
2
tron inertial scale, computationally they are a factor
∼(dH/de)=(mH/me)1/2more cost efficient in each
dimension per timestep than fully kinetic simulations.
Here dH,de,mHand meare the inertial lengths and
masses of protons and electrons, respectively. Therefore
hybrid simulations enable much larger domains to cap-
ture the essential physics of heavy ion acceleration. De-
spite the fluid approximation for electrons, hybrid sim-
ulations have demonstrated good agreement for the re-
connection rate and dynamics compared to fully kinetic
simulations [35–38]. In the Appendix, we show that hy-
brid and fully kinetic [20] simulations produce very sim-
ilar proton acceleration and flux rope dynamics, demon-
strating that the hybrid model is viable for studying ion
acceleration.
Our hybrid simulations, for the first time, achieved ef-
ficient acceleration of multiple ion species (with a wide
range of charge and mass up to 56Fe14+) into nonthermal
power-law energy spectra. We find that the 3D recon-
nection layers consist of fragmented kinking flux ropes
across different scales (mainly from the m= 1 flux-
rope kink instability), which are growing in both width
and length over time, as a distinct component of the
reconnection-driven turbulence. The origin and prop-
erties of reconnection-driven turbulence are frontiers of
research [39–42]. Similar strong and turbulent magnetic
fluctuations have also been observed in magnetotail re-
connection [10, 11]. This 3D dynamics plays a critical
role in the particle acceleration for all species, by facilitat-
ing transport to acceleration regions. Different ions are
pre-accelerated/injected into nonthermal energies when
first bouncing off an Alfv´enic outflow at a reconnection
exhaust (a single Fermi reflection). The injection pro-
cess leads to “shoulders” in the energy spectra, becom-
ing the low-energy bounds that control the nonthermal
energy content. At higher energy, all species undergo
a universal Fermi acceleration process between outflows
and form power-law energy spectra with similar indices
(p∼4.5). However, the onset times of Fermi acceler-
ation are delayed for lower charge-mass-ratio ions, un-
til the flux ropes and neighboring exhausts grow large
enough to magnetize them. Consequently, the maximum
energy per nucleon εmax ∝(Q/M)αwhere α∼0.6 for
low upstream plasma β, and both pand αincrease as β
approaches unity. These results are consistent with the
HCS and magnetotail observations [8, 12, 13], suggesting
that the observed energetic particles may be a natural
consequence of reconnection.
Numerical Simulations.— We use the Hybrid-
VPIC code [43, 44] that evolves multi-species ions as
nonrelativistic particles and electrons as adiabatic fluid,
which is coupled with Ohm’s law (with small hyper-
resistivity and resistivity to break the electron frozen-in
condition), Ampere’s law and Faraday’s law. The simu-
lations start from two identical current sheets (our anal-
yses focus on one) with periodic boundaries and force-
free profiles: Bx=B0[tanh((z−0.25Lz)/λ)−tanh((z−
0.75Lz)/λ)−1], By=qB2
0+B2
g−B2
x, with uniform
density and temperature. We use the initial electron den-
sity n0for the density normalization. B0is the reconnect-
ing field, Bgis the guide field, Lzis the domain size in
z, and λis the half thickness of the sheet set to be 1 dH.
bg=Bg/B0= 0.1 (corresponding to a magnetic shear an-
gle 169◦), which represents in general the low-guide-field
regime in the HCS and magnetotail [9, 45–48]. The do-
main size Lx×Ly×Lz= 1350×140.4×672d3
H, with grid
size ∆x= ∆y= ∆z= 0.6dHand 800 protons per cell (
4.7×1011 protons in total). Lyis sufficient for capturing
the m= 1 flux-rope kink mode for efficient acceleration
[20]. Small long-wavelength perturbations are included
to initiate reconnection at both current sheets. To limit
the influence of periodic boundaries, the simulations ter-
minate at time ∼1.3Lx/VA, during which less than 1/3
of the upstream magnetic flux is reconnected and the two
current sheets are not yet interacting. We include sev-
eral ion species 1H+,4He2+,3He2+ ,16O7+,56Fe14+ , with
abundance 95%, 5%, 0.1%, 0.1%, 0.1% respectively.
Our simulations are relevant for multi-X-line collision-
less reconnection, as well as plasmoid reconnection in a
thicker current sheet that may develop kinetic-scale cur-
rent sheets to trigger collisionless reconnection [49–53].
We present three runs with different initial temper-
atures Ti=Te= 0.04,0.09,0.25mHV2
A, where VA=
B0/√4πn0mHis the Alfv´en speed, resulting in pro-
ton βH= 0.08,0.18,0.5 respectively. We discuss the
βH= 0.18 run by default and use others for compar-
ison. Unless otherwise stated, the simulations employ
the same initial temperature for all ion species. We have
performed additional simulations to confirm that the con-
clusions are not sensitive to different initial temperatures
for different species.
Reconnection Current Sheet with 3D Frag-
mented Kinking Flux Ropes.— Figure 1(a)-(d)
shows Bzin the x−yplane in the center of one cur-
rent sheet. The unprecedented 3D domain size facilitates
strong m= 1 kink instability of flux ropes that com-
pletely fragmentizes the flux ropes – in contrast to previ-
ous smaller-domain simulations [20] with more coherent
flux ropes (see Appendix). This leads to turbulent mag-
netic fluctuations, as in magnetotail reconnection [10, 11].
As reconnection proceeds, these fragmented kinking flux
ropes keep growing over time both in width and length,
while they advect along the global bidirectional outflows
in x. We visualize these flux ropes in 3D in Figure 1(e-
f) from different perspectives. Flux ropes in (e) can be
directly compared to those in (c), with the same per-
spective and time. Panel (f) emphasizes that flux ropes
exist over a range of scales: one flux rope is newly born
from the reconnection layer (green box), and another has
grown to occupy a sizeable fraction of the domain (orange
box). This flux-rope kink instability produces chaotic