Magnetic Reconnection within the Boundary Layer of a Magnetic Cloud in the Solar Wind

Supposing a vertically oriented current sheet (Figure 6c.1), here we provide an order of magnitude predictions of the motion of MR outflow channel between the spacecraft.

In the MVA-rotated system, the reconnection LN plane is approximately perpendicular to the ecliptic, therefore, the crossing of the current sheet requires a vertical $V_{Z}$ speed (Figure 6c left). The positive $V_{Z}$ of the ambient plasma surrounding the MR outflow can already be seen in Figure 2e. At the same time the current sheet is convected by the solar wind bulk speed ($\sim V_{X}$). Along the outflow in the spacecraft frame, the outflow front speed is the sum of the solar wind bulk speed and the outflow speed. Here we should be able to answer the question why the vertical current sheet with $\mathbf{N}$ $\sim$ Z GSE undergoing a vertical motion in positive Z GSE direction can be seen by ACE, Geotail and WIND, but not by SOHO. To show that we will use the time-delay of outflow observations between the spacecraft, calculate the spatial vertical shift of the supposedly planar outflow channel with constant average velocity $V_{Z}$ and estimate an additional change in the vertical location of the flow channel in a given position which appears due to the inclination of $\mathbf{L}$ direction relative to the GSE-XY plane. The latter can be positive or negative depending on the position of a given location relative to a reference spacecraft. Although the $\mathbf{L}$ direction and the inclination angle slightly change between the spacecraft, initially we choose constant values for these quantities between the spacecraft.

Figure 7 shows the time delays in MR outflow observations (all in GSE) between ACE (Figures 7a-e), Geotail (Figure 7f) and WIND (Figures 7g-k) spacecraft. For each spacecraft the magnetic field magnitudes are shown in Figures 7a, f and g. The $B_{X}$, $B_{Y}$ and $B_{Z}$ components are in Figures 7b and h. The magnitude of radial $|V_{X}|$ and transverse $V_{Y,Z}=\sqrt{(}V_{Y}^{2}+V_{Z}^{2})$ velocities are in Figures 7c and i. The $V_{Y},V_{Z}$ solar wind velocity components and ion density Ni are in Figures 7d and j. The ion density and temperature are in Figures 7e and k. In this panel the same Y axis is used for Ni and Ti. For this time interval there are no reliable plasma measurements from Geotail in the solar wind.

The sharp boundary in $|B|$ at the right side of the outflow at each spacecraft can be used as a reference point to estimate the time delay between outflow observations. At ACE the sharp boundary in $|B|$ is seen at 15:09 UT (Figure 7a). The same structure is seen by Geotail t(ACE-Geotail)=50.2 minutes later (Figure 7f) and by WIND t(ACE-WIND)=108.6 minutes later (Figure 7g). The large-scale shear in $V_{Z}$, which was explained in Figure 2e, is seen here at both ACE and WIND (Figures 7d, j). However, at ACE, the outflow is embedded into a flow structure with $V_{Z}\sim$ 18 km/s (Figure 7d). Since the ambient magnetic field before and after the outflow is predominantly in the X-Y GSE plane (Figure 7b), the $V_{Z}$-GSE velocity is approximately perpendicular to the magnetic field there. The vertical crossing speed and the duration of the outflow gives for the current sheet thickness 12900 km ($\sim$ 2 $R_{E}$ or 157$d_{i}$).

The vertical motion of the outflow channel from ACE GSE Z = -7 $R_{E}$ (Figure 1) to Geotail location can be predicted. Supposing planar geometry for the outflow and a constant $V_{Z}$ between the spacecraft, the vertical shift of the flow channel is 8.5 $R_{E}$, i.e. the outflow structure moves from ACE GSE Z = -7 $R_{E}$ to the location Z GSE $\sim$ 1.5 $R_{E}$ (not shown). This value has to be corrected by $\Delta$Z which is due to the inclination of the $\mathbf{L}$ direction relative to the GSE-XY plane. Since at Geotail $\mathbf{L}$ =[-0.63 0.78 -0.06], the flow inclination angle is 3.4^∘ and over the distance of $\Delta L$ along $\mathbf{L}$ (Figure 6a), $\Delta Z$ =1.3 $R_{E}$. By adding it to 1.5 $R_{E}$, the predicted vertical location of the flow channel is Z GSE $\sim$2.8 $R_{E}$ at Geotail. This roughly matches the actual Geotail Z GSE = 2 $R_{E}$ location (Figure 6a), where the outflow is de-facto observed. The 0.8 $R_{E}$ difference can appear as a result of changing $V_{Z}$ between the spacecraft by 1-2 km/s or/and due to the error in determination of L inclination angle by $<$1^∘.

SOHO at GSE [X,Y] = [241,89] $R_{E}$ should see the $\mathbf{L}$ directional outflow at the same time as it goes through an upstream location A1=GSE [X,Y] = [226+$\Delta$A1,-30] $R_{E}$, where [226,-30] $R_{E}$ is the location of ACE (Figure 1) and $\Delta$A1=110 $R_{E}$ is determined by triangulation (not shown). On this basis SOHO should see the outflow $\sim$1630 s earlier than ACE. During this time the vertical shift of the outflow structure due to $V_{Z}$ = 18 km/s is $\sim$4.6RE. At ACE $\mathbf{L}$ =[-0.62 0.78 -0.08] and the inclination angle of $\mathbf{L}$ is 4.6^∘. This gives over a distance of $\Delta$L=152 $R_{E}$ (distance between upstream location A1 and SOHO, determined by triangulation) a positive shift $\Delta$Z=12 $R_{E}$, which adds to 4.6 $R_{E}$, leading to the predicted vertical location of the outflow of Z GSE =16.6 $R_{E}$, while the actual SOHO Z GSE =-13 $R_{E}$ (Figure 1). Since the thickness of the current sheet is $\sim$ 2 $R_{E}$, this might explain why SOHO does not observe the outflow (not shown).

