs or at the quasi-stationary heliospheric current sheet.
 ACE was launched 25 August 1997 into a halo orbit about the L1 Lagrange point ∼0.01 AU upstream from Earth in the solar wind. The spacecraft spins at ∼5 revolutions/min about an axis nominally pointed at the Sun. The Los Alamos Solar Wind Electron Proton Alpha Monitor (SWEPAM) on ACE includes separate spherical section electrostatic analyzers for three-dimensional measurements of solar wind ion and electron velocity distribution functions, f(v), [McComas et al., 1998]. In the mode utilized for this study, the ion instrument tracks the solar wind peak and provides complete measurements of the solar wind ion f(v) at 40 energy levels around the count rate peak every 64 s with an energy resolution of ∼5% and an angular resolution of ∼5°. Here we use the ion data in the form of proton and alpha particle integrated moments (densities, flow velocities, and temperatures) and reduced distributions of f(v) in the solar wind frame. Data from the electron analyzer used in the present study were obtained in a mode that provides measurements of the electron f(v) in ten 12% energy pass bands, with logarithmically spaced centers ranging from 73.3 to 1370 eV, at each of 60 spin angles in 64 s and repeated every 128 s. Pitch angle (PA) distributions of the electron f(v) at constant energy in the solar wind frame are obtained by correcting for the spacecraft potential and the measured solar wind velocity and by sorting the data into 9° PA bins using knowledge of the magnetic field direction. The event discussed here occurred prior to the beginning of the validated plasma data set available at the ACE Science Center, but the instrument was operating in the proper mode and we have validated the data for the time period of interest. The Magnetic Field Experiment (MAG) on ACE provides measurements of the magnetic field vector at a cadence of 24 vectors/s [Smith et al., 1998]. In this paper we use 16 s and 64 s averages of the magnetic field vector.
Figure 1 shows a 3-day plot of selected plasma and magnetic field data obtained by ACE on 22–24 November 1997 that provides overall context for the more detailed observations in later figures. A strong shock swept over ACE shortly after 0900 UT on 22 November, as is evident from the sharp increases in proton density, temperature, flow speed, and magnetic field strength at that time. The shock was driven by a moderately fast ICME that ACE first encountered at ∼1845 UT on 22 November. The leading edge of the ICME was distinguished by changes characteristic of many ICMEs observed near 1 AU [e.g., Gosling, 1990; Neugebauer and Goldstein, 1997], including the onset of a counterstreaming suprathermal electron beam at PA 180°, an increase in helium abundance, a large decrease in proton temperature, a large change in the orientation of the magnetic field, B, and decreases in the amplitude of fluctuations in B and in the flow velocity, V. The ICME leading edge was also characterized by an increase in proton density, a small increase in flow speed, and a large change in flow direction (not shown).
 Of particular interest here is the brief spike in flow speed centered at ∼1228 UT on 23 November and indicated by the red arrows in Figure 1. That speed increase coincided with a large rotation in B, a brief decrease in the magnitude of B, brief increases in proton density and temperature, an intensification and broadening of the suprathermal electron population at PA 180°, and short-lived changes in flow direction (not shown). This brief event may have occurred at the rear boundary of the ICME, since the above changes were followed by a relatively long-term recovery in proton temperature and long-term decreases in proton density, helium abundance, and flow speed. On the other hand, the rear boundary of the ICME may have come later. We note, for example, that (1) counterstreaming electrons, which are generally indicative of closed field lines within ICMEs [e.g., Gosling et al., 1987], persisted well into 24 November, with a brief 1-hour interruption centered on ∼0830 UT on 23 November; (2) the overall field strength remained high until at least 1830 UT on 23 November; and (3) a second long interval of anomalously low proton temperatures occurred throughout most of 24 November. This complexity suggests the possibility that the overall series of modulations evident in Figure 1 may have been associated with more than one ICME. Thus the spike in flow speed centered at ∼1228 UT on 23 November 1997 occurred either at the interface between two ICMEs or within the interior of a large and relatively complex ICME.
