The structure and origin of magnetic clouds in the solar wind

Plasma and magnetic field data from the Helios 1/2 spacecraft have been used to investigate the structure of magnetic clouds (MCs) in the inner heliosphere. 46 MCs were identified in the Helios data for the period 1974–1981 between 0.3 and 1 AU. 85% of the MCs were associated with fast-forward interplanetary shock waves, supporting the close association between MCs and SMEs (solar mass ejections). Seven MCs were identified as direct consequences of Helios-directed SMEs, and the passage of MCs agreed with that of interplanetary plasma clouds (IPCs) identified as white-light brightness enhancements in the Helios photometer data. The total (plasma and magnetic field) pressure in MCs was higher and the plasma-β lower than in the surrounding solar wind. Minimum variance analysis (MVA) showed that MCs can best be described as large-scale quasi-cylindrical magnetic flux tubes. The axes of the flux tubes usually had a small inclination to the ecliptic plane, with their azimuthal direction close to the east-west direction. The large-scale flux tube model for MCs was validated by the analysis of multi-spacecraft observations. MCs were observed over a range of up to ∼60° in solar longitude in the ecliptic having the same magnetic configuration. The Helios observations further showed that over-expansion is a common feature of MCs. From a combined study of Helios, Voyager and IMP data we found that the radial diameter of MCs increases between 0.3 and 4.2 AU proportional to the distance, R, from the Sun as R0.8 (R in AU). The density decrease inside MCs was found to be proportional to R−2.4, thus being stronger compared to the average solar wind. Four different magnetic configurations, as expected from the flux-tube concept, for MCs have been observed in situ by the Helios probes. MCs with left-and right-handed magnetic helicity occurred with about equal frequencies during 1974–1981, but surprisingly, the majority (74%) of the MCs had a south to north (SN) rotation of the magnetic field vector relative to the ecliptic. In contrast, an investigation of solar wind data obtained near Earth’s orbit during 1984–1991 showed a preference for NS-clouds. A direct correlation was found between MCs and large quiescent filament disappearances (disparition brusques, DBs). The magnetic configurations of the filaments, as inferred from the orientation of the prominence axis, the polarity of the overlying field lines and the hemispheric helicity pattern observed for filaments, agreed well with the in situ observed magnetic structure of the associated MCs. The results support the model of MCs as large-scale expanding quasi-cylindrical magnetic flux tubes in the solar wind, most likely caused by SMEs associated with eruptions of large quiescent filaments. We suggest that the hemispheric dependence of the magnetic helicity structure observed for solar filaments can explain the preferred orientation of MCs in interplanetary space as well as their solar cycle behavior. However, the white-light features of SMEs and the measured volumes of their interplanetary counterparts suggest that MCs may not simply be just Hα-prominences, but that SMEs likely convect large-scale coronal loops overlying the prominence axis out of the solar atmosphere.

Abstract. Plasma and magnetic ®eld data from the Helios 1/2 spacecraft have been used to investigate the structure of magnetic clouds (MCs) in the inner heliosphere. 46 MCs were identi®ed in the Helios data for the period 1974±1981 between 0.3 and 1 AU. 85% of the MCs were associated with fast-forward interplanetary shock waves, supporting the close association between MCs and SMEs (solar mass ejections). Seven MCs were identi®ed as direct consequences of Helios-directed SMEs, and the passage of MCs agreed with that of interplanetary plasma clouds (IPCs) identi®ed as whitelight brightness enhancements in the Helios photometer data. The total (plasma and magnetic ®eld) pressure in MCs was higher and the plasma-b lower than in the surrounding solar wind. Minimum variance analysis (MVA) showed that MCs can best be described as largescale quasi-cylindrical magnetic¯ux tubes. The axes of the¯ux tubes usually had a small inclination to the ecliptic plane, with their azimuthal direction close to the east-west direction. The large-scale¯ux tube model for MCs was validated by the analysis of multi-spacecraft observations. MCs were observed over a range of up to $ 60 in solar longitude in the ecliptic having the same magnetic con®guration. The Helios observations further showed that over-expansion is a common feature of MCs. From a combined study of Helios, Voyager and IMP data we found that the radial diameter of MCs increases between 0.3 and 4.2 AU proportional to the distance, R, from the Sun as R 0X8 (R in AU). The density decrease inside MCs was found to be proportional to R À2X4 , thus being stronger compared to the average solar wind. Four dierent magnetic con®gurations, as expected from the¯ux-tube concept, for MCs have been observed in situ by the Helios probes. MCs with leftand right-handed magnetic helicity occurred with about equal frequencies during 1974±1981, but surprisingly, the majority (74%) of the MCs had a south to north (SN) rotation of the magnetic ®eld vector relative to the ecliptic. In contrast, an investigation of solar wind data obtained near Earth's orbit during 1984±1991 showed a preference for NS-clouds. A direct correlation was found between MCs and large quiescent ®lament disappearances (disparition brusques, DBs). The magnetic con®gurations of the ®laments, as inferred from the orientation of the prominence axis, the polarity of the overlying ®eld lines and the hemispheric helicity pattern observed for ®laments, agreed well with the in situ observed magnetic structure of the associated MCs. The results support the model of MCs as large-scale expanding quasi-cylindrical magnetic¯ux tubes in the solar wind, most likely caused by SMEs associated with eruptions of large quiescent ®laments. We suggest that the hemispheric dependence of the magnetic helicity structure observed for solar ®laments can explain the preferred orientation of MCs in interplanetary space as well as their solar cycle behavior. However, the whitelight features of SMEs and the measured volumes of their interplanetary counterparts suggest that MCs may not simply be just Ha-prominences, but that SMEs likely convect large-scale coronal loops overlying the prominence axis out of the solar atmosphere.

Introduction
The term magnetic cloud (MC) was introduced by  to characterize the magnetic ®eld and plasma signatures of an interplanetary post-shock ow observed by ®ve spacecraft separated over more than 30 in solar longitude between 0.9 and 2 AU. The outstanding feature of this solar wind transient was the smooth rotation of the magnetic ®eld vector nearly parallel to a plane over a time interval of the order of one day at 1 AU. This peculiar transient¯ow showed extremely low plasma-b values, i.e., the plasma's magnetic pressure f 2 a8p dominated the thermal pressure nk due to its higher than average solar wind magnetic ®eld strength f and its unusually low plasma temperature and density (N). The MC maintained its characteristic magnetic structure during its outward propagation up to at least 2 AU, although it had considerably expanded in its radial direction.
Subsequent analyses of solar wind data from various spacecraft near Earth and in the outer heliosphere revealed that MCs are relatively common phenomena in the solar wind Burlaga and Behannon, 1982;Burlaga, 1991). MCs are often, but not always associated with interplanetary shocks Zhang and Burlaga, 1988). MCs that strike the Earth may cause intense geomagnetic storms (Wilson, 1987;Zhang and Burlaga, 1988;Tsurutani et al., 1992;Bothmer, 1993;Bothmer and Schwenn, 1995).