The prediction of the vertical shift of the outflow channel from ACE to WIND is connected with more uncertainties because of the large GSE-XY differences in their locations. ACE at GSE [X,Y] = [226,-30] $R_{E}$ should observe the $\mathbf{L}$ directional outflow at the same time as it goes through an upstream location $W$=GSE [X,Y] = [32.8+$\Delta W$,-252] $R_{E}$, where [32.8,-252] $R_{E}$ is the location of WIND (Figure 6a) and $\Delta W$= 367.2 $R_{E}$ is determined by triangulation. We consider here only the upstream point W at GSE [X,Y] = [400,-252] $R_{E}$ (Figure 6a). Since WIND observes the outflow 108.6 minutes later than ACE (Figure 7) the vertical shift of the outflow channel due to $V_{Z}=$ 18 km/s during this time is $\sim$ 18.4 $R_{E}$. At ACE, Geotail and WIND the $\mathbf{L}$ directional inclination angle relative to the GSE XY plane varies as 3.4^∘, 4^∘ and 4.6 ^∘, respectively. By taking the smallest inclination angle, this gives over a distance of $\Delta L$=279.5 $R_{E}$ (here this is the distance between upstream location $W$ and ACE, determined by triangulation) $\Delta Z$=-16.5 $R_{E}$, which adds to +18.4 $R_{E}$ (due to $V_{Z}$=18 km/s), leading to a net vertical shift of the flow channel by +1.9$R_{E}$. Thus, the flow channel from ACE GSE Z=-7 $R_{E}$ is predicted to shift to GSE Z=-5.1 $R_{E}$ at WIND while the actual WIND is at GSE Z=-4 $R_{E}$. By taking $V_{Z}$=19 km/s between upstream point W and WIND, the net vertical shift is modified to +2.9 $R_{E}$, leading to the predicted vertical location of the flow channel of -4.1 $R_{E}$ at WIND which is matching WIND GSE Z=-4 $R_{E}$ location pretty well. The above rough estimations of the vertical motion of MR outflow channel (the orientation of the LN plane is shown in Figure 6c.1) support the idea that the relatively thin flow structure (of width $\sim$ 2 $R_{E}$) can be sequentially observed by ACE, Geotail and WIND spacecraft, but not seen at SOHO location.

First the density structure around MR outflow is considered. The yellow dashed lines in Figures 7a-e (ACE) and in Figures g-k (WIND) show that at the density jumps, $|B|$, Ti and velocity components exhibit smaller or larger jumps. From Geotail $|B|$ is shown only (Figure 7f).

The jumps in different parameters supposedly correspond to discontinuities. Before the outflow in the ICME sheath (at vector $\mathbf{B_{1}}$ in Figure 6) the discontinuities are labeled as D1A (roughly at 14:12 UT), D1G (at 15:00 UT) and D1W (at 16:35 UT), as observed by ACE, Geotail and WIND, respectively. The discontinuities after the outflow at the boundary layer of the MC (at vector $\mathbf{B_{2}}$ in Figure 6) are marked similarly as D2A (at 15:38 UT), D2G (at 16:30 UT) and D2W (at 17:18 UT). Although there might be other discontinuities earlier or later in time, we are interested in those when $|B|$ decreases and Ni changes before and after the outflow. Here our aim is not to verify the Rankine-Hugoniot jump relations (e.g. Burlaga et al., 1995), but rather to understand possible interactions between discontinuities and the outflow. The other goal is to compare the evolution of discontinuities on the basis of available data at different distances from the MR X-line. The field and plasma parameters across discontinuities D1A and D1W in the ICME sheath are in Table 2. The same for D2A and D2W is shown in Table 3. The GSE field and plasma variables represent 10 minute averages before and after the discontinuity, not including the largest gradients at jumps. The downstream side of a discontinuity is where $|B|$ decreases, Ni and Ti increase. The discontinuity MVA normals in GSE are shown in the first rows in Tables 2 and 3.

Across D1A (Table2, left) $|B|$ decreases, $V$ decreases, both Ni and Ti increase. $V_{X}$ does not change, the transverse ion velocities $V_{Y}$, $V_{Z}$ and $V_{YZ}=\sqrt{V_{Y}^{2}+V_{Z}^{2}}$ (Figures 7c and d) decrease. Since the local Alfvén speed is $\sim$ 100 km/s the velocity changes are sub-Alfvénic. Initially we suppose that, D1A is neither a forward slow shock (SS), for which Ni, Ti and V should increase, B decrease nor a reverse SS, for which V and B should increase, Ni, Ti decrease. Across D1A the total pressure (magnetic and ion thermal pressure, electron thermal pressure is not available) changes from 0.15 to 0.1 NPa (not shown). However, without the electron thermal pressure it is not possible to classify D1A as a pressure balanced structure. Since $|B|$ and Ni change across D1A, it is not a rotational discontinuity. Although in anisotropic plasmas $|B|$ and Ni could change across a rotational discontinuity, for low beta plasmas (in our case beta $<$ 0.1) Ni could change by at most a factor of 1.1 and $|B|$ should be almost constant at 1 AU (Hudson, 1973). In our case Ni changes more than 1.6 times. D1A could also be a tangential discontinuity. In that case, supposing pressure balance, Ni, Ti and $\mathbf{B}$ can change across D1A, however, the normal component of $\mathbf{B}$ should vanish at the discontinuity surface. By calculating the scalar product $\mathbf{B}_{u,d}\cdot\mathbf{n}\sim$ 6 nT, we find that the normal component does not vanish at D1A. The subscripts u and d refer to the upstream and downstream values, $\mathbf{n}$ is the normal.