Figures 2 and 3 provide expanded views of the accelerated flow event and include some parameters not shown in Figure 1. It is evident in Figure 2 that the flow speed, proton density, and field strength were all higher before than after the event and that the opposite was true for the proton temperature, proton beta, and Alfven speed. From Figure 1 we note that the alpha particle/proton abundance ratio was also higher before than after the event. Thus the accelerated flow occurred at the interface between two quite different plasma/field states, although both states apparently occurred on closed field lines. During the accelerated flow the proton density, flow speed, temperature, temperature anisotropy, and proton beta were all enhanced, whereas the field strength and Alfven speed were slightly depressed. We note, however, that the flow speed enhancement began ∼3 min before the proton density started to rise and ended ∼5 min after the proton density had decreased. Further, throughout the 1-hour interval shown in Figure 2 the field strength and Alfven speed were higher than normal and the proton beta was quite low, as is often the case within ICMEs.
Figure 3 illustrates that (1) the changes in proton and alpha particle velocity during the accelerated flow event were approximately the same; (2) the accelerated flow event straddled a large (149°) change in the orientation of B; (3) B had a roughly constant orientation in the center of the event intermediate between that prevailing before and after; and (4) the changes in V and B were anticorrelated in the leading portion and correlated in the trailing portion of the event. Since the radial component of the field, Br, was positive before the event and negative after, the latter observation indicates that the accelerated flow event was bounded by Alfven waves propagating antisunward along B, an assertion we will provide more evidence for later in this paper.
 The apparent high proton temperatures and temperature anisotropies within the central portion of the accelerated flow event did not arise from simple thermal distributions. Rather, as illustrated in Figure 4, those temperatures and anisotropies were associated with a pair of relatively cold proton beams of comparable density streaming relative to one another along B. The relative streaming speed was ∼140 km/s and thus was about ∼1.4 times the local Alfven speed, VA. The thermal widths of the individual beams were comparable to those of the cold and relatively warm solar wind distributions observed immediately before and after the accelerated flow event, respectively, as can be ascertained in Figure 4 by comparing the individual beam widths in the accelerated flow region with those obtained before and after. When both beams were present, the overall proton density (Figure 2) was enhanced. On the other hand, double ion beams of this nature were not present near the beginning and end of the accelerated flow event where the overall proton densities were comparable to those of the adjacent solar wind flows.
4. Analysis and Interpretation
 Our working hypothesis is that the 23 November 1997 accelerated flow event observed by ACE was the result of quasi-stationary magnetic reconnection sunward of the spacecraft. Figure 5 provides an idealized sketch of the reconnection exhaust region on the antisunward side of a reconnection site in the solar wind that initially lies sunward of the spacecraft. The geometry sketched is roughly similar to that of the 23 November 1997 event. The exhaust region is bounded on either side by kinks in the magnetic field that propagate as Alfven waves in opposite directions along the reconnected field lines. Plasma enters into the exhaust region from both sides and is accelerated away from the reconnection site as it encounters the field line kinks. Since Alfven waves propagating parallel (antiparallel) to B produce anticorrelated (correlated) variations in B and V, one expects to observe anticorrelated changes in V and B along A1 in Figure 5 and correlated variations in V and B along A2. That is in fact what was observed at the front and back edges, respectively, of the 23 November 1997 accelerated flow event (see Figure 3).
 In a frame of reference moving with one or the other of the field line kinks, the plasma inflow is field-aligned at the local Alfven speed since in the solar wind rest frame the kinks propagate along B at that speed. The transformation velocities from the spacecraft rest frame to the kink rest frames are the deHoffman-Teller (HT) velocities [e.g., Sonnerup et al., 1990], which are different for the two kinks. Figure 6 illustrates the different frame transformations and shows as dotted arrows the inflow velocities in the kink rest frames. Kinetic energy is conserved in those frames, since the convection electric field vanishes there; consequently, the outflows in the kink frames are also field-aligned at the local Alfven speeds. The net result in the reconnection exhaust region is a pair of solar wind ion beams that initially stream along B relative to one another at a speed equal to VA1 + VA2, where VA1 and VA2 are the Alfven speeds at the boundaries of the exhaust region. In the spacecraft frame one should detect both an increase in overall flow speed and a deflection of the flow as well as interpenetrating ion beams of comparable density within the central portion of the exhaust region. This is, of course, essentially what was observed in the 23 November 1997 event (Figures 2, 3, and 4), although the relative beam streaming speed along B was only ∼140 km/s, i.e., somewhat less than the sum of the Alfven speeds (341 km/s) at the edges of the region. The latter result suggests that the relative streaming of the interpenetrating beams was limited by the electromagnetic ion beam instability which, for comparable beam densities, should limit the relative streaming speed to ∼1.5 VA [Goldstein et al., 2000], roughly as observed.