The search for the solar origin of MCs brought up a close association of MCs with solar mass ejections (SMEs) and erupting prominences (disappearing ®laments) Wilson andHildner, 1984, 1986;Marubashi, 1986;Bothmer and Schwenn, 1994). Note, that in this study we use the more general term solar mass ejection' (SME), rather than`coronal mass ejection' (CME), because the origin and evolution of transient events in the solar atmosphere is not yet known from current white-light coronagraph observations and in situ solar wind measurements. Further, evidence has been found that at times not only coronal but also chromospheric solar material is ejected into interplanetary space (Schwenn et al., 1980). Here we prefer to reserve the term SME for the solar event and we call the interplanetary counterparts of SMEs ejecta, interplanetary plasma clouds (IPCs), or magnetic clouds (MCs) if they exhibit a speci®c magnetic topology. According to previous studies MCs represent just a subset of all IPCs found in the solar wind (Gosling, 1990;Bothmer and Schwenn, 1996). Among all IPCs identi®ed by bidirectional electron (BDE)¯ows, $1/3 had a large coherent internal magnetic ®eld rotation characteristic of a MC (Gosling, 1990(Gosling, , 1993. Bothmer and Schwenn (1996) found from correlated white-light and in situ solar wind measurements that 41% of the driver gases (ejecta, IPCs), caused by SMEs directed towards the Helios spacecraft, were MCs. Goldstein (1983) ®rst suggested that the speci®c magnetic properties of MCs might be explained through force-free, large-scale, cylindrical magnetic¯ux tubes with r Â f af, i.e. the electric currents are¯owing everywhere parallel or antiparallel to the magnetic ®eld lines. Maximum magnetic ®eld strength should be observed at the center, i.e., on the axis of the¯ux tube under ideal force-free conditions. Burlaga (1988), Marubashi (1986) and Lepping et al. (1990) found a good agreement between the in situ observed magnetic ®eld characteristics of MCs and self-consistent magnetohydrodynamic (MHD)-calculations based on the force-free¯ux tube concept. In contrast to the model of cylindrical shaped clouds, Vandas and Fisher (1991) and Vandas et al. (1993) suggested that MCs might rather be explained in terms of spheroidal topologies, and Farrugia et al. (1995) found that the¯ux tube model better describes in in situ ®eld structure of MCs than the spheromak model.
The bidirectional¯ows of supra-thermal electrons along magnetic ®eld lines, as observed inside IPCs, support the assumption of closed ®eld lines, in agreement with observations that these plasma and magnetic ®eld entities in the solar wind stem from regions in the solar atmosphere where the magnetic ®eld lines form closed loops (Gosling et al., 1973;Montgomery et al., 1974;Bame et al., 1981;. It has been proposed that the ejecta from SMEs evolve into largescale magnetic¯ux ropes that remain magnetically connected to the solar atmosphere (e.g. Gosling, 1990;Gosling et al., 1992). Observations of solar energetic particles (SEPs) associated with MCs and IPCs (Kahler and Reames, 1991;Bothmer et al., 1996b), of¯are particles inside a MC (Farrugia et al., 1993) and of BDE-anisotropies  are also in favor of this picture. However, BDEs could be present on closed magnetic ®eld lines disconnected from the Sun as in plasmoids, or on closed ®eld lines that are rooted with its footpoints in the solar corona as in magnetic¯ux ropes. Unfortunately, current spacecraft measurements do not allow us to infer the overall topology of MCs unambiguously.
MCs have hitherto been studied extensively only at near Earth-orbit and beyond. The Helios mission has provided us with solar wind data in the inner heliosphere between 0.3 and 1 AU over a time interval covering more than one solar activity cycle. The purpose of our study is to extend the investigations of MCs to this range of the inner heliosphere. A further scienti®c goal of the work is to clarify the origin of MCs at the Sun, especially to validate the probable association with erupting solar prominences (disappearing ®laments).

Identi®cation of magnetic clouds in the Helios data
For the study presented here we used hourly averaged plasma and magnetic ®eld data of the instruments E1 & E2 onboard the Helios probes. A detailed description of the instruments can be found in the papers by Rosenbauer et al. (1981), Schwenn et al. (1975) and Musmann et al. (1975). The complete set of interplanetary magnetic ®eld (IMF) and plasma data of Helios 1 covers the time interval December 1974 through July 1981. Data of Helios 2 span the time-interval January 1976 through March 1980.
Only a small number of magnetic clouds have been studied previously in the inner heliosphere at distances closer to the Sun than 1 AU (e.g., . Thus, we had to establish appropriate identi®cation criteria applicable to identify MCs between 0.3 and 1 AU. According to Burlaga (1991), MCs can be identi®ed at 1 AU through the following criteria: (1) the magnetic ®eld direction rotates parallel to a plane through a large angle during a time-interval of the order of one day; (2) the magnetic ®eld strength inside a MC is higher than in the average solar wind; and (3) the temperature inside a MC is lower than average. Note, that the rotation of the magnetic ®eld direction can occur in any direction relative to the ecliptic. Previous MC observations at 1 AU have shown that the criteria (2) and (3) imply low plasma-b values for MCs . Each individual criterion might occur independently of a MC, e.g. in association with planar magnetic structures (PMS) (Nakagawa et al., 1989;Neugebauer et al., 1993), co-rotating interaction regions (CIRs) or sector boundaries (SBs) (Bothmer and Schwenn, 1992a;1996). For a unique identi®cation of a MC all three criteria have to be ful®lled simultaneously.
The enormous radial size of MCs ($ 0X25 AU at 1 AU) was ®rst evidence for an expansion of MCs in the heliosphere (Klein and Burgla, 1982). The Voyager spacecraft indeed observed larger radial sizes for MCs in the outer heliosphere (Burlaga and Behannon, 1982). The sizes of MCs observed near Earth's orbit cannot be considered to be typical of those that might be observed closer to the Sun. Further, the spacecraft's trajectory through a MC in¯uences the observed duration time and thus its apparent size. Due to these uncertainties we concentrated our search for MCs in the Helios data on solar wind events associated with the basic property of a MC, i.e., a smooth coherent directional change of the magnetic ®eld vector on time scales of several hours to days. These changes can be identi®ed in the magnetic ®eld components f x , f y , f z which are the Cartesian components of the IMF in solar ecliptic coordinates (SEC), and in the latitudinal and longitudinal excursions of the magnetic ®eld vector, i.e., by variations in the angles h and u. f x points from the spacecraft towards the Sun, f y points in the ecliptic plane normal to f x towards the East, and f z points normal to the ecliptic, towards the ecliptic north pole. Accordingly, h and u are the magnetic ®eld's polar (h 90 NY h À90 S and azimuthal (u 0 sunward direction, u 90 EY u 270 W angles. Based on these criteria we investigated plots of the plasma and magnetic ®eld data of the Helios spacecraft taking into account all events in which variations in the magnetic ®eld direction occurred in h or u and in the cartesian components over time intervals of several hours. We included events with irregular or weak ®eld variations and variations at sector boundaries. All events were then investigated with the minimum variance analysis (MVA) (Sonnerup and Cahill, 1967) described in the Appendix to identify smooth rotations of the magnetic ®eld vector on time scales of the order of several hours. The MVA-method can be accurately applied for directional changes of the magnetic ®eld vector exceeding $ 30 (Burlaga and Behannon, 1982). Thus MCs were identi®ed by a smooth rotation of the ®eld direction in the plane of maximum variance, if the plasma temperature was lower and the magnetic ®eld strength higher than in the surrounding solar wind (see criteria 2 and 3), and ®nally the error criteria of the minimum variance method were satis®ed (k 2 ak 3 ! 2Y k 2 and k 3 correspond to the eigenvalues of the directions of intermediate and minimum variance, see Appendix). Figure 1 shows an example of a MC in the Helios data for a radial distance of 0.4 AU from the Sun. The MC was basically identi®ed from the smooth south to north turning of the magnetic ®eld vector relative to the ecliptic. Note the rotation of h in Fig. 1 covering the time interval marked with dashed lines, and that of the magnetic ®eld vector in the plane of maximum variance. Here the diamond denotes the initial point of the plot and the arrow indicates the direction of the rotation of Top to bottom: solar wind momentum¯ux q p 2 , total plasma pressure 10 À10 dyn/cm 2 , plasma-b p calculated from the proton parameters, magnetic ®eld magnitude fn , polar and azimuthal angles h, u of the magnetic ®eld vector in solar ecliptic coordinates, proton velocity p , proton number density x p , and proton temperature p . The bottom part of the ®gure shows the rotation of the magnetic ®eld vector in the plane of maximum variance for the time interval of the MC. The diamond denotes the initial point of the plot and the arrow indicates the direction of the rotation of the magnetic ®eld the magnetic ®eld. The plasma-b p (calculated from the proton parameters only) was much lower inside the cloud whereas the total plasma pressure nk f 2 a8p was much higher compared with the surrounding¯ow. The solar wind momentum¯ux q 2 remained fairly constant during the time interval shown in Fig. 1. Another interesting feature is the high plasma density compared with the ambient solar wind which may indicate that the low densities often found in MCs may be a result of considerable expansion of MCs with increasing distance from the Sun (see Klein and Burlaga, 1982;Burlaga and Behannon, 1982 and the results presented in Sec. 7). Our search through the Helios data yielded a set of 46 certain MCs for the years 1974±1981 over the distance range 0.3±1 AU ( Table 1). Note that many more cases hampered by data gaps or other disturbances were excluded. It is interesting to remark that generally solar wind structures with unique internal ®eld rotations identi®ed in the inner heliosphere were associated also with low plasma-b values bK0X1. This ®nding agrees with the characteristics of MCs at 1 AU (Burlaga, 1991).

