Issue 
A&A
Volume 595, November 2016



Article Number  A84  
Number of page(s)  12  
Section  The Sun  
DOI  https://doi.org/10.1051/00046361/201628071  
Published online  01 November 2016 
Role of electron inertia and reconnection dynamics in a stressed Xpoint collapse with a guidefield
School of Physics and Astronomy, Queen Mary
University of London, Mile End
Road, London, E1
4NS, UK
email: j.n.grafvonderpahlen@qmul.ac.uk
Received:
3
January
2016
Accepted:
18
June
2016
Aims. In previous simulations of collisionless 2D magnetic reconnection it was consistently found that the term in the generalised Ohm’s law that breaks the frozenin condition is the divergence of the electron pressure tensor’s nongyrotropic components. The motivation for this study is to investigate the effect of the variation of the guidefield on the reconnection mechanism in simulations of Xpoint collapse, and the related changes in reconnection dynamics.
Methods. A fully relativistic particleincell (PIC) code was used to model Xpoint collapse with a guidefield in two and three spatial dimensions.
Results. We show that in a 2D Xpoint collapse with a guidefield close to the strength of the inplane field, the increased induced shear flows along the diffusion region lead to a new reconnection regime in which electron inertial terms play a dominant role at the Xpoint. This transition is marked by the emergence of a magnetic island – and hence a second reconnection site – as well as electron flow vortices moving along the current sheet. The reconnection electric field at the Xpoint is shown to exceed all lower guidefield cases for a brief period, indicating a strong burst in reconnection. By extending the simulation to three spatial dimensions it is shown that the locations of vortices along the current sheet (visualised by their Qvalue) vary in the outofplane direction, producing tilted vortex tubes. The vortex tubes on opposite sides of the diffusion region are tilted in opposite directions, similarly to bifurcated current sheets in oblique tearingmode reconnection. The tilt angles of vortex tubes were compared to a theoretical estimation and were found to be a good match. Particle velocity distribution functions for different guidefield runs, for 2.5D and 3D simulations, are analysed and compared.
Key words: magnetic reconnection / magnetic fields / plasmas / Sun: flares / Sun: magnetic fields / Sun: coronal mass ejections (CMEs)
© ESO 2016
1. Introduction
Magnetic reconnection is an important process in the study of solar plasma physics; it allows energy stored in magnetic fields to be converted into kinetic energy of superthermal particles and heat. Following Dungey’s original model of the open magnetosphere, where magnetic reconnection facilitates the interaction of the Sun’s magnetic field and the Earth (Dungey 1953), numerous models have been devised that use magnetic reconnection to explain solar eruptions (Shibata & Magara 2011; Chen 2011), coronal heating (Cranmer 2009; De Moortel & Browning 2015), and other energetic processes within the heliosphere. While no model of magnetic reconnection has currently been demonstrated to comprehensively describe energetic phenomena, several observational studies have found evidence for magnetic reconnection, such as great energy conversions in the heliospheric currentsheet, corresponding with fast reconnection models (Gosling et al. 2007), and signatures of Hallreconnection in the geomagnetic tail of the Earth (Eastwood et al. 2007). Furthermore, a recent study using magnetohydrodynamic modelling driven by solar magnetograms (Jiang et al. 2016) has found substantial new results regarding the transition from preeruptive to eruptive states in a magnetic fluxemerging region. A strong case thus exists for the importance of magnetic reconnection in solar processes; a thorough understanding of the magnetic reconnection process is vital to the further understanding of the dynamics within the heliosphere.
Dungey’s original analysis of magnetic energy conversion in the Earth’s magnetosphere describes a reconnection model now known as Xpoint collapse. Following reconnection models proposed by Sweet and Parker (Parker 1957) and later by Petshek (Petschek 1964), reconnection setups of sheared magnetic fields relying on the tearingmode became the dominant setup in computational studies of magnetic reconnection. As established in Tsiklauri & Haruki (2007, 2008) particleincell (PIC) simulations of Xpoint collapse in the collisionless regime exhibit many of the established features of tearingmode magnetic reconnection, e.g. the formation of a current sheet, magnetic Hall field generation, and independence of system size. However, simulations of Xpoint collapse have also uncovered several new features, such as initial oscillatory reconnection and vortex formation in the high guidefield regime (Graf von der Pahlen & Tsiklauri 2014a) and a distinct octupolar outofplane magnetic field (Graf von der Pahlen & Tsiklauri 2014b, 2015), which makes this setup a useful device for the ongoing study of magnetic reconnection. In this study we extend the results of Graf von der Pahlen & Tsiklauri (2014a) where collisionless Xpoint collapse with a magnetic guidefield was investigated in a 2.5 PIC simulation. In particular, this study analyses the reconnection mechanism (i.e. the term breaking the frozenin condition) and relevant plasma dynamics for increasing guidefields.
Fig. 1
Reconnection at an Xpoint in a 3D domain, showing the motion of two sets of magnetic fieldlines. The perspectives on the simulation domain are indicated for each row. in panel b) the direction of the reconnection electric field induced as the fieldlines pass the Xpoint is indicated. Thin arrows on panels c) and f) indicate the direction of the electron current generated by the reconnection electric field. As shown, the shape of the fieldlines guides the accelerated particles such that there is a shearflow in the xyplane. 