 We can make our comparison between the observations and the “model” more quantitative by investigating how well the observed velocity changes agree with the so-called Walen condition for Alfven waves. From the constancy of the tangential electric field across any discontinuity and from conservation of mass and tangential momentum, one arrives at an equation for the vector change in velocity (ΔV) experienced by plasma passing through an Alfven wave [e.g., Sonnerup et al., 1981]:
Here the subscripts 1 and 2 refer to points up and downstream from the wave, respectively, the + and − signs correspond to waves propagating antiparallel and parallel to B, respectively, ρ is the mass density, μo is the permeability of free space, and α is the pressure anisotropy factor defined by
where p∥ and p⊥ are the plasma pressures parallel and perpendicular to B. We note that the ΔV experienced should be independent of the mass/charge ratio of the individual particles and thus should be the same for protons and alpha particles, essentially as was observed in the 23 November 1997 event (Figure 3).
Figure 7 provides a comparison of measured values of the overall proton ΔV with values calculated using (1) and assuming an isotropic plasma upstream. The starting point for values of ΔV to the left of the vertical line in Figure 7 was prior to the accelerated flow event at 1220 UT, whereas the starting point for values to the right of the vertical line was after the accelerated flow event at 1237 UT. The left and right portions of Figure 7 thus relate to plasma flowing across the exhaust boundaries A1 and A2 in Figure 5, respectively. Our choice of where to divide the data was arbitrarily made so as to provide the best overall agreement between the observed and calculated values of ΔV, but it is a reasonable choice and the agreement between observed and calculated values is quite good. We conclude that the exhaust region was, in fact, bounded on its leading and trailing edges by Alfven waves propagating parallel and antiparallel to B, respectively.
 In the previous sections of this paper we have presented direct evidence that ACE passed through an exhaust region from a quasi-stationary (in time) reconnection site that initially was sunward of the spacecraft and convecting in the solar wind flow. The flow within the exhaust was accelerated relative to that of the surrounding solar wind and had a higher proton density and higher (apparent) proton temperature, a weaker magnetic field, and an intermediate field orientation. The plasma and field states on opposite sides of the exhaust region were distinctly different, although both were characterized by low proton beta (<0.07) and high Alfven speed (>150 km/s). The reconnection site was either within an ICME or at the interface between two ICMEs and passed to the west (i.e., in the direction of positive t in an r, t, n coordinate system) of the spacecraft. From the ∼16-min duration of the accelerated flow and field reversal region and assuming that the region was tilted 45° relative to the radial direction, we estimate that the exhaust region had a cross-section width of ∼4 × 105 km at the position of ACE. The exhaust region was bounded on either side by kinks in the reconnected field lines that propagated as Alfven waves in opposite directions along B. Solar wind ions entering the exhaust region from opposite directions along B interpenetrated one another in the central portion of the exhaust region. The interpenetrating ion beams were of nearly equal density and were streaming relative to one another at ∼1.4VA. The presence of these interpenetrating beams, whose relative streaming was probably limited by the electromagnetic ion beam instability, provides some of the strongest evidence for the above description of the exhaust region.