Magnetic clouds, interplanetary shocks and SMEs
We examined the high resolution plasma and IMF data set of Helios 1/2 to investigate the association of the identi®ed MCs with interplanetary shock waves. 39 (85%) of the 46 MCs observed between 0.3±1 AU were associated with fast-forward shocks (see Table 1). The arrival times of the MCs lagged the shocks by 5±25 h with an average delay of $ 11 h. Solar wind momentum ux q 2 often peaked near the``nose'' of the MCs several hours after arrival of the interplanetary shock (e.g., Figs. 5, 6, 7). In agreement with theoretical models for interplanetary shocks (see, e.g., Hundhausen, 1972), MCs seem to be shock drivers in the inner heliosphere. Figure 2 shows a histogram in 100 km/sec bins for the average plasma speeds, (derived from the protons only), measured inside the 46 MCs observed by Helios 1/2. Most MCs were found to have a propagation speed similar to that of typically slow (300±500 km/s) solar wind. Only a few MCs had speeds comparable to the typically high ($700±800 km/s) speed solar wind¯ow from coronal holes. The peak of the speed distribution in Fig. 2 lies in the range 400±500 km/s with an average value of h i 478 (rms=118) km/s. As indicated in Fig. 2 by the open portion of the bars, the MCs that were not associated with shocks were amongst the slowest MCs with speeds typically less than 400 km/s. The ®nding that most of the MCs with speeds beyond 400 km/s were associated with shocks is in agreement with the result obtained from direct correlations of coronagraph observations of SMEs directed towards Helios 1 and in situ shock disturbances at the spacecraft: When Helios 1 was located inside the angular extent of a SME seen in the coronagraph's ®eld of view as viewed from Earth's orbit, it was de®nitely hit by an interplanetary shock as consequence of the SME, if the SME's speed was greater 400 km/s, independent of the space- craft's distance to the Sun (Sheeley et al., 1985). The average speed of MCs shown in Fig. 2 is comparable to the speed distribution derived for SMEs (see Hundhausen et al., 1994) in the sense that the shape of the distribution drops o at speeds beyond 600 km/s, i.e., very fast MCs and SMEs are less frequent. However, the fact that the observed``low speed'' MCs were so well associated with shocks may be indicative for deceleration of the MCs on their way out from the Sun depending on the ambient solar wind¯ow through which they had to plow. This should be especially important during solar maximum when the heliosphere consists of very inhomogeneous solar wind¯ow.
The high percentage of MCs associated with shocks found here is similar to that obtained for MCs observed near Earth orbit during the years 1978±1982 (Zhang and Burlaga, 1988) but higher than that observed during the years 1967±1978 . None of the 46 MCs observed between 0.3±1 AU was associated with a forward-reverse shock pair, which has been observed for interplanetary counterparts of SMEs detected by the Ulysses spacecraft at high heliographic latitudes in fast solar wind coronal hole¯ow (Gosling et al., 1994). Since it has been uniquely shown that fast SMEs at the Sun are the prime sources of transient shocks in the interplanetary medium (Sheeley et al., 1985;Schwenn, 1986;Cane et al., 1987) our ®ndings support a close association of MCs with SMEs.
We compared our list of shock-associated MCs with the SME shock event list derived from correlated Helios-Solwind observations during 1979±1982 published by Sheeley et al. (1985) and found that 7 MCs occurred as direct consequences of SMEs towards Helios 1 (see Table 1). Another study based on this event list showed that 17 out of 25 Helios-directed SMEs which caused an interplanetary shock wave at Helios 1 were associated with driver gases, but only seven of the driver gases were MCs (Bothmer and Schwenn, 1996).
Note that this study included only those events of the original list of Sheeley et al. (1985) in which the presence or absence of a piston had been uniquely identi®ed and for which both plasma and magnetic ®eld measurements had been obtained.
Both observations are plausible since an interplanetary shock should commonly have a broader spatial extent than the associated driver gas (Borrini et al., 1982;Richardson and Cane, 1993) and IPCs might have dierent magnetic structures, as has been pointed out by Bothmer and Schwenn (1996).
The Helios photometers, designed to measure the zodiacal light, mounted at an angle of 16 , 31 and 90 with respect to the spacecraft's spin-axis, have been used as wide-angle coronagraphs in order to identify SMEs by their white-light brightness increases close to the Sun and also as IPCs in the interplanetary medium (Webb and Jackson, 1990). Striking evidence was found that plasma clouds resulting from SMEs did pass the Helios spacecraft. We compared our list of MCs with the list of IPCs detected by the Helios photometers (Jackson et al., 1994). Without considering the duty-cycle of the instruments we ®nd 23 of our MCs (see Table 1) associated with an interplanetary white-light event identi®ed as an IPC, i.e., MCs are intimately connected to IPCs from SMEs (see also Webb et al., 1993). Brightness enhancements in the 16 and 31 photometers usually preceded the peak of the 90 photometer by 1±2 days. The peak in the 90 photometer data is thought to indicate the passage of a plasma cloud right at Helios. We ®nd also that there were IPCs identi®ed from the Helios photometer data that had not the structure of MCs, in agreement with the ®nding that MCs are only a subset of all IPCs (Gosling, 1990;Bothmer and Schwenn, 1996).
4 Magnetic clouds as magnetic¯ux tubes Goldstein (1983) has proposed that the magnetic ®eld characteristics of MCs could be explained through force-free large-scale cylindrical magnetic¯ux tubes. The idealized sketch in Fig. 3 depicts the magnetic ®eld signatures expected for a cylindrical magnetic¯ux tube which moves radially away from the Sun, i.e., in the x-direction, and centrally passes a Helios spacecraft in the ecliptic. Here, the maximum change in the magnetic ®eld direction would occur normal to the ecliptic in the f Ã x -component due to the south (S) to north (N) turning of the magnetic ®eld vector. The component of minimum variance, f Ã z , corresponds to the radial component f x , because f x is zero in the whole MC. At the center of the¯ux tube, i.e., at the cloud's axis, the magnetic ®eld is directed in the azimuthal (+y-) direction [eastern (E)direction]. This is also the direction of intermediate variance, because f Ã y is zero at the cloud's outer boundaries (on the surface of the cylinder). In reality, the¯ux-tube axis may have any orientation relative to the ecliptic plane and to the radial direction i.e., the actual variance directions x Ã , y Ã , z Ã usually dier from the solar ecliptic coordinate system in x, y, z. Whether the model is appropriate to describe the structure of MCs can be inferred from the results of MVA  applied to measured data.