D1A can be considered also as a velocity stream interaction region separating faster, less dense and cooler plasma from slower, denser and a bit hotter plasma. Obviously, this is not a solar wind stream interaction region separating a fast stream of coronal hole origin from a slow solar wind (Richardson, 2018, e.g.), but rather, a velocity shear generated by the moving and possibly expanding MC interacting with the ambient plasma in the sheath. Part of this interaction happens at the MC boundary where the MR outflow is moving faster than the MC. The denser and hotter plasma carried by reconnected open field lines piles up and slows down at D1A stream interface (Figures 7a-e, Table 2 left). Actually, this could be an SS if the velocity was not increasing towards the upstream direction. However, the vertical flow shear seen in $V_{Z}$ and also to a less extent in $V_{Y}$ (Figure 7d and also in Figure 2e) can mask the real velocity jump conditions across D1A. This ’masking’ does not allow to calculate reliably the SS speed. For this reason we identify D1A as an SS-like discontinuity embedded into the flow stream. The angle $\Theta_{Bn}$ between $\mathbf{B}$ and $\mathbf{n}$ is $\sim$67 ^∘ which corresponds to a quasi-perpendicular shock geometry.

Geotail basically observes the same $|B|$ profile at D1G as ACE at D1A (Figure 7f). The time delay between the discontinuities D1A, D1G and the left border of the current sheet is $\sim$ 48 minutes (Figures 7a, f). Since Geotail is almost at the same GSE Y as ACE (Figure 6a), the distance to the X-line at both spacecraft presumably is not very different, which can result in similar $|B|$ profiles. D2A is also roughly at the same distance from the right boundary of the current sheet as D2G, though D2G is less pronounced. Since plasma data are not available from Geotail, the D1G and D2G discontinuities are not analyzed further.

Wind observes D1W only 11.5 minutes earlier than the left border of the outflow (Figures 7 g-k). Across D1W (Table 1, right) B decreases, $V$, Ni and Ti increase. As for velocities mainly $|V_{X}|$ and $V_{Z}$ increase. The normal components of the magnetic field are $\sim$ 15 nT, and $\Theta_{Bn}$= 12^∘ and 4^∘, respectively. These are the signatures of a quasi-parallel forward SS. The slow shock speed normal to the shock surface is estimated from the upstream and downstream values as $V_{sh}=\mathbf{n}[N_{d}\mathbf{V}_{d}\cdot\mathbf{n}-N_{u}\mathbf{V}_{u}\cdot\mathbf{n}]/(N_{d}-N_{u})$. Then the normal solar wind speeds in the shock frame are calculated as ${U_{n}}_{u}=\mathbf{V_{u}}\cdot\mathbf{n}-V_{sh}$ and ${U_{n}}_{d}=\mathbf{V_{d}}\cdot\mathbf{n}-V_{sh}$, respectively. Finally, the Alfvén-Mach number is estimated through ${M_{A}}_{u}={U_{n}}_{u}/(V_{A}cos\Theta_{Bnu})$ and ${M_{A}}_{d}={U_{n}}_{d}/(V_{A}cos\Theta_{Bnd})$ (Table 2, right). The small values of $M_{A}=$ 0.26 and 0.2 indicate that D1W is a weak SS. We note that the local flow shear masking the SS-like discontinuity (D1A) has a much larger amplitude than the flow shear observed by WIND at the weak SS (D1W). This is the reason why the SS could be observed by WIND.

Now we consider the inflow region after the right boundary of the outflow (location of $\mathbf{B2}$ vector in Figure 6) which is inside the boundary layer of the MC. The variables in upstream and downstream regions across D2A and D2W are shown in Table 3 left and right.

Across D2A $|\mathbf{V}|$ decreases, the largest changes are in $V_{Z}$ and $V_{Y}$ (Figures 7c, d and Table 3, left). Ti increases by 0.2 eV only. There is a density dip between D2A and the right boundary of the MR outflow. Instead of the sudden Ni jump as it was across D1A, the change from 4 to 10 cm^-3 is rather smooth across D2A (Figure 7e). In the $\pm$ 10 minute vicinity Ni changes from 7 to 9 cm^-3 (Table 3, left). Since Ni increases by 1.3-2.5 times D2A is not a rotational discontinuity in the low beta plasma. Since the normal component of the magnetic field is $\mathbf{B_{n}}\sim 15.9$, and $\Theta{Bn}$=8.9^∘ and 1.8^∘, respectively, D2A is not a tangential discontinuity. D2A is associated with a similar velocity stream interaction region as D1A with an embedded quasi-parallel SS-like discontinuity. Therefore, the shock speed cannot be reliably estimated.

Across D2W $\mathbf{|B|}$ slightly decreases, Ni increases 2.5 times, Ti increases by 0.1 eV, $\mathbf{|V|}$ decreases, however, again the transverse flow speeds $V_{Y}$ and $V_{Z}$ show a flow shear at the discontinuity (Figures 7g-k and Table 3, right). $\Theta_{Bn}$ is 14^∘ and 6^∘, respectively. We classify D2W as a quasi-parallel SS-like discontinuity embedded into a flow stream interface.

In the following the MR outflow structure is studied in more detail.