Open with DEXTER 
The reconnection rate in a 2D reconnection setup can be defined as the outofplane electric field where magnetic separatrices meet, i.e. at the Xpoint. The movement of magnetic fieldlines (representative of fluxtubes in 3D) in the xyplane corresponds to changes in the zcomponent of the magnetic vector potential, A_{z}, for a set gauge. Integrating out the inplane magnetic field components B_{x} and B_{y} over a given area allows values of A_{z} to be determined. The equipotential lines on a contour plot of A_{z} represent the inplane magnetic fieldlines, as for example in panels a) to c) of Fig. 1. When the magnetic field is frozen into the plasma then (1)where V represents the plasma velocity and B is the magnetic field. Thus, changes in the magnetic field are facilitated entirely by advection rather than diffusion. In the case of magnetic reconnection, the frozenin condition is broken and changes in A_{z} by definition result in the generation of an electric field, according to (2)Since the topology of fieldlines in 2D must change for reconnection to occur, fieldlines must pass through a nullpoint, thus making the outofplane electric field at the Xpoint a reliable measure of the reconnection rate. This process is shown in panels a) to c) in Fig. 1, where panel b) shows the electric field generated as fieldlines break and change topology.
In the collisionless regime, the diffusion region is dominated by electron dynamics. Therefore, a means of identifying the reconnection mechanism in collisionless 2.5D simulations is to identify the terms in the generalised electron Ohm’s law that sustain the outofplane electric field at the Xpoint, i.e. (3)where the terms on the righthand side are, from left to right, the advection term, the divergence of the electron pressure tensor, the time derivative of the electron bulk inertia, and the convective inertia (i.e. spatial derivative) term, and where ⟨v_{e}⟩ represents the mean electron particle velocity at the Xpoint. As shown by Hesse & Zenitani (2007) a relativistic version of this equation can be derived from the relativistic Vlasov equation and is given by (4)where P_{e′} = ^{∫}du_{e}((u_{e}u_{e}/γ)f − n_{e} ⟨u_{e}/γ⟩ ⟨u_{e}⟩) and u_{e} = γv_{e}, where γ is the Lorentz factor and f the electron velocity distribution function at the Xpoint. The reconnection mechanism in tearingmode reconnection setups has been investigated in many computational studies (Hesse et al. 1999, 2004; Horiuchi & Sato 1997; Pritchett 2001; Swisdak et al. 2005) and was consistently found to be the divergence of the electron pressure tensor. A recent exception to this trend is found in Melzani et al. (2014); the authors show that for tearingmode reconnection in relativistic conditions (i.e. where inflow magnetic energy exceeds plasma rest mass energy), convective inertia can make an approximately equal contribution to the reconnection electric field as the pressure tensor divergence. A further exception is found in Hesse & Zenitani (2007), where a sheared magnetic field setup was modelled with a relativistic electronpositron plasma, and contributions from the time derivative of the electron bulk inertia were observed. Swisdak et al. (2005) and Hesse et al. (2002) show that for increasing values of guidefield in a tearingmode setup, the convective inertial (spatial derivative) terms start to make an increasingly large contribution to the outofplane electric field adjacent to the current sheet. However, the contribution to the reconnection electric field at the Xpoint remained the divergence of the electron pressure tensor. In this study it is shown that similar results emerge in an openboundary Xpoint collapse setup, with the notable difference that for high enough guide fields the convective inertial terms can become asymmetric across the current sheet and shift to the Xpoint, becoming the dominant contribution to the reconnection electric field.
A 3D representation of reconnection at an Xpoint with an outofplane magnetic guidefield is shown in panels d), e), and f) in Fig. 1. As shown, reconnecting magnetic fieldlines now carry a vertical magnetic field component. As explained by Schindler et al. (1988), Priest et al. (2003), fluxtubes in 3D do not necessarily have to pass through an Xpoint/Xline in order to undergo reconnection. In 2.5D simulations however, all reconnecting fieldlines must meet at the Xpoint, making this a representative model of the relevant dynamics. As the vertical components of the magnetic field are carried into the Xpoint, shown in panels d), e) and f) in Fig. 1, the outofplane electric field at the Xpoint is now partially parallel to the magnetic field (see panel b)) and thus accelerated electrons are “guided” along the fieldlines. Panels c) and f) of Fig. 1 show the resulting electron current. As shown in panel c) this electron current represents a shear flow in the xyplane. Kleva et al. (1995) discuss this effect and the resulting density asymmetry along the across the current sheet.
An alternative possible modification to a 2D tearingmode reconnection setup is the addition of a shear flow parallel or antiparallel to the inplane magnetic field. This reconnection setup has been studied by several authors (Mitchell & Kan 1978; Cassak & Otto 2011; Nakamura et al. 2008; Chacón et al. 2003; Jain & Büchner 2015) and is considered representative of reconnection and vortex formation in the magnetosheath (Otto & Fairfield 2000). For shear flows where the shear velocity is below the Alfvén speed, the reconnection rate is shown to be to be inhibited by greater shear flows (Mitchell & Kan 1978; Cassak & Otto 2011). However, for shear speeds greater than the Alfvén speed it has been shown that the reconnection dynamics can be altered and the reconnection rate increased. Twodimensional simulation results by Nakamura et al. (2008) show that for large enough shear flows, tearingmode reconnection can be coupled with vortex reconnection, and in Chacón et al. (2003) the parameters necessary for the mixing of these two reconnection modes are mathematically established. Further, Jain & Büchner (2015) established the growth rates of tearing and nontearing modes in reconnection with an electron shear flow using a linear stability analysis.
In this study, for an openboundary Xpoint collapse setup with a guidefield close to the strength of the inplane field, we show that electron shear flows are generated that are strong enough to change the reconnection dynamics and alter the term that breaks the frozenin condition. It has previously been demonstrated that a large enough guidefield can lead to vortical electron flows and island formation in both tearingmode reconnection (Fermo et al. 2012) and in Xpoint collapse (Graf von der Pahlen & Tsiklauri 2014a). Here we demonstrate how these results can be enabled by the shear flow and the resulting convectional inertia contribution to the reconnection electric field, generated through guidefield reconnection. Furthermore, by exploring the same reconnection setup extended into the third dimension, we investigate how electron and vortex dynamics proceed in 3D. While there is no generally accepted method of identifying vortices in a fluid (Cai et al. 2015), for this particular type of investigation we promote the use of the Qvalue. The Qvalue represents a Galileantransformation invariant measure of vortical flow (Haller 2005; Chakraborty et al. 2005; Hunt et al. 1988), defined as the second invariant of the velocity gradient tensor, ∇v, given by (5)When positive at a given point in a domain it indicates the presence of a vortical flow at that location or, as originally stated in Hunt et al. (1988), in “eddy zones” more than about 3/4 of the area has Qvalues greater than 1. In this study the Qvalue is used to show that vortical flows in 2.5D simulations correspond to 3D vortex tubes with structures that are not apparent from the 2.5D simulations.
An additional possible feature, unique to 3D reconnection with a guidefield, is the generation of oblique modes as demonstrated in Liu et al. (2013), Baalrud et al. (2012) and Akçay et al. (2016). In 2.5D simulations of magnetic reconnection, reconnection must occur at a magnetic Xpoint where it is possible for fieldlines to change in topology. In a symmetric setup, this means reconnection occurs at the centre of the diffusion region (here along the x = 0 line). However, in 3D reconnection with a guidefield, a more generic requirement for reconnection applies: reconnection occurs on surfaces where k×B = 0, where k represents the wave vector of a perturbation associated with reconnection and B the magnetic field. In the case of a sheared magnetic field with a guidefield, extended over a 3D domain, this implies that reconnection sites may exist adjacent to the midplane of the diffusion region, generating current sheets at oblique angles, relative to the zdirection, of (6)The angle θ does thus correspond to the inclination of the outofplane magnetic field. In a sheared magnetic field reconnection setup, the strength of B_{y} increases with distance from the midplane of the diffusion region, meaning that reconnection sites further from the centre should lead to greater obliqueness. While the generation of oblique current sheets and fluxtubes has been demonstrated in 3D PIC simulations with tearingmode setups (Liu et al. 2013; Akçay et al. 2016), this study similarly demonstrates the oblique nature of vortex dynamics in 3D reconnection in an alternative reconnection setup, i.e. Xpoint collapse.
2. Simulation model
2.1. Stressed Xpoint collapse reconnection model
As in previous studies (see Graf von der Pahlen & Tsiklauri 2014a, 2014b, and 2015) the setup of the inplane magnetic field used in this study in known as Xpoint collapse and it is mathematically described by the expressions (7)where B_{0} is the characteristic magnetic field intensity, L is the characteristic lengthscale of reconnection, and α is the stress parameter (see e.g. Chap. 2.1 in Birn & Priest 2007). In this setup, B_{x} and B_{y} lie in the xyplane, while a uniform current, j_{z}, is imposed satisfying Ampere’s law, i.e. (8)In this scenario, for an initial stress parameter greater than unity, the magnetic field leads to a J × B force that pushes the fieldlines inwards along the Xdirection. This serves to increase the initial magnetic stress, which leads to an increase in j_{z}, which in turn increases the inwards force. Owing to the frozenin condition, this leads to a build up of plasma near the Xpoint. Given that conditions for magnetic reconnection are met, magnetic pressure does not counter the inwards force but dissipates through reconnection and the field collapses. This is accompanied by the formation of a current sheet.
As an additional modification, a uniform outofplane magnetic guidefield is imposed at the beginning of the simulation. The strengths of the guidefield are chosen to be fractions of the maximum field amplitude within the plane, B_{P}, i.e. (9)where n is an integer ranging from 1 to 6. When extended into the third dimension, this configuration was not altered, i.e. the initial values of current, density, and magnetic field do not vary with z.
2.2. PIC simulation setup
Following Graf von der Pahlen & Tsiklauri (2014a,b, and 2015), this study continues the investigation of Xpoint collapse using a relativistic and fully electromagnetic PIC code. While in previous studies the simulation runs were limited to 2.5D, in this study results were extended into 3D. The PIC code was developed by the EPOCH collaboration (Arber et al. 2015) and is based on the original PSC code by Hartmut Ruhl (Ruhl 2006), employing the Villasenor and Buneman scheme (Villasenor & Buneman 1992) to update simulation parameters. It is a kinetic and relativistic code and all the relevant physical quantities are represented, allowing for the calculation of all the terms in the generalised Ohm’s law (see Eq. (4)). As with all PIC codes, a simulation domain is initiated with a set of pseudo particles, representing multiple physical particles of a specified temperature and momentum. Furthermore, electric and magnetic fields are set up over a discrete grid of cells. After initialisation the code carries out a leapfrog algorithm where in turn fields on grid cells are updated by the particle motion and particle motion is updated by fields defined on grid cells.
The parameters in the simulation were chosen such that temperature, particle densities, and magnetic fields corresponded to observed values for coronal flaring loops (Shibata & Yokoyama 1999; Aschwanden 2005). Observational studies have found flaring temperatures to be in the range 10^{6} K to 10^{8} K, number densities of electrons 10^{15} m^{3} to 10^{17} m^{3}, and magnetic field strengths of the order of 0.01 Tesla. While the corresponding lengthscales of flaring processes range from 10^{6} m to 10^{8} m, PIC simulations using today’s technology do not have the capacity to simulate plasma over such vast scales and a reduced area is considered, focussed on the reconnection processes.
Accordingly in the simulation n_{e} = n_{p} = 10^{16}m^{2}, T_{e} = T_{p} = 6.0 × 10^{7} K, and the characteristic electron Alfvén speed as , fixing the magnetic field parameter as B_{0} = 0.03207 T, and satisfying v_{Te} = v_{ae0}, where v_{Te} is the electron thermal velocity. In order to computationally afford to run simulations with such particle densities, the mass of protons was set as 100 times the electron mass, i.e. m_{p} = 100m_{e}, to speed up the code. The initial stress parameter is set as α = 1.2, corresponding to a small initial compression of an Xpoint magnetic field. In the 2.5D case, lengths of grid cells were set as the Debye length, i.e. Δx = Δy = λ_{D} = v_{te}/ω_{pe}, over a grid of 400 × 400 cells, amounting to a system length of approximately four ion inertial lengths, i.e. 4c/ω_{pi}. While the system size does not extend to characteristic coronal lengths scales, it is large enough to capture both particle species dynamics. Five hundred pseudo particles per cell were used which was shown to be a suitable number in convergence tests. When extended into 3D, the height of the simulation box was set to half the simulation width, i.e. L_{z} = 0.5L. The size of grid cells was set to 2 Debye lengths, i.e. Δx = Δy = Δz = 2λ_{D}, making up a grid of 200 × 200 × 100 cells using 200 particles per species per cell. While this is less computationally reliable than the 2.5D simulation runs, it is shown in this study that a strong correspondence exists between the two setups.
In order to avoid energy losses or gains due to the finite grid instability (Okuda 1972; Langdon 1970), the 2D and 3D simulation runs both adhere to the established condition that the size of grid cells, Δx, is of the order of the Debye length, λ_{D}. Also using EPOCH, 2.5D simulation runs with a similar configuration and closed boundary conditions, with grid cells of Δx = λ_{D} and Δx = 2λ_{D}, were conducted and it was found that the total energy was conserved within an error of about 1 per cent for the same simulation period (Graf von der Pahlen & Tsiklauri 2015). As with all PIC codes, particles are not subject to discretisation and momentum is conserved to machine precision. In all simulation runs the simulation time step is predetermined by the simulation code as , where c is the speed of light in vacuum. This is sufficient to resolve the propagation of both light and Langmuir waves, i.e. cΔt< Δx and ω_{pe}Δt< 2.
2.3. Boundary conditions
The choice of boundary conditions in a pure Xpoint collapse configuration is not trivial. Unlike in tearingmodetype reconnection, it is not possible to apply periodic boundary conditions, since fieldlines at opposite boundaries are not equidirectional. In Graf von der Pahlen & Tsiklauri (2014a) two types of boundary conditions, open and closed, were used and compared. In the closed case, particles are reflected and fieldlines are kept fixed at the boundary. The latter is ensured by imposing zerogradient boundary conditions on both the electric and magnetic fields, forcing the tangential component of electric field to zero and keeping the normal component of the magnetic field constant. This ensures that no particle energy is lost and that no flux can escape through the boundary. In open boundary conditions, particles reaching the boundary are removed from the system. Fields at the boundary are allowed to evolve freely, as set out in Ruhl (2006), Chap. 2.4 (see radiating boundary conditions), allowing electromagnetic waves to escape. The magnetic field perpendicular to the boundary is kept fixed here. This corresponds to the initial simulation domain being embedded in a larger Xpoint collapse setup. In the open case it was demonstrated that the system allowed for greater reconnection rates and for a smoother system evolution when guidefields were applied. Furthermore, in the open case, for guidefield strengths close to the inplane field, the reconnection dynamics significantly changed. Magnetic islands and electron flow vortices start to emerge. For these reasons the open boundary case was chosen for this study. Also, while closed boundaries are more relevant to laboratory reconnection experiments, open boundaries are more relevant to reconnection events in nature, e.g. reconnection in the geomagnetic tail. When extending the 2D simulation into 3D, the previously ignorable direction (z) now requires welldefined boundaries. These were set as periodic.
3. Reconnection mechanism and dynamics for varying strengths of guidefield in 2.5D
Fig. 2
Reconnection electric field at the Xpoint for 2.5D simulation runs, with guidefield strengths as indicated. 