In Fig. 4 the in situ characteristics of a MC associated with an interplanetary shock observed by Helios 1 are displayed together with the results of the variance analysis of the magnetic ®eld data. A smooth rotation of the magnetic ®eld direction from south to north (note the changes of h in Fig. 4) starts in coincidence with a drop to very low b p values (b p is calculated from the proton parameters only) at 01 UT on day 61. After the ®eld vector has reached its most northern dip at 01 UT on day 62 it relaxes back to the average Parker spiral direction of the IMF. Note also the high total internal plasma pressure inside the MC. For shock-associated MCs we found the -values usually to be highest between the shock and the leading edge of the cloud where the plasma turbulence is high. This region is commonly called the cloud's sheath (in analogy to the Earth's magnetosheath, see also McComas et al., 1988). Maximum pressure often occurred right at the nose of the MCs, thus indicating that MCs were indeed the drivers of the associated shocks. The large decrease of the magnetic ®eld strength from the front to the rear portion of the cloud certainly stems from compression of the cloud's leading edge while plowing through slower moving ambient solar wind ahead. The in situ observations show that MCs are certainly not ideal force-free structures in reality.
The boundaries of the cloud have been determined from two parameters: the b p -slope and the time interval of rotation of the magnetic ®eld vector. Note that unique criteria to determine MC-boundaries have not been established in the literature. The size of the MC as calculated from the cloud's average plasma speed and its  Goldstein, 1993) turning of the magnetic ®eld vector is associated with passage of the circular ®eld lines at the cloud's outer boundaries. At the center of the¯ux tube (at the axis) the magnetic ®eld is directed along the f Ã y (east)-direction. A semicircular rotation appears when f Ã y is plotted versus f Ã x (top diagram) Fig. 4. Solar wind parameters (top panel ) and results from the minimum variance analysis (middle panel ) for a MC (dashed lines) observed with Helios 1 at 0.9 AU, day 61, 01 UT ± day 62, 01 UT in 1978. The MC was associated with a shock on day 60, 12 UT. The draping of the IMF in the cloud's sheath region and behind the cloud's rear portion is labelled (h). The calculated orientation of the cloud's axis, sketched in the bottom panel, was u 119 , h 44 X The direction of minimum variance was u 189 , h À20 duration was 0.27 AU, a typical size for MCs observed at 1 AU .
Results of the MVA of the magnetic ®eld data are presented in Fig. 4 in the lower diagrams. During the time interval of the MC, f Ã z was small compared to the magnitude of the total magnetic ®eld fhj f Ã z ji ahfi 0X09. The ratio of the intermediate eigenvalue k 2 to the minimum eigenvalue k 3 was k 2 ak 3 16X8Y and the angle between the ®rst and the last value of the ®eld vector was 150 (see Table 1 which includes the results of the MVA for all 46 MCs). The variance directions were well determined according to the error criteria given by Siscoe and Suey (1972) and Lepping and Behannon (1980). Plotted in the bottom panel in Fig. 4 The hodogram shows that the magnetic ®eld vector rotates smoothly through a large angle in the plane of maximum variance.
The comparison of the in situ observed magnetic characteristics of this MC with that expected from the cylindrical¯ux tube model proposed by Goldstein (1983) is in excellent agreement. The calculated direction of the cloud's axis, i.e., the direction of intermediate variance, was u 119 (azimuthal direction in the ecliptic), h 44 (inclination relative to the ecliptic). In contrast to the in-ecliptic orientation of the¯ux tube axis sketched in Fig. 3, the axis of this MC had a considerable inclination relative to the ecliptic. A sketch of the axis direction at Helios 1 is given in Fig. 4 at the bottom. The calculated minimum variance direction was pointing southward of the ecliptic h À20 , slightly towards the west u 189 , close to the probable direction of propagation, which is radially outwards from the Sun.
Although more sophisticated cloud models have been developed, including also a possible oset of the spacecraft relative to the cloud's axis (Burlaga, 1988;Lepping et al., 1990;Vandas et al., 1993), a MVA of the magnetic ®eld data is appropriate to estimate the cloud's axis orientation to ®rst order.

Con®gurations and orientations of MCs in interplanetary space
The concept of cylindrical¯ux tubes allows dierent magnetic con®gurations. In the context of the investigation of the orientations of MCs in interplanetary space, it will be shown that all the predicted types of ux-tubes actually occur in the solar wind. Assume as in Fig. 3 that the axis of a MC lies in the ecliptic plane, normal to the sun-spacecraft line. A¯ux tube (MC) cloud possess one of the four dierent magnetic con®gurations presented in Table 2, which dier in the orientation of the magnetic ®eld lines at the cloud's outer boundaries and on its axes. These MCs would produce dierent characteristic magnetic signatures when passing over a spacecraft. According to Table 2, MCs can be classi®ed into SEN (SWN) clouds where the magnetic ®eld vector turns from south (S) to east (E) (west, W) on the cloud's axis and ®nally to the north (N) at its rear boundary, and vice versa into NES (NWS) clouds. More generally we can de®ne Dh b 0 for SNclouds (Dh`0 for NS clouds) with 180 `u`0 360 b h b 180 for SEN, NES (SWN, NWS) clouds. The classi®cation introduced here involves the property of magnetic helicity (see Burlaga, 1988;Lepping et al., 1990). SEN and NWS clouds possess left-handed (LH), SWN and NES clouds, right-handed (RH) helicity as viewed by an observer looking towards the Sun. In situ observations of these four dierent MC-types will be presented in the following section.
5.1 Magnetic cloud types 5.1.1 SEN-type MC. Figure 5 shows a shock-associated SEN-cloud observed by Helios 1 in August 1977. The magnetic ®eld direction changes smoothly from south to north during the cloud's passage over the spacecraft. Note that the southward dips of the magnetic ®eld vector in the cloud's sheath region (see the variations of h, labelled D, in Fig. 5) were caused by draping of the IMF ahead of the MC (Gosling and McComas, 1987;McComas et al., 1988). At the cloud's center the magnetic ®eld had an eastward direction. In the plane of maximum variance we ®nd a rotation of the magnetic ®eld vector from negative to positive f Ã x -values and f Ã yvalues. In this case, the component f Ã z along the direction of minimum variance had a constant, but substantial non-zero value (see Table 1) which might be indicative of twisted rather than pure circular outer magnetic ®eld lines. The ratio of k 2 ak 3 was 120, and the direction of minimum variance was u 205 , h 17 in SECs. The cloud's axis was lying almost parallel to the ecliptic plane h 12 , normal to the sun-spacecraft-line u 112 . Note also the high internal total plasma pressure and the very low plasma-b p inside the cloud. Other interesting features of this MC were its high magnetic ®eld strength with a very symmetric time pro®le and the extreme drop of the plasma temperature near the center of the cloud.

SWN-type MC.
A shock-associated SWN-cloud observed in December 1978 is presented in Fig. 6. The magnetic ®eld changes of h and u look relatively moderate, but a unique rotation of the ®eld vector is present in the plane of maximum variance. Note that this cloud might have gone undetected had the data simply been inspected by eye. The axis of the cloud was lying close to the sun-spacecraft line u 177 , pointing slightly towards the south h À30 . The minimum variance direction was u 230 , h 47 . Inside the cloud b p is low and is high, as is typical. The strong decrease of the magnetic ®eld strength from the cloud's front to its rear has probably been caused by interaction of this MC with the ambient solar wind.