Figure 8 shows WIND observations of MR outflow (red vertical dashed lines at 16:46:15 and 16:58:12 UT) in a 1-hour long time interval (between 16.25 and 17:25 UT) in the MVA rotated system (Figure 6c) on 2000-10-03. The discontinuities D1W and D2W are indicated by dashed yellow vertical lines and marked as SS. Although D2W, as described above, is more an SS-like discontinuity in a flow stream, it is also marked as SS. The top three subplots 8a-c contain the pitch-angle (e-Pa) distribution of 255 eV electrons, the omni-directional electron fluxes for energies from 27 to 520 keV, the omni-directional proton/ion fluxes for energies from 0.4 to 3 keV. Figures 8 d,e show the amplitude and the LMN components of the magnetic field. The ion density and temperature (Ni and Ti) are in the same subplot 8f with the same Y axis in $cm^{-3}$ and eV, respectively. The GSE Vx, Vy and Vz ion velocity components are in Figures 8g and h. With the average ambient solar wind speed removed the LMN ion velocity components are shown in Figure 8i. Additionally, the perpendicular and parallel LMN directional velocity components to the local magnetic field are calculated (Figures 8j,k). The local proton Alfvén velocity is also added to the subplot 8j. Finally, in the last subplot 8l, the on-board calculated electron velocity magnitude $|Ve|$ is shown without the strongly fluctuating components. The LMN components of the low time resolution (one vector/98 s) ground calculated electron velocity are shown in the same subplot 8l. From the electron velocities the average ambient electron speed was also removed.

The MR outflow is as expected in fluid models: simultaneous correlated/anticorrelated rotations of magnetic field and velocity components at the borders (at red dashed vertical lines, Figures 8e,i), inside the outflow decreased $|B|$ (Figure 8d), increased Ti and Ni (Figure 8f), and enhanced $\mathbf{L}$ directional velocity $V_{L}$ (Figure 8i). The component of the velocity perpendicular to the magnetic field $V_{\perp L}$ is almost reaching the local Alfvén speed $V_{A}$ (Figure 8j). In papers by Wang et al. (2016) and by Zhao et al. (2019) the so-called Walén test has been successfully verified for this event, which is a fluid level test for tangential stress balance for a rotational discontinuity across a reconnecting current layer. Across the outflow boundaries (red dashed vertical lines) both $\mathbf{V_{i}}$ and $\mathbf{B}$ rotate, additionally $\mathbf{V_{i}}$, $\mathbf{V_{e}}$ and $T_{i}$ increase towards the outflow center (Figures 8e-h, l). Also, $N_{i}$ and $N_{e}$ (see below) clearly increase at the right outflow boundary (towards the center of the outflow at 16:58:12 UT). Such simultaneous jumps and rotations can correspond to a merged Alfvén discontinuity and slow-shock (Sasunov et al., 2012). However, the exact jump conditions and discontinuity normal directions cannot be determined because of the high variability of plasma and field data. Since there are merged Alfvén discontinuities and SSs at the boundaries of MR outflow, it is not obvious that the discontinuities D1W and D2W in the inflow regions are Petschek-type SSs. In fact, by using a compressible resistive MHD model with sub-Aflvénic shear flow across the MR outflow it was shown that for low-$\beta$ plasmas a pair of slow shocks (SSs) is generated in the inflow region away from separatrices (Li et al., 2012). These SSs in the inflow region in simulations are not the Petschek-type shocks at the separatrices. In what follows we will refer to the standalone SSs (D1W and D2W in Figure 7) in the inflow regions indicated by the yellow vertical dashed lines as left SS (at 16:34:33) and right SS (at 17:17:42 UT), respectively. Here left and right just correspond to the time sequence of events, not the spatial locations.

Before continuing with a more detailed description of the outflow structure we shortly describe, without showing, the parameters as they would be in the MVA system (Figure 6b). The L components of variables would not change. The M and N components would be exchanged. For example, ${V_{\perp}}_{N}$ and $V_{||N}$ (Figures 8j and k) would be exchanged by ${V_{\perp}}_{M}$ and $V_{||M}$. It is easy to notice that, in the MVA rotated system (Figure 6c and the whole Figure 8), at the left and right boundaries of the outflow, the $\mathbf{N}$ directional parallel (minus then plus) and perpendicular ion speeds (in green) are enhanced and there is a small positive $V_{||L}$ (Figures 8j and k). In fact, these are the MR inflow ion signatures near separatrix when $B_{N}$ is large (Figure 8e). Similar inflow signatures are seen also in simulations (e.g. Lapenta et al., 2015). However, having the MVA coordinates (Figure 6b), the enhanced ${V_{\perp}}_{M}$ and $V_{||M}$ components at the outflow boundary would be nonphysical. This seems to support the MVA rotated configuration (Figure 6c.1). We add that by rotating the coordinate system around L in smaller steps (eg. by 10^∘) from the vertical geometry (Figure 6c.1) in clockwise direction Bg would increase in negative direction while in case of counter-clockwise rotation Bg would increase in positive direction.

