
2
pling of superconducting order parameters. We model
our junctions using a circuit network of coupled resis-
tively and capacitively shunted junctions (RCSJs). We
show that the semi-classical RCSJ model reproduces the
observed resonant features, which are similar in nature to
those predicted for multiplet Cooper pairings [23,25,30].
To elucidate the underlying mechanisms giving rise to
these features, we calculate the contribution of the quasi-
particle current which reveals the pair current contribu-
tion to the total current. We further consider materials-
specific properties such as the Fermi surface geometry of
the normal material and junction transparency by incor-
porating the relevant current-phase relation (CPR) into
the RCSJ model. We show that while the Fermi sur-
face geometry is primarily responsible for the shape of
the CCC [26], the resonant features outside of the CCC
are robust to changes in the Fermi surface and junc-
tion transparency. Finally, our study demonstrates that
circuit-network effects, predicted by the RCSJ model,
lead to macroscopic signatures in differential resistance
maps that are identical to those ascribed to distinctly
quantum processes such as multiplet pairings.
The paper is organized as follows. In Sec. II, we dis-
cuss the details of device fabrication. In Sec. III, we de-
scribe the transport data for the asymmetric device. We
also theoretically analyze our junctions using an RCSJ
model. In Sec. IV, we discuss the transport results in
the symmetric device. We establish that multiple An-
dreev reflections (MARs) are responsible for the discrep-
ancies between the experimental and theoretical data. In
Secs. Vand VI, we consider the effect of different CPRs
ranging from ballistic to diffusive transport in the RCSJ
model and analyze the resultant differential resistance
maps. We conclude in Sec. VII.
II. DEVICE FABRICATION
We assemble hBN/graphene/hBN van der Waals het-
erostructures using a standard dry transfer technique,
followed by annealing in H2/Ar gas at 350 ◦C to re-
move polymer residues from the heterostructures [32,33].
The heterostructures are then patterned with electron
beam lithography, followed by dry etching (O2/CHF3),
to define the junction area. Another e-beam lithography
step is performed to define the contact patterns. Finally,
Ti(10 nm)/Al(100 nm) is evaporated to create supercon-
ducting edge contacts. Atomic force microscope (AFM)
images of our four-terminal JJs are shown in Fig. 1(a).
Devices A and B are four-terminal asymmetric and sym-
metric junctions, respectively. The channel length of de-
vice A is 0.8µm and 3 µm along I13 and I24 directions,
respectively. Device B has a circular geometry with a
diameter of 1.3µm. The Dirac point of device A and
B is at Vg=−4.5 V and Vg=−11.25 V, respectively.
In Fig. 1(a), we show the current directions used for the
transport measurements. We set I1=−I3and I2=−I4
in our experiments. We perform all of the measurements
FIG. 1. (a) The AFM image of device A. Inset shows the
AFM image of device B. The Josephson junctions are made
of hBN-graphene-hBN (outlined by white dashed line) edge
contacted with superconducting Al terminals. Arrows show
the directions of bias currents. (b) Color map of the differ-
ential resistance (dV13 /dI1) versus applied dc current biases
I1=−I3and I2=−I4, as indicated in panel (a) of device
A. The differential resistance is measured using a lock-in am-
plifier. Dashed red lines indicate superconducting branches
corresponding to Vjk = 0, where jand kare pairs of super-
conducting terminals as labeled in (b). Vertical lines in the
map along the I2direction are multiple Andreev reflections
corresponding to eV13 = 2∆/n with ∆ ≈169 µeV. All the
measurements are performed at Vg= 50 V and T= 12 mK.
at T= 12 mK, if not otherwise specified.
III. TRANSPORT IN THE ASYMMETRIC
DEVICE GEOMETRY
We now turn our attention to the transport char-
acteristics of the multi-terminal junctions in device A.
Fig. 1(b) shows a differential resistance dV13/dI1map
versus I1and I2measured at Vg= 50 V. The directions
indicated by the red dashed lines are branches along the
local minima of dV13/dI1and satisfy Vjk = 0 conditions
as labeled. These conditions correspond to supercurrent
flow between terminals jand k. The vertical lines in
the map along the I2direction are signatures of MARs.
Along these lines, eV13 ≈2∆/n, where eis the electron
charge, ∆ is the superconducting gap, and nis an in-
teger. We calculate an induced superconducting gap of
∆≈169 µeV by fitting eV13 to 2/n. The calculated gap is
consistent with the gap we obtain from the temperature-
dependent measurement. The coherence length based on
this gap size is ξ= ¯hvF/π∆≈1.2 µm, where ¯his the
reduced Planck constant and vF≈1×106m/s is the
Fermi velocity of graphene. The MAR signatures are
only observed along terminals 1 and 3, with distance ∼
0.8 µm< ξ, whereas no MAR is observed along terminals
2 and 4, with distance ∼3µm> ξ. This observation is
consistent with phase-coherent Andreev processes where
the superconducting phase keeps (loses) its coherence be-
fore being reflected multiple times between terminals 1
and 3 (2 and 4).
To more closely examine the fine structure around the