NES-type MC.
A shock-associated NES-cloud observed by Helios 1 in January 1977, directly at a sector boundary (see the changes of u) is shown in Fig. 7. The peak of the magnetic ®eld strength at the cloud's front suggests strong interaction with the ambient plasma, which certainly also deformed the magnetic ®eld lines. Although the ®eld vector changes direction from 90 N to 0 , with u % 90 at the cloud's center, the plane of maximum variance revealed a rather complex rotation, presumably a result of compression and distortion of the ®eld lines due to the cloud's interaction with the ambient slower moving solar wind. An extremely high overabundance of He has been measured inside this cloud (Schwenn et al., 1980) being most probably cold chromospheric material caused by an erupting prominence (Cane et al., 1986). The cloud's axis was lying in the ecliptic h 8 , normal to the sun-spacecraft line u 103 . The direction of minimum variance was near the ecliptic, approximately radially outwards u 191 Y h À15 . Schwenn et al. (1980) explicitly stated that they did not ®nd any evidence for the end-time of the driver gas within 48 h following the arrival of the shock wave at $ 01 UT on January 29. However, the rotation of the magnetic ®eld vector ends 15 h earlier, around $ 10 UT on January 30. In this case, the MC might have been the leading portion of a much larger plasma cloud from a SME, consistent with observations of intermittent intervals of BDEs following the MC (private communication, K. Ivory, Max-Planck-Institut fuÈ r Aeronomie, Katlenburg-Lindau, Germany), as was the case for an IPC observed recently with the Ulysses spacecraft (Gosling et al., 1995a;Bothmer et al., 1996b). Events like these need further investigation to clarify their nature.

NWS-type MC.
A shock-associated NWS-cloud passed Helios 1 in June 1980 (Fig. 8, see also . It was directly associated with a Heliosdirected SME detected with the Solwind-coronagraph onboard the P78/1 satellite (Sheeley et al., 1985). The MC shows low b p and high . The magnetic ®eld vector turns from north (N) over west (W) ®nally to the south (S) at the cloud's rear. Note also the draping of the IMF in the sheath region between the shock and the MC, associated with high b p values, and the ®eld variations following the cloud when the surrounding IMF drapes back to its undisturbed orientation. The cloud's axis was lying close to the EW-direction u 87 , slightly inclined towards the ecliptic h À17 . The angles for the direction of minimum variance were u 168 Y h 26 .

Statistical results
For all 46 MCs we determined their orientation and magnetic con®guration (see Table 1). The frequency distribution of the calculated axis orientations in terms of h and u is shown in Fig. 9. In the top diagram of Fig. 9 we compared the frequency distribution of the axes inclinations to the ecliptic between 90 S±90 N found for the Helios 1/2 MCs with a random distribution of orientations for the same number of events in bins of 20 . Solid bars represent the number of MCs in each bin, open bars represent the number of cases as expected from the random distribution. Assume that the sky is divided into bins of equal areas of one square each, then for h 0 there are 360 possibilities for u, but there is only one possibility for h 90 . Thus random orientations at 0 should be 360 times more likely than those at . A more sophisticated approach shows that the area is proportional to 5 for the measured range of 5, where 5 is the colatitude. Figure 9 shows that the distributions are roughly comparable, i.e., MCs can have in principal any inclination with respect to the ecliptic. However, there is an indication for an overabundance of axes inclinations in the range 10 S±10 N. Since MC axis do not spread randomly over 180 in azimuth the number of cases in the interval 10 S±10 N should be too peaked in the random distribution in Fig. 9 supporting the assumption that the axes of MCs commonly lie rather close to the ecliptic than highly inclined to it.
The distribution for the azimuthal axes direction of MCs in the ecliptic between 0 ±180 is shown in the lower diagram of Fig. 9, also in bins of 20 . The axes scattered about the east-west direction, i.e., normal to the sun-spacecraft line. The average azimuthal orientation of the¯ux tube axes was hui 91 , rms=44 , and the average orientation of the minimum variance direction was hui 188 , rms=38 and hhi 3 , rms=27 . Our ®ndings con®rm the results for IPCs and shocks found by Smith and Phillips (1997). Most of the clouds' axes were lying inside a cone angle of AE45 with respect to the east-west direction in the ecliptic, distinctive from the average Parker spiral direction of the IMF u % 135 at 1 AU). The large scatter of the axes orientations around the east-west direction may result from passage through the individual MCs under dierent aspect angle relative to the center of the MCs. Often left or right wings of the¯ux tubes may have been encountered by the Helios probes. BDEs seem to be best explained through heat¯uxes of supra-thermal electrons along magnetic ®eld lines that remain connected back at both ends to the solar corona (e.g., Gosling, 1990) so that expanding¯ux tubes should in reality be curved obstacles, which could explain the observed scatter of the axes orientations in the azimuthal direction. Table 2 shows the frequencies of the dierent¯ux rope types observed by Helios 1/2. SEN-and SWNclouds were observed approximately three times more frequently than NES-and NWS-type MCs during 1974 to 1981. On the other hand, comparing the frequency of MCs by magnetic helicity, the number of left-handed (SEN, NWS) and right-handed (SWN, NES) MCs was the same, as one might expect if these types occur with equal probability. Zhang and Burlaga (1988) found that 79% of the MCs they investigated for the years 1978± 1982 were associated with SN-rotations of the magnetic ®eld vector, but they did not gave a physical explanation for this preference.
A time-period of the solar cycle similar to the years 1974±1981 is 1984±1991. For these years we inspected the near Earth solar wind data (OMNI-database at NSSDC, for a description of this data set see King, (1991), for MCs showing a unique coherent internal SNor NS-rotation of the magnetic ®eld vector according to the identi®cation criteria of MCs at 1 AU (Burlaga, 1991). Our search yielded only nine clearly de®ned MCs, since the data are sparse and we did not investigate events with complex structure or time durations shorter than $1 day. Table 3 shows that seven of the nine MCs identi®ed between 1984±1991 had NS-rotations, contrary to the years 1974±1981 where SN-MCs were dominant.
For the years 1974±1981 (solar cycle 21), magnetic ®eld lines in the Sun's northern hemisphere had predominantly northern magnetic polarity, i.e. they were pointing away from the Sun, whereas magnetic ®eld lines in the Sun's southern hemisphere were predominantly directed sunwards (southern magnetic polarity). During 1984±1991 (cycle 22) the magnetic polarities in the Sun's hemispheres were reversed. The observations are in agreement with the ®ndings of Bothmer and Rust (1997) who suggested a solar cycle dependence of the magnetic con®guration of MCs. This topic will be further discussed in Sec. 8,9. 6 Large-scale structure of MCs MCs are thought to be large-scale solar wind structures. Thus, they should be observable over a substantial range in solar longitude/latitude with the same overall mag-netic¯ux tube structure. One spectacular event in which the same MC was observed by Helios, Voyager and near-Earth spacecraft at distances between 1±2 AU has been studied in detail by Burlaga et al. (1990). The authors concluded that the spacecraft observations are in agreement with the model of a MC as a large-scale magnetic¯ux-rope.