Open with DEXTER 
Reconnection setups for guidefield strengths of up 0.6B_{P} were run until 500ω_{pe}. In Graf von der Pahlen & Tsiklauri (2014a) it was shown that peak reconnection was reached for all guidefield cases within this time. Figure 2 shows the outofplane electric field (E_{z}) at the Xpoint for 2.5D runs with different values of guidefield, representing the reconnection rate according to Eq. (2). As discussed in the previous work, a greater guidefield leads to increasingly delayed onsets of reconnection. Graf von der Pahlen & Tsiklauri (2014a) also addresses the initial periods of intense highfrequency oscillations, linked to oscillatory reconnection. In the 0.6B_{P} guidefield case, a magnetic island and thus a secondary Xpoint emerge. Rather than plotting both reconnection rates, E_{z} from the Xpoint with the greater reconnection rate is used. As shown, the reconnection rate in the 0.6B_{P} briefly exceeds the reconnection rate in all other cases. This occurred shortly after the emergence of the second Xpoint. It should be noted that the locations of the Xpoint here were tracked and E_{z} was sampled at those locations, whereas in Fig. 4 of Graf von der Pahlen & Tsiklauri (2014a)E_{z} was simply sampled at the centre of the simulation domain. Since the Xpoint starts to move for B_{z0} = 0.6B_{P}, tracking of the Xpoint location becomes necessary to accurately measure the reconnection rate.
Fig. 3
Contributions of different terms in the generalised Ohm’s law to the outofplane electric field along cuts through the current sheet along the xaxis through the Xpoint. The vertical line in each plot marks the horizontal position of the Xpoint, as determined by tracking the magnetic null. The solid black lines show the outofplane electric field (not including the advective electric field component), dashed lines the contribution of the divergence of the pressure tensor, dashdotted lines the contribution of the convective inertia, and dotted lines the contribution from the rate of change of bulk inertia. The solid grey line represents the sum of the contributing terms. 