 The reconnection exhaust we have identified in the solar wind was “Petschek-like” [Petschek, 1964] and resembled that for asymmetric reconnection first described by Levy et al. . In the fluid model outlined by Levy et al., which pertains to cases where the plasma and field conditions on opposite sides of the exhaust are distinctly different, the plasmas entering into the exhaust region from opposite directions along B do not interpenetrate and are separated near the middle of the exhaust region by a contact surface. In that case the transitions from outside to inside the exhaust region are accomplished by rotational discontinuities (Alfven waves) followed by slow mode shocks. We have had difficulty identifying transitions in the 23 November 1997 exhaust region that we would be willing to call slow mode shocks. Although the increased proton density and “temperature” and decreased magnetic field strength observed in the central portion of the exhaust were qualitatively consistent with the Levy et al. prediction, Figure 4 demonstrates that significant plasma interpenetration occurred and that the resulting ion distributions in the central portion of the exhaust were only partially thermalized. Such interpenetration does not occur in fluid theory and is probably why we have been unable to identify slow mode shocks in this event. However, interpenetrating and partially thermalized ion beams having relative speeds along B of ∼1VA are apparent in large-scale hybrid simulations of reconnection exhausts under conditions representative of the geomagnetic tail [Krauss-Varban and Omidi, 1995]. The foregoing authors argue that in this situation slow mode shocks are not fully developed. On the other hand, to the best of our knowledge, interpenetrating ion beams have not yet been identified in distant tail observations of reconnection exhausts, in contrast to the present solar wind observations, and a number of authors have presented evidence that slow mode shocks do form during reconnection in the distant tail [e.g., Feldman et al., 1984, 1985; Oieroset et al., 2000, 2004; Eriksson et al., 2004]. More work is required to compare and reconcile the physical nature of reconnection exhausts in the solar wind and in the distant geomagnetic tail.
 The solar wind exhaust region discussed here had characteristics quite similar to that of the Earth's flank magnetopause when reconnection occurs there [Gosling et al., 1986]. For example, at the flank magnetopause (1) the accelerated flow is generally confined to the current layer where B changes from its magnetosheath to its magnetospheric orientation; (2) the field magnitude is generally depressed within the current layer; (3) the plasma within the current layer typically is a mixture of interpenetrating plasmas that have entered from opposite sides of the current layer; and (4) the reconnection process favors intervals of low proton beta on both sides of the magnetopause. On the other hand, in the flank magnetopause events there is a considerably greater disparity in the plasma densities, temperatures, and Alfven speeds on opposite sides of the exhaust layer than is the case in the solar wind event.
 In our analysis we have assumed that ACE sampled the antisunward-directed exhaust region from a quasi-stationary (in time) reconnection site that initially was sunward of the spacecraft and have shown that that assumption fits the observations quite well. The analysis also implicitly assumed that the flow was super-Alfvenic, in agreement with the observation that the Alfven Mach number, MA, of the surrounding flow was somewhat greater than 2. Had ACE instead encountered the sunward-directed exhaust region from a reconnection site sunward of the spacecraft (a reasonable possibility since for MA > 2 the plasmas and fields in both exhaust regions are convected away from the Sun by the solar wind flow), it would have detected a comparable deceleration of the flow within the exhaust region.
 It is worth considering if we should expect that a nearly stationary spacecraft in the solar wind near 1 AU, such as ACE, would ever encounter flow associated with a sunward-directed exhaust region from a reconnection site at a greater heliocentric distance. The answer depends on the Alfven Mach number of the surrounding solar wind flow. If MA < 1 (rarely observed in the solar wind at 1 AU), the reconnected field lines and the plasma in the sunward-directed exhaust propagate back toward the Sun against the solar wind flow and the exhaust flow velocity, VE, would be directed sunward with VA < VE < 2 VA. If 1 < MA < 2 (also relatively rare), the field lines in the sunward-directed exhaust are convected anti-sunward, but the flow in the exhaust still propagates sunward with VE < VA (see, for example, the discussion related to Figure 9 in the work of Gosling et al. ). In both of the above cases, however, the SWEPAM experiment on ACE would not actually detect the exhaust flow since the ion instrument does not view in the antisunward direction. Finally, if MA > 2 (the usual case in the solar wind at 1 AU), the exhaust flow would be directed anti-sunward and thus would not be detected by a spacecraft at 1 AU.
 The significance of the measurements reported here is that (1) they demonstrate rather conclusively that quasi-stationary magnetic reconnection does at least occasionally occur in the solar wind near 1 AU; and (2) they optimally reveal the physical nature of a Petschek-type reconnection exhaust under conditions not previously explored observationally. On the other hand, we have closely examined ∼6.5 years of ACE plasma and magnetic field data and have been able to identify only five additional events of this nature. All of these additional reconnection exhaust events were associated with ICMEs and all were remarkably similar to the 23 November 1997 event. In particular, they all (1) were characterized by accelerated or decelerated plasma flow with the alpha particles and protons experiencing essentially the same changes in velocity; (2) had higher proton densities, higher (apparent) proton temperatures, weaker magnetic field strengths, and intermediate field orientations as compared with the surrounding solar wind; (3) had comparable cross-sectional widths (between about 2 and 4 × 105 km); (4) were bounded by Alfven waves propagating in opposite directions along B on reconnected field lines; (5) occurred at relatively abrupt shears in B separating solar wind characterized by low proton beta and high Alfven speed yet having distinctly different plasma (density, temperature, and composition) and magnetic field states; and (6) were associated with reconnection sites that initially were sunward of the spacecraft.