To further clarify the structure of MCs, we investigated near-Earth solar wind data collected in the OMNI-database at NSSDC to see if the MCs observed by Helios 1/2 also passed the Earth (see Table 1). In a few cases, when only plasma data were available and when the MC observed by the Helios s/c had a substantial southward magnetic ®eld, we additionally investigated geomagnetic activity which sensitively responds to large southward components of the IMF (see, e.g., Bothmer and Schwenn, 1995). Doubtful events due to major gaps in the solar wind data and other uncertainties were discarded, leaving 9 cases (note in Table 1 that Helios 1 and Helios 2 observed the same  MC in the events in January and February 1978). Figure 10 gives an example. The left panel shows a MC that was observed by Helios 2 in April 1979 at 0.7 AU, 27 east of the Earth. Due to data gaps it was dicult to de®ne the rear boundary of the MC unambiguously. The time interval that we identi®ed as the MC is comparable to the start-and end-times given for this event by . This MC later passed Earth's orbit as shown by IMP/ ISEE observations in the right panel. The MC was associated with an interplanetary shock and caused a strong geomagnetic storm with Kp! 8-(see also . From the results of the MVA the cloud was identi®ed as a¯ux tube of type SWN. The spacecraft measurements of IMP/ISEE and that of Helios 2 show a striking similarity, although the MC was observed in dierent ambient solar wind streams. At Earth's orbit the MC was propagating through fast ambient wind, whereas it plowed through low speed wind at Helios 2. This observation is consistent with the slightly higher plasma speed of the MC at Earth's orbit (V$500 km/sec), compared with the plasma speed measured at Helios 2 (V$450 km/sec). Note that Helios 2 observed the MC at a sector boundary (see the change in u from $ 315 to $ 135 in Fig. 10). The orientation of the cloud's axis as calculated from the minimum  Figure 11 is a simple sketch for the possible largescale con®guration of this MC at the arrival time of the MC at Helios 2 based on the calculated axis orientations at the positions of Helios 2 and at Earth's orbit. The axis orientation of the MC at the heliolongitude of the Earth at the arrival time of the MC's center at Helios 2 was determined back by using the speed of the MC measured at 1 AU and the time delay between the MC's arrival at Helios 2 and 1 AU, assuming that the MC propagated along the Sun-Earth line with constant speed. Note that Helios 1 at 0.7 AU, 43 east of Helios 2, did observe the shock, but no signatures of a MC or ejecta in general could be found in the data.
In nine cases where a MC was detected from both Helios and Earth-bound satellites, longitudinal separations of up to $ 60 occurred. We ®nd that in cases with separations b 60 , the MC was observed by only one spacecraft. Figure 12 shows the histogram obtained from our study. In two cases, in which both Helios s/c observed the same MC, we used the value of the spacecraft which was separated largest in longitude from Earth. Based on these observations the azimuthal broadness of a MC (60 in heliolongitude 91 AU at 1 AU) should be larger than its radial size (the typical diameter of a MC is $0.25 AU at 1 AU). Figure 12 might imply a smaller extent for MCs in heliolongitude compared to the value of $ 100 found previously for shock-associated IPCs (Richardson and Cane, 1993), but we have not found any case where a MC was observed with two spacecraft separated by more than 60 . A possible explanation for this discrepancy could be that the events studied by Richardson and Cane (1993) were all¯are-associated and thus probably more energetic and larger in their spatial extent. Burlaga and Behannon (1982) investigated MCs in the outer heliosphere using Voyager data. They found larger radial sizes for MCs observed further away from the Sun. To investigate the expansion of MCs in the heliosphere, we used the MCs observed by Helios 1/2 between 0.3±1 AU and investigated whether they had passed also the Voyager spacecraft in the outer heliosphere (s/c data other than Helios were taken from the COHO data base at NSSDC, described, e.g., in NSSDC News vol. 8(2), 1992). We found a number of cases in which the same MC could be uniquely identi®ed in both data sets with time delays corresponding to the outward propagation of the MCs from the Helios to the Voyager spacecraft with approximately constant speed.

Expansion of Magnetic clouds
Here we present an example observed by Helios 1 at 0.9 AU and by Voyager 1/2 at 2.6 AU in March 1978 (for the identi®cation of the MC in the Voyager data see also Burlaga and Behannon, 1982). The left side of Fig. 13 shows magnetic ®eld data for Helios 1 (top panel) and Voyager 1/2 (middle, bottom panel), and the right side shows the corresponding results of the MVA. The spacecraft were separated by 18 in solar longitude and by 1.6 AU in solar distance. No signatures of a solar wind transient were found in the plasma data at Earth's orbit, 55 west of Helios 1, i.e., 73 west of Voyager 1/2, Fig. 11. Sketch showing the possible large-scale geometry of the MC observed by Helios 2 and IMP/ISEE in April 1979 (see Fig. 10) based on results of the MVA of the magnetic ®eld data. Helios 1 did observe the shock, but not the MC. Arrows denote the orientation of the magnetic ®eld lines at the cloud's outer boundaries and on its axis so that we can assume that there was also no MC at the Earth although there was a gap in the magnetic ®eld data. Note the striking similarity of the directional changes of the magnetic ®eld vector at all three spacecraft, which all detected a SEN-type MC. The arrival time of the MC at Voyager 1, at 2.6 AU (08 UT on day 66) and at Voyager 2 (02 UT on day 66) agreed well with the propagation time from Helios 1 to Voyager 1/2 as estimated from the plasma speed of $446 km/sec observed during the cloud's passage over Helios 1 at 0.9 AU. The cloud showed a much longer time-duration at the position of Voyager 1 (t 50 h) and Voyager 2 (t 48 h) compared to Helios 1 (t 24 h), thus giving direct evidence for the expansion of this MC. The MVA yielded similar results for the orientation of the cloud's axis at all three spacecraft, with values of u 119 , h 44 at Helios 1, u 91 , h 30 at Voyager 1, and u 83 , h 28 at Voyager 2. Figure 14 presents a log-log plot showing the growth of the radial sizes of MCs between 0.3 and 4.2 AU obtained from Helios 1/2, Voyager 1/2, and Pioneer 10 data (Bothmer and Schwenn, 1994). The smallest MCs had radial sizes`0X1 AU and were observed close to the Sun, the largest MCs had radial sizes b 0X4 AU and were observed beyond 2 AU.
From linear regression it is found that the radial size, s () in AU, of MCs increases with radial distance, , from the Sun as: There is a clear trend for an increase in the radial sizes of MCs with distance from the Sun, con®rming that radial expansion is a common feature of MCs in the heliosphere.
MCs observed considerably closer to the Sun compared to 1 AU had often higher plasma densities than the surrounding solar wind (see Fig. 1). The density inside the¯ux tubes might fall o rapidly with increasing distance from the Sun as indicated by the radial expansion found for MCs. We determined the average proton density, hx i, inside each of the MCs observed by the Helios probes. Figure 15 shows in a log-log plot lnhx p i versus ln(). For comparison the decrease of the proton density measured in slow (`400 km/s; dashed line) and fast ( b 600 km/s; dotted line) solar wind streams over the range 0.3±1 AU has been included. These slopes were calculated according to results summarized in Schwenn (1990) with hx p i x p Â À2X1 cm À3 ( in AU).
From linear regression we found that the density of MCs decreases with increasing radial distance, , as: Thus the density inside MCs decreases stronger with radial distance from the Sun than that of the average solar wind which falls o as À2X1 in agreement with what one would expect for expanding structures. Figure 15 shows that the density in MCs is generally higher than that of the average fast solar wind and that it is also generally higher than that of the slow solar wind at closer distances to the Sun. Note that this result implies, without taking density enhancements into account that have been caused by compression due to the interaction of a MC with the ambient solar wind, that IPC type MCs should be observed as transient white-light depletions rather than enhancements in the interplanetary medium at larger (J0X5 AU) distances from the Sun, in agreement with the suggested higher preference for MC associations with photometer IPCs in Table 1 for MCs observed closer to the Sun. MCs in which the densities are found to be considerably lower compared to that of the ambient slow solar wind should have undergone strong expansion on their way out from the Sun. Consistent with this result, recent observations of the Ulysses spacecraft showed that IPCs including MCs were over-expanding at high latitudes in high speed solar wind due to their higher internal pressure compared to the surrounding¯ow (Gosling et al. 1994a,b).