The right boundary of the outflow at 16:58:12 UT is attached to the MC boundary (time interval indicated by horizontal arrow line, marked as MC over the top panel) which can be readily seen from the appearance of bi-directional suprathermal electrons of 255 eV (e-Pa Figure 8a). The same MC boundary can also be seen in ACE e-Pa 272 eV distributions in Figure 2g. Earlier in time , before the MC boundary, predominantly the parallel electron population is seen on open field lines, with a smaller flux of anti-parallel electrons starting at the left SS (Figure 8a). The closed field lines in MC, rooted on the Sun, carry 0.5 MeV low-flux population of electrons of solar origin (Figure 8b). There is also an enhanced electron flux over 30-120 keV between 17:00 and 17:04 UT (Figure 8b). The ion energy spectrogram (Figure 8c) shows two population of ions, one centered at $\sim$950 eV and another centered at $\sim$1900 eV. Inside the MR outflow, before 16:50 UT, both populations are accelerated $\sim$1.4 times. This indicates that both populations of ions are protons. The higher energy proton population appears downstream of the left SS (Figure 8c) where also the low flux anti-parallel electron population is generated (Figure 8a). Similar intensification of the proton flux is seen downstream of the right SS. It is easy to notice that near the left and right quasi-parallel SS (D1W and D2W in Figure 7 and in Tables 2 and 3) there are enhanced fluctuations in $\mathbf{B}$ components (Figure 8e). Magnetic turbulence or waves at quasi-parallel SSs can be responsible for particle acceleration or/and reflection. The more energetic protons, possibly on the freshly reconnected open field lines, can propagate from their source within the MR outflow until the left SS where they can be back reflected. Alternatively or concurrently, the reflected protons can propagate back to the outflow where they are further accelerated by MR electric field together with the low energy proton population. The pitch angle distribution of protons in the keV range is not available. At the right SS a similar increase in the flux of protons/ions in the downstream region is observed, however, the source of the seed population of particles is not clear.

More importantly, there is an asymmetry at the left and right boundaries of the MR outflow including the vicinity where the discontinuities are observed (at $\mathbf{B1}$ and $\mathbf{B2}$ vectors in Figure 6). The left side is located in the ICME sheath, while the right side is in the boundary layer of MC. Let’s consider the left side of the outflow first.

The left SS (D1W above) is associated with a pile-up of plasma and reflection of protons back towards the outflow. Here Ni is increased between the left SS and the left boundary of the outflow.

Both ACE and WIND observe increased Ni after the left SS with slowly decreasing values towards the left outflow boundary (Figures 7c, g). As a result, along WIND crossing, $N_{r}$=$<$Ni(outflow)$>$/$<$Ni(inflow)$>$ $\sim$ 1.1 near the left boundary of the outflow, where $<$Ni(inflow)$>$ is the average ion density in the inflow region and $<$Ni(outflow)$>$ is the average ion density inside the outflow near its left boundary. Possibly, as the left SS is getting farther or closer to the outflow border, $<$Ni(outflow)$>$/$<$Ni(inflow)$>$ can change and have different impact on the outflow. Following this line of thought, a bundle of freshly reconnected field lines can form a flux tube or flux rope (Semenov et al., 2012; Sasunov et al., 2012; Vörös et al., 2014). When we consider a bundle to form a twisted flux rope embedded into ambient plasma with a small external twist, Kelvin-Helmholtz (KH) instability can develop even for sub-Alfvénic boundary flows (Zaqarashvili et al., 2014). The criterion for KH instability in such a case is $|m|M_{A}^{2}>1+N_{r}(2+|m|)/|m|$, where $m$ is the azimuthal mode number of KH vortices and $M_{A}$ is the Alfvén Mach number inside the flux rope/outflow. In this context, $|m|$ is the number of vortices which can develop around a flux rope. The analytical and numerical calculations leading to the above instability criterion start from cylindrical flux ropes moving along their axis in an ambient plasma with and without external magnetic twist. Without the external twist, e.g. when the flux ropes move along the ambient magnetic field, that external field appears as an axial one, the KH instability is suppressed for sub-Alfvénic flow shear. In our case the internal field inside the current sheet rotates and the external ambient field has a different direction than the $\mathbf{L}$ directional outflow (Figures 8e, i). In such a case the external field has a transverse component and KH instability can develop for a sub-Alfvénic flow shear (Zaqarashvili et al., 2014). Since $M_{A}\sim 0.7$ at the left boundary of the outflow and Nr$\sim$1.1, for azimuthal mode numbers $|m|$=5 the KH instability criterion is fulfilled. In fact, between 16:43 and 16:46 UT low-frequency fluctuations in magnetic field (Figure 8e) and plasma parameters (Figures 8f, h-j) are observed. Although the size and the geometry of the flux rope bundle is not known (probably the flux rope is not a cylinder), the fluctuations clearly seen in Ti and Ni (Figure 8f) 1-3 minutes before the left outflow boundary could be associated with KH vortices for the sub-Alfvénic flow shear. More importantly, the left SS generating the elevated density Ni(inflow) and therefore smaller ratio $N_{r}$ can result in KH unstable outflow boundary with smaller $|m|$ (larger size structures). This can modify the outflow structure through mixing the higher Ni and Ti outflow plasma with the lower Ti and Ni sheath plasma. This is not the case at the right outflow boundary where $N_{r}\sim$3 (Figure 8f).

Let us consider now the right side of the outflow. The right SS-like discontinuity (D2W above) with more pronounced transverse flow shear (Figure 8h) seems to reflect ions (Figure 8c) and increase density downstream of SS (Figure 8f). However, there is a more pronounced density cavity between the right boundary of the outflow and right SS (Figure 8f). The low Ni can appear as a result of pressure balance when MR outflow interacts with the ambient magnetic field compressing it. A similar density cavity has already been observed in magnetotail reconnection associated with interactions between MR outflow and ambient magnetic field in the magnetotail plasma sheet (Cai et al., 2009). In our case, the cavity after the right boundary of MR outflow is observed by both WIND (Figure 7k) and ACE (Figure 7e). WIND observes a longer low density region, Ni around 5cm^-3, roughly 20 minutes. The long duration density cavity might also been explained by the slow motion of WIND across the structure. Although $\mathbf{|B|}$ slowly increases in the density cavity, it is unlikely that this can alone generate the observed density cavity due to pressure balance. When the current sheet configuration is vertical (Figure 6c) the right boundary of the outflow is slowly convected in +Z GSE direction with $V_{Z}\sim$ 6 km/s (Figure 7j). The fast motion with the bulk solar wind speed in GSE-XY directions moves the vertical structure mainly in the L-M plane (Figure 6c.1), therefore, WIND possibly remains near the separatrix surface where the low Ni is observed. We note the separatrix is a line in 2D geometry while it is a surface in 3D case.