Open with DEXTER 
Fig. 4
From top to bottom: time progressions of the current density; the dominant component of the convective inertia, ; and the Qvalue. Superimposed on all plots is the inplane magnetic field, and panels g), h), and i) also show electron velocities as well as coloured areas where the Qvalue is greater than zero, indicating the existence of a vortex. The lengths of the arrows next to the plots represent the greatest speeds reached, in each panel approximately 0.22c, i.e. 2.2v_{ae0}. The simulation times, indicated at the top of each panel, were chosen around the occurrence of island and vortex formation. 

Open with DEXTER 
Changes in the shape of the current sheet and reconnection region with increasing guidefield in Xpoint collapse are discussed by Graf von der Pahlen & Tsiklauri (2014a). However, it was not investigated how these changes affected the reconnection mechanism, which is one of the goals of the present study. Similarly to Swisdak et al. (2005), cuts were made through the widths of the current sheet at the Xpoint to show E_{z} at locations along the reconnection region. The terms in the generalised Ohm’s law contributing to E_{z} were calculated. Since for greater guide field cases, electron speeds in the simulation reached increasingly relativistic speeds (see Sect. 5), the modified Ohm’s law (see Eq. (4)) was used. For each guidefield case, cuts were taken when peak reconnection was reached, as shown in Fig. 3. This progression of plots shows that, for greater guidefields, the convective inertial contributions become increasingly asymmetrically distributed relative to the Xpoint. This can be interpreted as shearflows tilting the flow across the diffusion region but not quite forming a vortex. The pressure tensor terms remain the dominant contribution at the Xpoint up to guidefield strengths of 0.4B_{P}, but their area of influence gets increasingly narrow. This can be understood in terms of the increased shear flow: as the electron flow speed along the current sheet increases, electrons that normally would have been undergoing meandering motion in the diffusion region (i.e. adding to the pressure tensor contribution) are now accelerated outwards and thus contribute to the convective inertial term instead. This is in line with the theoretical prediction stated in Eq. (4) in Tanaka (2001).
In the case of a guidefield of 0.6B_{P}, a shift in the previously observed dynamics occurs. After the formation of a magnetic island, and thus a secondary Xpoint, a significant contribution to the reconnection electric field is made up by the convective inertia term. As shown in the final panel of Fig. 3, both the pressure tensor and convective inertia contributions to the electric field are now highly asymmetric across the current sheet along the lower Xpoint (interestingly, this asymmetry is reversed at the upper Xpoint). This change in dynamics is coupled with the emergence of an electron flow vortex in the proximity of the Xpoint. Figure 4 shows the time evolution of the relevant quantities during this shift of dynamics. Closer analysis shows that the convective inertial contribution to the reconnection electric field at the Xpoint is mainly provided by the component . Panels d), e), and f) of Fig. 4 show the evolution of this term superimposed on the inplane magnetic field. As shown in panel d), which corresponds to a time shortly before the reconnection peak, this contribution initially plays a role only adjacent to the current sheet, similar to the contributions of the convective inertia in panels a), b), and c) in Fig. 3. However, as shown in panel e) of Fig. 4, this contribution shifts to the location of the Xpoint and also to the location where the secondary Xpoint is formed, thus playing a role at both reconnection sites.
Panels a), b), and c) in Fig. 4 show the evolution of the outofplane electron current density (j_{z}) and the inplane magnetic field. After the formation of the magnetic island, a strong current starts to develop at its centre. This can be attributed to the compression of the magnetic field, and thus increased curl of the magnetic field, owing to the continued reconnection at the two Xpoints, as explained by Huang et al. (2013). Panels g), h), and i) of Fig. 4 show the electron motion at the same time steps, as contour plot of the Qvalue. This value represents a Galileantransformation invariant measure of vortical flow (Haller 2005; Chakraborty et al. 2005; Hunt et al. 1988), defined as the second invariant of the velocity gradient tensor, ∇v, given by Eq. (5). When greater than zero, the Qvalue indicates the existence of a vortex. in panel g) there is only a shear flow, as predicted by the nature of guidefield reconnection, while in panel e) it can be clearly seen that a vortical flow emerges in the vicinity of the Xpoint. The vortex visible in the velocity field and the Qvalue show good correspondence. Thus, panels b), e), and h) demonstrate that in this reconnection simulation, an increased reconnection rate and the emergence of a secondary Xpoint are brought on by a convective inertial contribution to the reconnection electric field, which is coupled to the emergence of an electron flow vortex. This strongly suggests that the vortex reconnection mode (described in Nakamura et al. 2008; and Chacón et al. 2003) rather than only Xpoint collapse facilitates the reconnection process. As shown by the arrows on the panels, electron flow speeds exceed the electron Alfvén speed (0.1c), as is required for an increase in reconnection rate due to vortex interaction to occur (Mitchell & Kan 1978; Cassak & Otto 2011). Panels f) and i) show the state of the convective inertia contribution and electron motion shortly after peak reconnection. As shown, multiple vortices have formed spreading in a somewhat chaotic fashion along the current sheet. The 3D equivalent of this outcome is investigated in Sect. 4.
4. Reconnection dynamics in 3D
As described in the previous section, the reconnection dynamics of a standard Xpoint collapse simulation are significantly altered by the inclusion of an outofplane guidefield of a strength close to the inplane field, i.e. B_{z} = 0.6B_{P}. The formation of a vortex and magnetic island occurred in a straightforward fashion when the system reached peak reconnection rate, but shortly after the system developed into a chaotic state. To see what these dynamics may correspond to in a real reconnection event, the ignorable direction z was extended outofplane to make a 3D reconnection setup. While an analysis of the term breaking the frozenin condition was not possible here, because no universally agreed definition of 3D reconnection rate exists, finding similar dynamics to the 2.5D case would represent strong evidence that a similar shift from Xpoint collapse to vortex induced reconnection occurred.
Fig. 5
In the vertical direction, panels a), c), and e) show the time evolution of the 3D outofplane current density (j_{ez}) at t = 337ω_{pe},t = 362ω_{pe}, and t = 387ω_{pe}. Current density is represented as an isosurface of 2/3 the maximum current density at the respective time. Again, in the vertical direction, panels b), d), and f) show isosurfaces of the Qvalue (see Eq. (5)) at the same times in the simulation. For clarity, isosurfaces left of the current sheet are shown in blue and isosurfaces right of the current sheet are in yellow. In all panels, inplane magnetic fieldlines are superimposed on fluxtubes, showing the magnetic field at several horizontal slices through the simulation box. The simulation times shown were chosen around the time of island/vortex formation, which occurred slightly earlier in the 3D case. 