 Interpenetrating proton beams of comparable density were detected within three of these additional five exhaust regions and possibly within one of the other exhausts as well. Where temporally resolved thermal electron measurements were available, electron temperatures within the exhaust regions were intermediate between those observed on opposite sides of the exhausts. In particular, the electrons were not heated as they entered into the exhaust regions; this constitutes further evidence for the lack of slow mode shocks in these events. In one of these five events (7 July 1998) ACE encountered the sunward-directed exhaust region from a reconnection site that initially was sunward of the spacecraft. In that case a substantial decrease in flow speed was observed as ACE encountered the exhaust region and the Alfven waves bounding the exhaust region were propagating sunward in the solar wind frame, as expected. In two other events (30 December 1997 and 13 December 1999) the exhaust regions were oriented nearly transverse to the radial direction and ACE detected substantial changes in flow direction but not in overall flow speed. Notably, none of these events occurred at the interface between the ICMEs and ambient (i.e., non-ICME) plasma ahead. Nor have we been able to identify events of this nature at the quasi-stationary heliospheric current sheet that encircles the Sun, another suggested site of reconnection in the solar wind. This suggests that quasi-stationary magnetic reconnection occurs relatively infrequently in the solar wind near 1 AU and only under certain special conditions.
Table 1 documents when these various reconnection exhaust regions were encountered by ACE, their start and stop times, the total change in magnetic field orientation across the events, the Alfven speeds, Alfven Mach numbers and proton betas of the external flows, the presence or absence of interpenetrating proton beams within the exhaust regions, the field topologies on opposite sides of the exhausts as inferred from suprathermal electron measurements, and the spatial widths of the events. We note that in two of these events (7 July 1998 and 3 October 2000) the reconnection was between closed and open magnetic field lines; in the other four events the reconnection was entirely between closed field lines. It is of particular interest that the observed field orientation changes in this set of events ranged between 98° and 162°. Although the orientation changes measured at ACE were not necessarily the same as those at the actual reconnection sites, these observed changes in field orientation do suggest that field lines need not be nearly antiparallel for reconnection to occur in the solar wind (that also appears to be the case at the dayside magnetopause but not at the flank magnetopause [Gosling et al., 1990a, 1991]). For quasi-stationary reconnection to occur it appears to be more important that there be a thin current sheet separating moderate to large changes in field orientation in plasmas characterized by low proton beta and high Alfven speed. In the solar wind these conditions occur most frequently within ICMEs and seldom, if ever, occur at the global heliospheric current sheet.
Table 1. Solar Wind Reconnection Exhausts Encountered by Advanced Composition Explorer
Field Angle Change
4 × 105
4 × 105
2 × 105
3 × 105
4 × 105
3 × 105
 Reconnection converts magnetic field energy to bulk flow energy and alters the topology of the heliospheric magnetic field. However, we wish to note explicitly that local reconnection within an ICME does not reduce the amount of new magnetic flux carried out into the heliosphere by the ICME, although reconnection must occur elsewhere to maintain a rough long-term balance of magnetic flux in the heliosphere [e.g., Gosling, 1975; McComas, 1995; Crooker et al., 2002]. In the case of the events discussed here, the newly reconnected field lines continued to convect outward in the heliosphere since the solar wind flow was super-Alfvenic. In order for reconnection to affect the overall balance of magnetic flux in the solar wind it must occur in regions where the solar wind flow is sub-Alfvenic, which generally implies that it must occur relatively close to the Sun.
 Work at Los Alamos was performed under the auspices of the U.S. Department of Energy with support from NASA as a part of the ACE program and an SR&T grant. Work at Southwest Research Institute and the University of New Hampshire was supported by the NASA/ACE program.
 Shadia Rifai Habbal thanks Terry G Forbes and Dennis K Haggerty for their assistance in evaluating this paper.