At the outflow boundaries the high-frequency field and plasma fluctuations are enhanced. Although in MR separatrix parallel electric fields, waves, instabilities and kinetic physics are expected (Khotyaintsev et al., 2020; Hesse et al., 2018), the limited time resolution of field and plasma measurements makes the investigation of sub-ion and electron scale structures more difficult. Nevertheless, we demonstrate here that WIND can observe signatures of electron-scale physics in MR outflow.

First, we concentrate on the narrow structure occurring inside the outflow at 16:48:39 UT inside the magenta dashed box in Figure 8. At and around that time, the accelerated protons reach their maximum energy (Figure 8c); $B_{L}$ changes sign sharply decreasing from 5 to -12 nT, $B_{N}$ decreases sharply from 12 to 6 nT, $B_{M}$ changes from -6.5 to 0 nT; the guide field Bg $\sim$ -2 nT (Figure 8e); Ni decreases and Ti increases (Figure 8f); $V_{\perp N}$ which is predominantly negative within the outflow locally reaches +40 km/s (Figure 8j); $V_{||N}$ reaches locally -70 km/s, $V_{||L}$ +100 km/s (Figure 8k) and the magnitude of onboard calculated electron velocity $|V_{e}|\sim 220$km/s which is by 130 km/s larger than the ambient electron outflow velocity (Figure 8l). The lower resolution ground-calculated ${V_{e}}_{L}$ has the same peak as the onboard computed $|V_{e}|$. The accelerated protons, the increased parallel velocities $V_{||L,N}$ and increased $|V_{e}|,{V_{e}}_{L}$ in Figure 8 indicate that there might be a local parallel electric field $E_{||}$ which accelerates particles and therefore generates currents. In principle, the current density could be determined from the curl of $\mathbf{B}$ through $J_{L}\sim-\partial B_{M}/\partial N+\partial B_{N}/\partial M$, however, the field and plasma parameters, including the Hall-field ($B_{M}-Bg$) fluctuate wildly. Turbulent Hall fields associated with MR have already been observed in the solar wind (Xu et al., 2015). Qualitatively, the currents responsible for the Hall field can be generated by electrons flowing along the separatrices, ie. near the outflow border, and at the narrow current structure (in dashed magenta box) where proton $V_{||L}$ (Figure 8k) and $Ve_{L}$ (Figure 8l) are large. The latter narrow current structure occurs closer to the left border of the outflow (closer to $\mathbf{B1}$ in Figures 6b or 6c). The deflection of the $J_{L}$ current from the midplane of the outflow has been reported in simulations, (Goldman et al., 2011), far from the X-line in the solar wind (Xu et al., 2015; Mistry et al., 2016) and in magnetotail reconnection (Zhou et al., 2014). For $Ve_{L}>V_{||L}>$0 the current is in negative $\mathbf{L}$ direction, pointing towards the X-line. The deflection of electrons is supposed to occur due to the Lorentz force in the presence of a weak guide field given by -q.$Ve_{L}\times Bg$, where q is the electric charge. Since Bg = -2 nT (Figure 8e, Bg/$|\mathbf{B}|\sim$ 0.13) and $Ve_{L}$ is positive (in the magenta dashed box, Figure 8i), the Lorentz force deflects the electron jet towards the left boundary of the outflow. For the vertically oriented current sheet with Bg in $\mathbf{-M}$ direction (Figure 6c.1) it is the boundary at $\mathbf{B1}$. However, it is also valid for the MVA LMN coordinates when LN is in GSE-XY plane in Figure 6b. In that case the large positive guide field Bg$\sim$7.5 nT (Figure 5b, Bg/$|\mathbf{B}|\sim$ 0.5) ensures that the Lorentz force deflects the electron jet again towards $\mathbf{B1}$.

The current layers associated with deflected electron flows are expected to be narrow of 0.8-6 di (Mistry et al., 2016). di stands for ion inertial length. Since di$\sim$ 100 km in our case, when the LN plane is in GSE-XY (Figure 6b), the thin electron flow structure should be convected with $V_{SW}$ velocity in less than 2 s across WIND. WIND does not have a good enough temporal resolution to see such a thin electron flow. Nevertheless, it is observed, as shown within the dashed magenta box in Figure 8l. However, in the MVA rotated system the LM plane is in the GSE-XY plane (Figure 6c) and the vertical convection speed of the structure is $\sim$16 km/s (Figure 7j) or less. With this speed the electron flow structure could be observed by WIND for more than 30 s, provided that the flow channel is more extended along the X-line, in out-of-plane direction. It is worth noticing that simultaneously with the occurrence of the narrow electron and ion flows in the dashed magenta box, $B_{L}$ transiently changes sign and $B_{N}$ decreases (Figure 8e). This might be the signature that the thin flow structures locally change the magnetic field geometry inside the outflow. Interestingly, Geotail, which is at a larger distance from the X-line as WIND, observes the sign-changing $B_{L}$ roughly 10 s earlier (Figures 5d, e). If we accept that the transient short sign-change of $B_{L}$ is the signature of the interaction of the flow with the magnetic field then the electron flow is possibly even more deflected towards the left border of the outflow at Geotail location, which is in a larger distance from the X-line.