Open with DEXTER 
Figure 5 shows the time progression of electron current density and the Qvalue representing electron flow vortices. The snapshots for the panels in Fig. 5 were taken to show the progression of island and vortex formation, which occurred slightly earlier than in the 2.5D case. The electron current density is represented by an isosurface (i.e. a surface where the current density has a constant value); the chosen isosurface values are approximately twothirds of the maximum current density at each snapshot. As shown in panel a), at first the current density has the shape of a standard current sheet, as would be the case for zero guidefield. However, in panel c) the current sheet starts to fragment, eventually leading to a tubular structure as shown in panel e). Similarly to the 2D case, the locations of elevated current density correspond to the centre of the fluxtubes, which have been shown to be the 3D equivalent of magnetic islands (Karimabadi et al. 1999) where magnetic fields are compressed and currents are increased (Huang et al. 2013). Interestingly, while the initial fragmentation in panel c) appears to be random, the final isosurface shows a distinct tubular structure. Furthermore, rather than connecting back on itself, as would be the 3D equivalent of panel c) in Fig. 4, the current density is tilted along the yaxis, similarly to studies of 3D reconnection with a guidefield in a tearingmode setup (Liu et al. 2013).
Another difference in the 3D results can be seen in the evolution of the Qvalue, as shown in Fig. 5, panels b), d), and f). Each snapshot shows two isosurfaces where the Qvalue exceeds zero. Yellow shows the isosurfaces for positive Qvalue on the righthand side of the diffusion region (i.e. right of the x = 0 line), while the blue isosurfaces correspond to positive Qvalues on the lefthand side of the diffusion region. This distinct colour scheme was chosen to represent the arrangement of vortical flows clearly throughout the simulation domain. Panel b) represents the initial instance of vorticity and, as for the current density, there initially appears to be no distinct structure. However, as seen in panel f), eventually two sets of distinct vortex tubes emerge on each side of the diffusion region, tilted in opposite directions. Again, this represents a structure that could not be adequately represented in a 2.5D simulation, and only the 3D simulation reveals the orderly, realistic dynamics. This gives new insight into panel i) of Fig. 4, as this apparently disordered arrangement of vortical flows actually corresponds to a welldefined structure in 3D. The motion of the vortex tubes in the plane appears to be in the opposite direction of the inplane shear flow along the current sheet, which seems to contradict basic theoretical considerations of the motions of vortex tubes in shear flows (Kuo 1969). However, this is in fact a misconception as the vortex tubes move downwards, along with the bulk electron current flow. Owing to their inclination relative to the zaxis, the illusion of motion in the xyplane is created.
Fig. 6
As in Fig. 5 panel f), showing vortical flows in 3D simulation runs with a guidefield of 0.6B_{P} according to the Qvalue at t = 387ω_{pe}. Panels use perspectives as indicated by axes. In panel a), the dashed lines signify the distance of vortex tubes from the centre of the domain and thus from the centre of the diffusion region. Dashed lines in panel b) are inclined at the calculated value of θ, based on Eq. (6), and show a strong correspondence with the inclination of vortex tubes. 

Open with DEXTER 
Following the analysis of tilted (oblique) current sheets in Baalrud et al. (2012), panel a) of Fig. 6 shows vortex tubes as they appear in the xzplane. Dashed lines on plots signify the locations of vortex tubes, which are shown to be left and right of the centre of the domain. By taking the mean value of the strengths of the sheared magnetic field, B_{y}, and the magnetic guidefield, B_{z}, at these locations, a prediction for the angle of the oblique modes, according to Eq. (6), is found to be θ = ± 16°. Panel b) of Fig. 6 shows vortex tubes as they appear in the yzplane. Here, dashed lines on the plot are inclined at the calculated value of θ, effectively representing the inclination of the outofplane magnetic field. As shown, there is a clear correspondence between the tilt of the vortex tubes and θ. Unlike in Baalrud et al. (2012), Liu et al. (2013) and Akçay et al. (2016) where a tearingmode setup is used, it is not possible in Xpoint collapse to relate the locations of oblique structures to initial simulation parameters since Xpoint collpase is inherently time variant and the width and shape of the diffusion region is not fixed by the setup. Furthermore, as there is no asymptotic magnetic field, there is no limit on the angle of obliqueness. However, by taking the B_{y} profile across the diffusion region during vortex formation to be of the form , where λ is the half width of the diffusion region, and noting that B_{z} across the diffusion region is approximately constant, while B_{y} ≈ B_{z} at the diffusion region edge, we arrive at a distance relation similar to Baalrud et al. (2012) given by x_{s} = λtan(θ). This gives the distance of the oblique vortex tubes from the centre of the domain as x_{s} = ± 0.3λ. In physical distance, this equates to approximately x_{s} = ± 0.03m, based on the measured width of the diffusion region, and is a good match as is shown in panel a) of Fig. 6.
5. Particle distribution function dynamics
Fig. 7
Electron and ion particle velocity distribution functions for guidefield cases of 0B_{P} and 0.6B_{P} in the 2.5D simulation runs. Black lines show the distribution functions at the beginning of the respective simulation. The dark grey lines show the distribution functions at peak reconnection, i.e. t = 250 /ω_{pe} for zero guidefield and 425 /ω_{pe} for the 0.6B_{P} guidefield case. The light grey lines show distribution functions at the end of the simulation, i.e. t = 500 /ω_{pe}. The particles included in the plots were chosen from an area around the diffusion region, i.e. (− 2c/ω_{pe}) <x< (2c/ω_{pe}) and (− 8c/ω_{pe}) <y< (8c/ω_{pe}). 