To test whether our results are consistent with thē ux-tube model for MCs, we assume that a MC has the topology of an ideal cylindrical¯ux tube of length 1 and diameter s, so that its volume is proportional to s 2 Â 1. We assume further mass is conserved inside the¯ux tube, i.e. $ 1ax p X From Eq. (1) the diameter of the¯ux tube increases due to the expansion as 0X78 X Based on observations of prominence eruptions and white-light observations from space-coronagraphs (see e.g. Webb and Hundhausen, 1987;Webb, 1988) we assume that 1 increases linearly with . Thus, the volume of the¯ux tube should increase as 2X56 , i.e., the density should decrease as À2X56 . If we consider the simplicity of the model assumptions, we ®nd this value to be in qualitative agreement with the À2X4 dependence in Eq. (2). The present results obtained for MCs observed in the ecliptic by Helios 1/2 are in agreement with the possible largescale topology of a MC sketched in Fig. 11. x p 6X47 AE 0X85 Â À2X4 AE 0X3 cm À3 (R in AU). Dashed and dotted lines represent slopes for proton densities measured by the Helios probes in slow (S) and fast (F) solar wind-streams in the inner heliosphere (taken from Schwenn, 1991) 8 Magnetic clouds and disappearing ®laments Studies of the relationship of SMEs with other features of solar activity have shown that SMEs are better associated with prominence eruptions (disappearing ®laments) than with solar¯ares (Harrison, 1986;Webb and Hundhausen, 1987). Yet, we do not know what causes SMEs nor which processes lead to the speci®c topology of the subset of SMEs associated with MCs in the interplanetary medium. A possible explanation is that a SME is caused by magnetic nonequilibrium leading to the rise of a prominence (e.g., Hundhausen et al. 1994;Priest, 1988;Feynman and Martin, 1995). The typical three-part structure of SMEs seen in white-light images (see, e.g., Webb, 1988 and references therein) is interpreted as a leading bright front comprised of coronal ®eld lines with a cavity beneath it, followed by the rising prominence in the trailing portion of the SME.
Erupting quiescent prominences often show a mag-netic¯ux-rope topology, so that it may seem plausible to speculate whether they are related to MCs (the term`e ruption'' means here that a prominence disappears in the Ha-observations, i.e., an erupting prominence might remain attached to the Sun). The concept of force-free magnetic¯ux tubes was ®rst introduced by LuÈ st and SchluÈ ter (1954) for quiescent solar prominences and was later applied to MCs (Goldstein, 1983; see also the review on magnetic¯ux ropes given by Priest, 1990 and the paper on``theory of prominence eruption and propagation'' by Chen, 1996).
The`Catalog of Solar Filament Disappearances 1964±1980' (Wright, 1991) oered us the most useful data set to investigate whether quiescent ®laments (disparition brusques, DBs) and MCs can be associated and to compare their magnetic structures (see also Bothmer, 1993;Rust, 1994). In order to establish unique associations between MCs that were observed in situ in interplanetary space by the Helios probes and DBs observed from Earth we used the following criteria: (1) Since the observations of DBs were taken from Earth we did not include any MC that was observed by the Helios probes when it was positioned further away than 60 in solar longitude from the Sun's central meridian (as seen from Earth), in order to avoid DB-events which might have gone undetected due to limb darkening or because the DB might have occurred behind the Sun's limbs which reduced the number of events to 25 cases. (2) We regarded only those associations of MCs and DBs as safe when spacecraft and ®lament were separated not more than $ 20 in solar longitude, and when the DBs had disappeared at the Sun in time windows (including the error in the given DB disappearance) that were calculated from the maximum and minimum proton speeds measured inside the MCs.
These criteria led to the 10 MC events listed in Table 4. In ®ve cases, MCs could be con®dentially associated with a DB at the Sun and we compared their magnetic structures (see also Bothmer and Schwenn, 1994). The solar latitude of the ®laments was $ 30±60 (see Table 4). We remark that for the ®rst two associations in Table 4 more than one ®lament disappeared during the estimated time windows for the solar events. In these cases the DB closest in longitude to the Helios probes has been selected. It is interesting to note that the daily rate of DBs between 45 E and 45 W was only 0.1±0.4 during 1975±1980.  Figure 16 shows one of the ®ve MCs of Table 4 that was associated with a DB. This MC was observed by Helios 1 in December 1977 at 0.8 AU. The ®gure shows the magnetic¯ux tube type as inferred from the MVA results and the inferred magnetic structure of the associated DB. The magnetic structure of the ®lament has been inferred from the polarity of the magnetic ®eld with respect to the prominence axis based on the observation that the ®eld structure in quiescent prominences is composed of loop-like arcades overlying the prominence axis (neutral line) (see, e.g., Dynamics and Structure of Quiescent Solar Prominences, edited by Priest, 1989), by taking into account the hemispheric magnetic patterns found for mid-to high-latitude ®laments (Rust, 1967;Martin et al., 1994;Rust and Kumar, 1994a) and by applying the helicity rule found by Rust and Kumar (1994a) who inferred that for prominences in the northern solar hemisphere, the twist of ®eld lines is preferentially left-handed, whereas it is preferentially right-handed for prominences in the southern hemisphere (compare with Fig. 18). This ®nding is consistent with sunspot observations which show that anti-clockwise whorls are three times as common in the north and clockwise ones in the south (Richardson, 1941;Yang et al., 1988). Anti-clockwise whorls are related to dextral and clockwise ones to sinistral ®laments. It is interesting to note that the Helios observations showed SN-MCs to be three times more frequent than NS-MCs which is similar to the ratio of clockwise and anti-clockwise whorls in sunspots (Richardson, 1941).
The agreement of the magnetic topologies of the MC and the DB in Fig. 16 is striking. The individual¯uxtube types given for the DBs in Table 4 have been labeled according to the notation that we introduced for MCs in Section 4 of this study (to compare with Table 2). Similar agreements between the magnetic structure of the ®lament and that of the associated MC were found in four of the ®ve events listed in Table 4, with three dierent types of¯ux-tubes. In the one event in which we have not found an agreement, the MC had a complex internal magnetic ®eld structure.  Table 4). Left bottom panel: results from the MVA of the magnetic ®eld data inside the MC. Right bottom panel: in situ observed¯ux tube type of the MC In the same way as for the Helios data, we have searched near Earth solar wind data from the OMNI and COHO data-bases at NSSDC for the years 1978± 1980 when the data coverage was very good. We identi®ed 12 MCs during these years. In four events DB observations were missing, one MC was not associated with a DB and in one event the ®lament was separated substantially in longitude from central meridian. Table 5 lists the six events that were found to be associated with a DB using the same criteria as for the Helios observations. Note that the ®rst two events in Table 5 are MCs that have also been observed with the Helios probes so that Table 5 yields a total set of four new associations between MCs and DBs. We found an agreement between the magnetic¯ux tube type of the MC and that of the associated DB in all cases. Similar to the Helios observations, the ®laments associated with the MCs, were observed typically at higher latitudes, in agreement with the recent ®ndings by Bothmer and Rust (1997) based on an extended study of MCs observed at 1 AU and DBs.
Our results are summarized in Fig. 17 (adapted from Bothmer and Schwenn, 1994): four dierent magnetic ux tubes are expected in interplanetary space from the magnetic structure of disappearing ®laments in the Sun's southern and northern hemispheres. The results indicate that the magnetic helicity of MCs observed in interplanetary space is the same as that of the associated ®laments, i.e., the magnetic helicity inherent in magnetic ®elds at the Sun might be conserved outward into interplanetary space, as has been pointed out by Rust and Kumar (1994b) in agreement with ®ndings of helicity conservation in laboratory plasmas. Note that on the contrary, Martin and McAllister (1997) claimed that the helicity of the overlying arches above a prominence axis which may evolve into¯ux tube is opposite to that of the ®lament beneath it.