There are other kinetic effects observed by WIND which can also shape the MR outflow structure. The right boundary of the outflow was previously classified as a merged slow shock and Alfvén discontinuity at the MHD level of description. Now we consider possible kinetic effects at the same boundary.

In Figure 9a the MVA rotated LMN components of $B$ and its magnitude are shown. Figure 9b contains Ni and Ne, the latter estimated from the WAVES experiment (based on electron plasma frequency) on WIND. Ti and Te are shown in Figure 9c. The electron temperature anisotropy $Te_{||}/Te_{\perp}$ is depicted in Figure 9d. Figure 9e shows again the omnidirectional electron energies in keV-MeV range and in Figures 9f-j are the pitch angle distributions (e-PA) for energy ranges from 40 to 310 keV. The frequency spectra of the average electric field in dB above the background and the normalized average voltage from the Thermal Noise Receiver (TNR) experiment on Wind/WAVES are shown in Figures 9k and l, respectively. The outflow is again indicated by red vertical dashed lines.

The magenta dashed box in Figure 9 indicates the time interval between the right end of the outflow and right SS. At the beginning of this time interval the electric field fluctuations are enhanced and wave activity is present in Figures 9k and l. This is the time interval where both Ni and Ne are low and, as it was explained above, the outflow interacts with the ambient plasma and magnetic field of the MC. While Ti is increased within the outflow, Te is enhanced in the density cavity (Figure 9c) where there is also a clear temperature anisotropy $Te_{||}/Te_{\perp}\sim$1.5 (Figure 9d). This indicates that electron heating is preferentially in parallel (or anti-parallel) direction to the magnetic field. In the density cavity the e-PA distributions for 40 and 86 keV energy range show perpendicular populations of electrons (Figures 9f,g). The local enhanced flux over this energy range is also clearly seen in the omnidirectional electron energy spectrum (Figure 9e). Over the 180-310 keV energy range the electrons are predominantly anti-parallel (Figures 9i,j) with a smaller flux in the parallel direction. There is also an isotropic electron background population with smaller flux which is already present inside the outflow near its right boundary at higher electron energies (Figures 9i, j). For the 180 - 310 keV electrons the e-PA is bi-directional indicating the presence of closed field lines in the MC. We have also checked a 3-hour long time interval after the outflow and bidirectional e-Pa distributions associated with MC closed field lines are present for the energy range of 110-520 keV (not shown). The strongest fluxes of anti-parallel electrons occur in the density cavity. We have also checked the lower energy electron e-PA distributions in the wider time interval. Although there exist perpendicular or more isotropic electron populations for energies 27-86 keV even 3 hours later (not shown), the perpendicular electron population in the density cavity close to the right outflow boundary (Figures 9f, g) has the largest flux. Also, the density cavity is the only place during the 3 hour long wider time interval (not shown), where the electric field fluctuations peak around 10 kHz (Figure 9k) and where wave activity is observed over the frequency range of 5-20 kHz (Figure 9l).

Now we provide a possible physical explanation for e-Pa distributions and wave activity seen in the density cavity and before, inside the outflow (Figure 9).

Let us start with the observations of electron acceleration in the Earth’s magnetosphere. For example, in the Earth’s magnetotail electrons accelerated up to $\sim$300 keV have been observed during a crossing of MR diffusion region (Øieroset et al., 2002). However, electron acceleration by MR in the magnetotail can be a multi-stage process, when electrons are accelerated to a few keV in the vicinity of the X-line by MR electric field, and further nonadiabatically accelerated to tens of keV in the outflow pile-up region due to $\mathbf{\nabla}B$ or/and curvature drift where the electron flux is also substantially increased (Wu et al., 2015).

In our case, from one-point measurements the magnetic field curvature or $\mathbf{\nabla}B$ cannot be estimated. We provide a qualitative description of $\mathbf{\nabla}B$-drift and a rough estimate of the spatial scales associated with a possible electron acceleration.

Near the right boundary of the outflow $\mathbf{|B|}$ increases from 11 to 14 nT in $\sim$35 s, which is mainly due to an increase of $\mathbf{|B_{L}|}$ from 5 to 14 nT (Figures 8 e,f). When the current sheet normal is vertical (Figure 6c) and the right boundary of the current sheet moves slowly in vertical direction ($V_{Z}\sim$6$\pm$6 km/s in Figure 7j), the $\sim$35 s time interval corresponds roughly to 200$\pm$200 km spatial scale. Since the gyroradius of 40-310 keV relativistic electrons is between 50 and 200 km, comparable to the spatial scale of $\mathbf{\nabla}B$, nonadiabatic chaotic motion of electrons can be expected (Wu et al., 2015). This can be readily seen in the e-PA spectra of 180-310 keV electrons of larger gyroradius (Figures 9i, j), where an isotropic background distribution, possibly due to the chaotic motion, is visible near the right end of the outflow.

When $\mathbf{|B_{L}|}$ increases by 9 nT at the right outflow boundary, GSE $\mathbf{|V_{X}|}$ decreases from 430 to 390 km/s (Figure 8g), GSE $\mathbf{V_{Y}}$ decreases from 50 to -5 km/s and GSE $\mathbf{V_{Z}}$ changes from -20 to 20 km/s (Figure 8h). Although the rotation of the velocity direction is an indication for the Alfvén discontinuity, however, $\mathbf{|V|}$ locally decreases by $\sim$35 km/s. This indicates that when the outflow interacts with the ambient plasma and magnetic field of the MC, the plasma inside the outflow near the right outflow boundary decelerates and the magnetic field is compressed. The largest compression occurs near the right outflow boundary, but $\mathbf{|B|}$ continues increasing roughly until the right SS (Figure 8d), within the whole density cavity. As the kinetic energy of the plasma decreases particles are energized at $\mathbf{\nabla}B$ structure. Due to the changing velocities it is not possible to estimate precisely $\mathbf{\nabla}B$ during this time interval from single-point measurements. In case of the vertically oriented current sheet (Figure 6c) $|B_{L}|$ ($=\mathbf{B_{2}}$) is increasing in $\mathbf{-N}$ direction. Due to $\mathbf{\nabla}B$-drift the electrons move into $\mathbf{+M}$ out-of-plane direction. In the density cavity the magnetic compression is smaller, however, $\nabla B$ - drift can still generate a current and the electrons and protons move in opposite but perpendicular directions to $\mathbf{B}$ and $\mathbf{\nabla}B$. This can explain the occurrence of the perpendicular population of accelerated electrons near the outflow right boundary in Figures 9f and g. However, the largest flux of perpendicular electron populations is seen 2-6 minutes after the right outflow border or the strong $\mathbf{\nabla}B$ structure (Figure 9d,e). The question is what is the real spatial distance between the location of the right outflow boundary and the location of perpendicular electrons (2-6 minutes later in time) or the right SS. How wide is the magenta dashed box in space with $\sim$20 minutes duration in time? Here again, the geometry of the current sheet matters (Figure 6b versus Figure 6c). When the outflow structure is convected vertically in GSE Y-Z direction (case Figure 6c) WIND can remain close to the right flow boundary or near the 2D separatrix for longer time. The average GSE $V_{Z}$ near the right outflow boundary is 6 km/s, however, fluctuating (Figure 7j). Within 2-6 minutes the outflow border / separatrix can move to the distance of 2000$\pm$2000 km in vertical direction, within 20 minutes the width of the magenta dashed box is 7000$\pm$7000 km. These are certainly rather rough estimates of the spatial distances. We note that for the geometry in Figure 6b, the distance of the location of perpendicular electron population to separatrix (where $\mathbf{\nabla}B$ is significant) would be 50000-100000 km (8-16 $R_{E}$) and the width of the magenta box would be 516000 km (81 $R_{E}$).

We suggest here that both the perpendicular and the anti-parallel relativistic electron populations can be the result of a multi-stage acceleration by MR. In the vertical configuration (Figure 6c) the 20 minute long magenta dashed box can coincide with the reconnection inflow region, spatially close to the 2D separartix. While the lower energy electrons react to the $\mathbf{\nabla}B$ structure which is strong near the right outflow boundary, the higher energy electrons are anti-parallel, which at the right outflow boundary (at $\mathbf{B2}$ in Figure 6c) means propagation from the X-line. Although the details of multi-stage electron acceleration are not clear, the enhanced electric field fluctuations and wave activity in Figures 9i and j can be associated with the relativistic electrons.

The WIND/WAVES experiment allows to detect high-frequency plasma waves between local ion (fpi) and electron (fpe) plasma frequencies. In this frequency range, by using the TNR spectra and the electric waveforms from the Time Domain Sampler (TDS), Huttunen et al. (2007) observed Langmuir waves, electron solitary waves and Doppler shifted ion acoustic waves (around $\sim$4 kHz) inside the MR outflows, preferentially near the boundaries of MR outflows in the solar wind. Since TDS waveforms are not available for the event under study we cannot distinguish between Langmuir and upper hybrid waves, both occurring near fpe which is about $\sim$20 kHz in the magenta dashed box in Figure 9j. The waves observed below fpe down to 5 kHz in TNR spectra are supposedly Doppler shifted ion-acoustic waves (Huttunen et al., 2007). Lower frequency waves cannot be detected by the WIND/WAVES experiment. According to 2D particle-in-cell simulations of antiparallel MR (Fujimoto, 2014) the Langmuir waves can be generated by intense parallel electron beams triggering the electron two-stream instability. In these simulations an electrostatic potential is created due to electron-ion decoupling near the separatrix - inflow region forming the Hall current system supported mainly by electron beams. In the potential well the electrons are accelerated and the density decreases to conserve the mass flux. As a result, a density cavity can be formed in the inflow/separatrix region (Fujimoto, 2014). In the simulation the 2D box was 131 $\times$ 65.5 di only, i.e. the distant MR outflow and separatrix/inflow regions were not studied. In our case however, there exist Hall-field signatures. The antiparallel electrons in the right separatrix/inflow region (at $\mathbf{B2}$ in Figure 6c) indicate a current flowing towards the X-line which should generate a positive Hall magnetic field in the inner part of the right outflow boundary. Indeed, Figure 9a (or Figure 8e) demonstrates that relative to the small Bg, B_M is fluctuating, but positive. As it was noticed above, it was unlikely that the increased $\mathbf{B}$ and the pressure balance could explain the occurrence of the density cavity in the inflow region. In the presence of the Hall current system the subsequent acceleration of electrons in the potential well and the requirement for mass flux conservation can explain the generation of the density cavity. Alternatively, the electron beams can trigger the Buneman instability resulting in waves with frequencies near the lower hybrid range (Fujimoto, 2014) (out of the range of WIND/WAVE). The Buneman instability is expected to be excited for stronger guide fields (Drake et al., 2003).