Open with DEXTER 
Figure 7 shows the distribution functions for electron and ion particle velocities in the 2.5D simulation runs for different guidefield cases. Electrons and ions initially have opposite velocities in the z direction and are oppositely accelerated by the reconnection electric field. In each case, the three lines on the plot show velocity distributions at the start of the simulation, at peak reconnection, and at the end of the simulation.
It can be seen that for greater guidefield cases, increased outofplane electron acceleration is observed. In the 0.6B_{P} guidefield case a bumpontail distribution in v_{ez} emerged at peak reconnection, stretching into the relativistic regime, and subsequently flattened out again. A similar effect was observed in simulations in Tsiklauri & Haruki (2008), Fig. 6, when an increased stress parameter of α = 2.24 was used in a 2.5D simulation of closed boundary Xpoint collapse. This indicates that there is an equivalence to using greater initial guidefields and greater initial stress in the inplane magnetic field.
While electrons in the zero guidefield case experience less acceleration in the zdirection, the acceleration of ions is in fact greater, leading to a slight bump in v_{iz}. However, the acceleration of ions in the ydirection in the 0.6B_{P} guidefield case greatly exceeds that in the zero guidefield case. This implies that ions are moved out of the diffusion region faster in the 0.6B_{P} guidefield case and thus experience less outofplane acceleration by the reconnection electric field, which explains the reduced acceleration in the zdirection.
Equivalent results for 3D simulation runs are shown in Fig. 8. While it was not possible to determine the time of peak reconnection in this case, intermediate time steps for the distribution function were chosen to be the points when the reconnection current reached a peak value, which was shown to approximately correspond to peak reconnection rates in 2.5D simulations of Xpoint collapse (Graf von der Pahlen & Tsiklauri 2014a). For high guidefields, peak current was reached sooner in the 3D case than in the 2.5D case, which is consistent with the earlier onset of vortex formation. Although timescales of processes were affected, the resulting distribution functions are notably similar, including features such as the bumpontail distribution in the outofplane electron velocity in the high guidefield case. While different dynamical features can emerge in the 3D case, this result shows that a high degree of correspondence exists in the bulk particle acceleration.
There are some differences in the plots, although they appear to be the result of a mismatch in the simulation times of snapshots considered. For example, v_{iz} in the low guidefield case and v_{ez} in the high guidefield case seem to vary only in the intermediate distribution function, while the initial and final distributions mostly take on the same shape. They thus show the same progression shifted in time. For v_{ex} in the high guidefield case two distinct bumps appear in the final snapshot of the 2.5D simulation, while in the 3D case they have already thermalised by the time of the final snapshot, again showing that reconnection proceeds slightly faster in 3D.
Fig. 8
Electron and ion particle velocity distribution functions for guidefield cases of 0B_{P} and 0.6B_{P} in the 3D simulation runs. The black lines show the distribution functions at the beginning of the respective simulation. The dark grey lines show the distribution functions when peak reconnection currents were reached, which occurred in both guidefield cases at around t = 375 /ω_{pe}. The light grey lines show distribution functions at the end of the simulation, i.e. t = 500 /ω_{pe}. The particles included in the plots were chosen from an area around the diffusion region, i.e. (− 2c/ω_{pe}) <x< (2c/ω_{pe}) and (− 8c/ω_{pe}) <y< (8c/ω_{pe}). 

Open with DEXTER 
6. Conclusions
By studying Xpoint collapse with open boundary conditions and an outofplane guidefield close to the strength of the inplane field, new insights have been gained into the specifics of reconnection dynamics. Using 2.5D simulations it was shown that reconnection dynamics were significantly altered by the increased induced shear flow. It was shown that, while the increased guidefield initially suppressed the reconnection rate, later in the simulation a brief period of peak reconnection was attained where the reconnection electric field exceeded that of lower guidefield cases. The reconnection electric field at this point was substantially supported by the convective inertia term in the generalised Ohm’s law rather than the divergence of the pressure tensor (see Fig. 3). This stands in stark contrast to previous studies of tearingmode reconnection with a guidefield, where no change in the reconnection mechanism was observed. The shift in reconnection mechanism during peak reconnection coincided with the formation of a secondary Xpoint and with an electron flow vortex (see Fig. 4). We conclude that owing to the induced shearflow along the current sheet, vortexinduced reconnection takes effect, which allows for the change in reconnection dynamics.
While particle velocity distribution functions show that bulk particle acceleration proceeds in a similar fashion in 2.5D and 3D simulations (see Sect. 5), in the high guidefield case 3D structures emerged that are not present in the 2.5 simulation. At later simulation times in the 2.5D simulation the vortical flows took on an apparently chaotic shape. However, when the simulation setup was extended into 3D geometry, vortical flows were shown to selfassemble into oblique 3D tubes and to take on a distinct structure that cannot be represented in a 2.5D simulation (see panel (f) of Fig. 5). Similarly, magnetic fluxtubes (i.e. magnetic islands in 2D ) and tubular regions of elevated current density appeared to be sheared along the zdirection. It was shown that the tilt angles of the vortex tubes correspond well with predictions for tilts due to oblique modes, as discussed in Liu et al. (2013, see Fig. 6. As oblique modes are suppressed in 2.5D simulations, this further shows that the emergent structure observed is unique to the 3D case.
Since purely 2D reconnection setups are an unlikely occurrence in nature, guidefield reconnection setups and their induced shear flows are important aspects of the study of magnetic reconnection and are likely to be needed to accurately model reconnection scenarios in the solar corona and the Earth’s magnetosphere. As discussed in Otto & Fairfield (2000), vortex formation due to shear flow have already been observed in the Earth’s magnetosheath. We hope the results of this study may further the progress in this field and other studies of reconnection where guidefields could lead to large shear flows. We hope to inspire the further investigation of vortical flows in in situ observations to see if there may be correspondence to the 3D structures found in this study.
Acknowledgments
The authors acknowledge the use of the particleincell code EPOCH and the support by the development team (http://ccpforge.cse.rl.ac.uk/gf/project/epoch/). Computational facilities used are those of the Astronomy Unit, Queen Mary University of London, and the STFCfunded UKMHD consortium at St. Andrews and Warwick Universities. J.G.V.D.P. acknowledges support from the STFC Ph.D. studentship. D.T. is financially supported by STFC consolidated Grant ST/J001546/1 and The Leverhulme Trust Research Project Grant RPG311.
References
 Akçay, C., Daughton, W., Lukin, V. S., & Liu, Y.H. 2016, Physics of Plasmas, 23, 012112 [NASA ADS] [CrossRef] [Google Scholar]
 Arber, T. D., Bennett, K., Brady, C. S., et al. 2015, Plasma Physics and Controlled Fusion, 57, 1 [NASA ADS] [CrossRef] [Google Scholar]
 Aschwanden, M. J. 2005, Physics of the Solar Corona, An Introduction with Problems and Solutions 2nd edn. (Springer), 383 [Google Scholar]
 Baalrud, S. D., Bhattacharjee, A., & Huang, Y.M. 2012, Physics of Plasmas, 19, 022101 [NASA ADS] [CrossRef] [Google Scholar]
 Birn, J., & Priest, E. R. 2007, Reconnection of magnetic fields: magnetohydrodynamics and collisionless theory and observations (Cambridge University Press) [Google Scholar]
 Cai, D., Lembège, B., Hasegawa, H., & Nishikawa, K.I. 2015, in EGU General Assembly, Vienna, Austria, EGU2015–10087 [Google Scholar]
 Cassak, P. A., & Otto, A. 2011, Physics of Plasmas, 18, 074501 [NASA ADS] [CrossRef] [Google Scholar]
 Chacón, L., Knoll, D. A., & Finn, J. M. 2003, Phys. Lett. A, 308, 187 [NASA ADS] [CrossRef] [Google Scholar]
 Chakraborty, P., Balachandar, S., & Adrian, R. J. 2005, J. Fluid Mech., 535, 189 [NASA ADS] [CrossRef] [Google Scholar]
 Chen, P. F. 2011, Liv. Rev. Sol. Phys., 8 [Google Scholar]
 Cranmer, S. R. 2009, Liv. Rev. Sol. Phys., 6 [Google Scholar]
 De Moortel, I., & Browning, P. 2015, Phil. Trans. Roy. Soc. Lond. Ser. A, 373, 20140269 [NASA ADS] [CrossRef] [Google Scholar]
 Dungey, J. W. 1953, Phil. Mag., 44, 725 [Google Scholar]
 Eastwood, J. P., Phan, T.D., Mozer, F. S., et al. 2007, J. Geophys. Res. (Space Physics), 112, 6235 [NASA ADS] [CrossRef] [Google Scholar]
 Fermo, R. L., Drake, J. F., & Swisdak, M. 2012, Phys. Rev. Lett., 108, 255005 [NASA ADS] [CrossRef] [Google Scholar]
 Gosling, J. T., Eriksson, S., Phan, T. D., et al. 2007, AGU Spring Meeting Abstracts, SH32A01 [Google Scholar]
 Graf von der Pahlen, J., & Tsiklauri, D. 2014a, Physics of Plasmas, 21, 012901 [NASA ADS] [CrossRef] [Google Scholar]
 Graf von der Pahlen, J. G., & Tsiklauri, D. 2014b, Physics of Plasmas, 21, 060705 [CrossRef] [Google Scholar]
 Graf von der Pahlen, J., & Tsiklauri, D. 2015, Physics of Plasmas, 22, 032905 [NASA ADS] [CrossRef] [Google Scholar]
 Haller, G. 2005, J. Fluid Mech., 525, 1 [NASA ADS] [CrossRef] [Google Scholar]
 Hesse, M., & Zenitani, S. 2007, Physics of Plasmas, 14, 112102 [NASA ADS] [CrossRef] [Google Scholar]
 Hesse, M., Schindler, K., Birn, J., & Kuznetsova, M. 1999, Physics of Plasmas, 6, 1781 [NASA ADS] [CrossRef] [Google Scholar]
 Hesse, M., Kuznetsova, M., & Hoshino, M. 2002, Geophys. Res. Lett., 29, 1563 [NASA ADS] [CrossRef] [Google Scholar]
 Hesse, M., Kuznetsova, M., & Birn, J. 2004, Physics of Plasmas, 11, 5387 [NASA ADS] [CrossRef] [Google Scholar]
 Horiuchi, R., & Sato, T. 1997, J. Plasma Phys., 4, 277 [CrossRef] [Google Scholar]
 Huang, C., Lu, Q., Wu, M., Lu, S., & Wang, S. 2013, J. Geophys. Res. (Space Physics), 118, 991 [NASA ADS] [CrossRef] [Google Scholar]
 Hunt, J. C. R., Wray, A. A., & Moin, P. 1988, in Studying Turbulence Using Numerical Simulation Databases, 2, 193 [Google Scholar]
 Jain, N., & Büchner, J. 2015, J. Plasma Phys., 81, 905810606 [CrossRef] [Google Scholar]
 Jiang, C., Wu, S. T., Feng, X., & Hu, Q. 2016, Nature Commun., 7, 11522 [NASA ADS] [CrossRef] [Google Scholar]
 Karimabadi, H., KraussVarban, D., Omidi, N., & Vu, H. X. 1999, J. Geophys. Res., 104, 12313 [NASA ADS] [CrossRef] [Google Scholar]
 Kleva, R. G., Drake, J. F., & Waelbroeck, F. L. 1995, Physics of Plasmas, 2, 23 [NASA ADS] [CrossRef] [Google Scholar]
 Kuo, H. L. 1969, J. Atmos. Sci., 26, 390 [NASA ADS] [CrossRef] [Google Scholar]
 Langdon, A. B. 1970, J. Comp. Phys., 6, 247 [NASA ADS] [CrossRef] [Google Scholar]
 Liu, Y.H., Daughton, W., Karimabadi, H., Li, H., & Roytershteyn, V. 2013, Phys. Rev. Lett., 110, 265004 [NASA ADS] [CrossRef] [Google Scholar]
 Melzani, M., Walder, R., Folini, D., Winisdoerffer, C., & Favre, J. M. 2014, A&A, 570, A111 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Mitchell, Jr., H. G., & Kan, J. R. 1978, J. Plasma Phys., 20, 31 [NASA ADS] [CrossRef] [Google Scholar]
 Nakamura, T. K. M., Fujimoto, M., & Otto, A. 2008, J. Geophys. Res. (Space Physics), 113, 9204 [NASA ADS] [Google Scholar]
 Okuda, H. 1972, J. Comp. Phys., 10 [Google Scholar]
 Otto, A., & Fairfield, D. H. 2000, J. Geophys. Res., 105, 21175 [NASA ADS] [CrossRef] [Google Scholar]
 Parker, E. N. 1957, J. Geophys. Res., 62, 509 [NASA ADS] [CrossRef] [Google Scholar]
 Petschek, H. E. 1964, NASA SP, 50, 425 [NASA ADS] [Google Scholar]
 Priest, E. R., Hornig, G., & Pontin, D. I. 2003, J. Geophys. Res. (Space Physics), 108, 1285 [NASA ADS] [CrossRef] [Google Scholar]
 Pritchett, P. L. 2001, J. Geophys. Res., 106, 3783 [NASA ADS] [CrossRef] [Google Scholar]
 Ruhl, H. 2006, Introduction to Computational Methods in Many Body Physics (Rinton Press) [Google Scholar]
 Schindler, K., Hesse, M., & Birn, J. 1988, J. Geophys. Res., 93, 5547 [NASA ADS] [CrossRef] [Google Scholar]
 Shibata, K., & Magara, T. 2011, Liv. Rev. Sol. Phys., 8, 6 [Google Scholar]
 Shibata, K., & Yokoyama, T. 1999, ApJ, 526, L49 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Swisdak, M., Drake, J. F., Shay, M. A., & McIlhargey, J. G. 2005, J. Geophys. Res. (Space Physics), 110, 5210 [NASA ADS] [CrossRef] [Google Scholar]
 Tanaka, M. 2001, Earth Planets Space, 53, 463 [NASA ADS] [CrossRef] [Google Scholar]
 Tsiklauri, D., & Haruki, T. 2007, Physics of Plasmas, 14, 112905 [NASA ADS] [CrossRef] [Google Scholar]
 Tsiklauri, D., & Haruki, T. 2008, Physics of Plasmas, 15, 102902 [NASA ADS] [CrossRef] [Google Scholar]
 Villasenor, J., & Buneman, O. 1992, Comp. Phys. Commun., 69, 306 [NASA ADS] [CrossRef] [Google Scholar]
All Figures
Fig. 1
Reconnection at an Xpoint in a 3D domain, showing the motion of two sets of magnetic fieldlines. The perspectives on the simulation domain are indicated for each row. in panel b) the direction of the reconnection electric field induced as the fieldlines pass the Xpoint is indicated. Thin arrows on panels c) and f) indicate the direction of the electron current generated by the reconnection electric field. As shown, the shape of the fieldlines guides the accelerated particles such that there is a shearflow in the xyplane. 

Open with DEXTER  
In the text 
Fig. 2
Reconnection electric field at the Xpoint for 2.5D simulation runs, with guidefield strengths as indicated. 

Open with DEXTER  
In the text 
Fig. 3
Contributions of different terms in the generalised Ohm’s law to the outofplane electric field along cuts through the current sheet along the xaxis through the Xpoint. The vertical line in each plot marks the horizontal position of the Xpoint, as determined by tracking the magnetic null. The solid black lines show the outofplane electric field (not including the advective electric field component), dashed lines the contribution of the divergence of the pressure tensor, dashdotted lines the contribution of the convective inertia, and dotted lines the contribution from the rate of change of bulk inertia. The solid grey line represents the sum of the contributing terms. 

Open with DEXTER  
In the text 
Fig. 4
From top to bottom: time progressions of the current density; the dominant component of the convective inertia, ; and the Qvalue. Superimposed on all plots is the inplane magnetic field, and panels g), h), and i) also show electron velocities as well as coloured areas where the Qvalue is greater than zero, indicating the existence of a vortex. The lengths of the arrows next to the plots represent the greatest speeds reached, in each panel approximately 0.22c, i.e. 2.2v_{ae0}. The simulation times, indicated at the top of each panel, were chosen around the occurrence of island and vortex formation. 

Open with DEXTER  
In the text 
Fig. 5
In the vertical direction, panels a), c), and e) show the time evolution of the 3D outofplane current density (j_{ez}) at t = 337ω_{pe},t = 362ω_{pe}, and t = 387ω_{pe}. Current density is represented as an isosurface of 2/3 the maximum current density at the respective time. Again, in the vertical direction, panels b), d), and f) show isosurfaces of the Qvalue (see Eq. (5)) at the same times in the simulation. For clarity, isosurfaces left of the current sheet are shown in blue and isosurfaces right of the current sheet are in yellow. In all panels, inplane magnetic fieldlines are superimposed on fluxtubes, showing the magnetic field at several horizontal slices through the simulation box. The simulation times shown were chosen around the time of island/vortex formation, which occurred slightly earlier in the 3D case. 

Open with DEXTER  
In the text 
Fig. 6
As in Fig. 5 panel f), showing vortical flows in 3D simulation runs with a guidefield of 0.6B_{P} according to the Qvalue at t = 387ω_{pe}. Panels use perspectives as indicated by axes. In panel a), the dashed lines signify the distance of vortex tubes from the centre of the domain and thus from the centre of the diffusion region. Dashed lines in panel b) are inclined at the calculated value of θ, based on Eq. (6), and show a strong correspondence with the inclination of vortex tubes. 

Open with DEXTER  
In the text 
Fig. 7
Electron and ion particle velocity distribution functions for guidefield cases of 0B_{P} and 0.6B_{P} in the 2.5D simulation runs. Black lines show the distribution functions at the beginning of the respective simulation. The dark grey lines show the distribution functions at peak reconnection, i.e. t = 250 /ω_{pe} for zero guidefield and 425 /ω_{pe} for the 0.6B_{P} guidefield case. The light grey lines show distribution functions at the end of the simulation, i.e. t = 500 /ω_{pe}. The particles included in the plots were chosen from an area around the diffusion region, i.e. (− 2c/ω_{pe}) <x< (2c/ω_{pe}) and (− 8c/ω_{pe}) <y< (8c/ω_{pe}). 

Open with DEXTER  
In the text 
Fig. 8
Electron and ion particle velocity distribution functions for guidefield cases of 0B_{P} and 0.6B_{P} in the 3D simulation runs. The black lines show the distribution functions at the beginning of the respective simulation. The dark grey lines show the distribution functions when peak reconnection currents were reached, which occurred in both guidefield cases at around t = 375 /ω_{pe}. The light grey lines show distribution functions at the end of the simulation, i.e. t = 500 /ω_{pe}. The particles included in the plots were chosen from an area around the diffusion region, i.e. (− 2c/ω_{pe}) <x< (2c/ω_{pe}) and (− 8c/ω_{pe}) <y< (8c/ω_{pe}). 

Open with DEXTER  
In the text 
Current usage metrics show cumulative count of Article Views (fulltext article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 4896 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.