Magnetic¯ux ropes/tubes may be formed through magnetic reconnection of coronal ®eld lines in a rising SME as has been pointed out by Gosling (1990) and Gosling et al. (1995c,d). In support of this view, the Ulysses spacecraft has directly observed MCs and IPCs after formation of new coronal loops in solar regions that map to the spacecraft, as observed by the Japanese Yohkoh-satellite (Gosling et al., 1995c;Bothmer et al., 1996b;Lemen et al., 1996;Weiss et al., 1996). We believe that our ®ndings are further evidence for an evolution of arcades of coronal loops into magnetic¯ux tubes (MCs). Weiss et al. (1996) found no characteristic dierences in the coronal soft X-ray signatures of IPCand MC-type IPC events, but they pointed out that the interplanetary plasma-b of SMEs appears to be crucial in determining the magnetic ®eld evolution. This is consistent with our ®nding of low plasma-b values in MCs between 0.3±1 AU. It is further interesting to remark that the magnetic ®elds of quiescent ®laments are presumably low plasma-b regions in the solar atmosphere (see, e.g., Priest, 1989).
We found evidence for a solar cycle variation of the magnetic structure of MCs in Sec. 5.2. In Fig. 18 (Bothmer and Rust, 1997) we propose an extended association between the polarity of sunspots, the magnetic structure of ®laments and that of MCs, that could explain our observations.
It is well known that Ha ®laments commonly occur in regions where sunspots are no longer visible. Usually the preceding sunspot (in the sense of solar rotation) lies closer to the solar equator than the following one which is of opposite magnetic polarity. The left column of  Bothmer and Schwenn, 1994). The abbreviations for the magnetic ®eld structures have been chosen in analogy to the dierent¯ux tube types of MCs presented in Table 2. LH (left-handed) and RH (right-handed) denote the corresponding magnetic helicity Fig. 18 shows the magnetic polarities of sunspots in both solar hemispheres during two subsequent cycles. The top row of Fig. 18 shows the magnetic polarities for odd cycles, e.g., for cycle 21, when the Helios observations were taken. The bottom row shows the polarities during even cycles. The middle column sketches the magnetic structures for mid-to high-latitude ®laments during the two cycles based on the hemispheric helicity rule found by Rust and Kumar (1994) who inferred that in ®laments in the northern solar hemisphere the twist of the magnetic ®eld lines is preferentially left-handed, whereas it is preferentially right-handed in the southern hemisphere. The right column of Fig. 18 ®nally presents the magnetic structure of MCs based on the interpretation of our observations (see also Bothmer and Schwenn, 1994). Figure 18 implies that during odd cycles, e.g., during solar cycle 21, ®laments should preferentially evolve in association with SMEs into SN¯ux tubes (MCs) in the interplanetary medium, whereas in the following cycle preferentially NS¯ux tubes (MCs) should evolve, in agreement with our ®ndings. Further evidence for the solar cycle dependence of MCs and their agreement with the magnetic structure of ®laments based on the helicity rule has been recently presented by Bothmer and Rust (1997).

Conclusions and summary
Using the complete set of plasma and magnetic ®eld data of the Helios 1/2 spacecraft, we have identi®ed 46 MCs between 0.3±1 AU during the period 1974±1981 between 0.3 and 1 AU. Additionally, we used solar wind data from the IMP, Pioneer and Voyager spacecraft for multi-spacecraft studies of MCs and to investigate the expansion of MCs in the outer heliosphere. Based on these investigations, we summarize the main properties of MCs in the heliosphere: 1. Our results support the close relationship between MCs and SMEs. 85% of the MCs observed by Helios 1/2 were associated with a fast forward shock. Seven of the MCs were observed after a Helios-directed SME was detected by the Solwind-corongraph on the satellite P78/1. Without taking the duty cycle of the instruments into account, half of the MCs were found to be associated with passage of an IPC detected by the Helios-photometer as a white-light brightness enhancement in the interplanetary medium. 2. Between 0.3 and 1 AU, MCs have higher total plasma pressure and lower plasma-b values b>1 compared to the normal solar wind. 3. MCs can best be described as large-scale cylindrical magnetic¯ux tubes with observed longitudinal extents up to $60 . The axis of the¯ux tube commonly has a small inclination to the ecliptic, with its azimuthal direction being roughly directed near the EW-direction, i.e., normal to the sun-spacecraft line. 4. MCs can occur in four dierent magnetic con®gurations. The individual types are distinguished by the magnetic orientation at the center (the axis) of the MC and on its outer boundaries. The four con®gurations result from two possible directions of the ®eld axis and two possible values of magnetic helicity. 5. MCs expand during their outward propagation. Their radial diameter, s, was found to increase between 0.3 Fig. 18. Sketch showing the extended association between sunspots, ®laments and MCs. The left panel shown the polarity of sunspots during two subsequent cycles. Note that the solar cycle number, n, is assumed to be an odd number. The middle panel shows the expected preferential magnetic structure of ®laments in the two solar hemispheres and the right panel shows the expected structure of MCs in interplanetary space. During odd (n) cycles SN-MCs are dominant whereas NS-MCs dominate during even n 1 cycles (Bothmer and Rust, 1997) and 4.2 AU proportional to 0X8 . Consistent with this expansion, the plasma density inside MCs decreases, faster than the normal solar wind, proportional to À2X4 between 0.3 and 1 AU. According to this result, on average, an IPC type MC should be identi®able as a transient white-light brightness enhancement in the interplanetary medium up to a distance of~0.5 AU from the Sun, whereas, without taking brightness enhancements into account that have been caused by compression due to the interaction of a MC with the ambient solar wind, a MC should at greater distances from the Sun lead to a transient white-light brightness depletion. 6. We found nine MCs identi®ed in Helios and near-Earth solar wind data to be uniquely associated with quiescent ®lament disappearances (DBs) at mid to high solar latitudes outside active regions. In eight of the nine cases we found agreement between the magnetic¯ux tube structure of the MC and that of the associated ®lament as inferred from the orientation of the ®lament axis and its magnetic polarity, on the assumption of left-handed (right-handed) magnetic helicity dominance of ®laments in the northern (southern) solar hemisphere. Further, the preferred east-west orientation of mid-to high-latitude quiescent prominences caused by dierential rotation may explain the preference of MCs to have small axis inclinations to the ecliptic in the interplanetary medium. However, we note that some MCs may well be related to other solar activity features, such as ares. Also, other results do not necessarily imply that the ®lament itself evolves into the MC in interplanetary space. From a comparison of the spatial sizes of the features seen in Ha-®ltergrams and white-light coronagraph observations it seems likely that the ®lament may be either the bottom part of a large¯ux tube or just an indicator of the overlying coronal ®eld structure. 7. There is a solar cycle dependence of the magnetic structure of MCs. Although MCs of opposite magnetic helicity occurred with equal frequency during 1974±1981, $3/4 of the MCs were associated with SN-rotations of the magnetic ®eld vector, but only $1/4 had NS-rotations. During the subsequent cycle predominately NS-rotations were found. 8. We suggest that the solar cycle variation of the magnetic structure of MCs can be explained by the magnetic ®eld structure of disappearing ®laments at the Sun and their solar cycle changes. 9. It is likely that magnetic reconnection occurs in the coronal loops overlying the prominence axis of a disappearing ®lament in association with a SME.
Finally, the measured components of f in the minimum variance system are: