
3
IC
+IC
+
a) b) c)
IC
+
IC
-IC
-IC
-
FIG. 2. I−Vcurves of the left junction measured at 3 values of IR: a) −0.3µA, b) 0µA, and c) 0.3µA. In all cases, the control
current IRis well within the dissipationless range of the right junction. In all panels, ILis swept back and forth, showing
negligible hysteresis. In panel (b), the I−Vcurves are reciprocal, as expected. In (a) and (c), either I−
cor I+
care approaching
zero, resulting in a nearly ideal superconducting diode effect.
tions. An example is the left dashed line in Fig. 1b, which
cuts through the hexagonal region away from its center.
As a result, the system stays dissipationless only for pos-
itive ILand becomes dissipative at negative IL(Fig. 2a).
Note that all three junctions play a role in this this pro-
cess: the bottom junction switches at IL=I−
c≈0 while
the left junction is responsible for the upper limit of the
dissipationless range, I+
c; finally, it is the biasing of the
right junction which establishes the required asymme-
try. For zero IR, one recovers a symmetric cross-section
along the ILdirection (central white line in Fig. 1b and
Fig. 2b). Finally, at the opposite value of the control
current IR(right dash line in Fig. 1b and Fig. 2c) the
system stays dissipationless only at negative IL≤0 and
the curves are reversed compared to Fig. 2a. Note that
the right junction is biased below its critical current in
all three cases.
Collectively, the boundaries of the hexagon enable tun-
ing of the transport non-reciprocity between any pair of
contacts by adjusting the current applied to the third
contact. This leads to a diode efficiency,
η=I+
c+I−
c
I+
c−I−
c
,(1)
that in practice can be tuned to exceed 90%, as will be
seen in Fig. 2, where we plot the I−Vcurves corre-
sponding to the three cross-sections in Fig. 1b. Each set
of curves is measured in both directions, showing negligi-
ble hysteresis. When IR= 0 (Fig. 2b) the curves are ex-
pectedly symmetric, so I+
c=−I−
cand η= 0. However,
as IRis tuned away from zero, the non-reciprocity grows
and |η|increases until it reaches ±94% at IR=±0.3µA.
Further increase of ηis possible by applying higher IR.
In fact, formally ηcan exceed unity when I−
cbecomes
positive (same sign as I+
c). However, this regime should
be avoided if we are interested in rectifying small cur-
rents. Hence we stop increasing |IR|at the point where
the high slope “knee” of the I−Vcurve approaches the
point IL= 0. We then define I−
cconservatively as a point
at half the slope of the knee (see arrow in Fig. 2a), re-
sulting in the η≈94%. Finally, either positive (Fig. 2a)
or negative (Fig. 2c) currents can be rectified depending
on the desired operation.
To demonstrate the potential utility of the device in
quantum circuits, we decrease the gate voltage to zero
and apply a square wave of amplitude ±60 nA to the
left contact (Fig. 3a). IRcan then be set to produce
desirable device responses: namely, at IR= 50 and −50
nA, the negative and positive portions of the square wave
are respectively rectified. Further, when IR= 0, the de-
vice is fully superconducting for the entire square wave,
as its amplitude is smaller than the critical current of
the device. As a result, the entire square wave passes
through the device without dissipation. Interestingly, we
can change the biasing scheme to further utilize the tri-
ode’s three-terminal nature. In Fig. 3b, we continue
measuring VLB but now use ILas a control parameter,
while applying a square wave to IR. As a result, the volt-
age at the left contact, VLB , switches depending on the
sign of the square wave applied to the right contact.
Fig. 4 explores the behavior of the sample as a func-
tion of temperature. The top row demonstrates the ef-
fect of temperature on the map of RLB vs IL,R, first
shown in Fig. 1. At the base temperature of 60 mK
(Fig. 4a), the hysteresis is evident as the central super-
conducting pocket is shifted upward in the sweep direc-
tion. This hysteresis is still evident but much reduced at
1.3K (Fig. 4b), and finally it is clearly absent at 1.9 K
(Fig. 4c). This is important, as hysteresis would prevent
the diode from properly rectifying small currents. The
bottom row presents the IL−VLB curves, measured at
the same three temperatures with the IRadjusted to op-
timally tune the diode efficiency. (The values of IRare
indicated as white lines across the corresponding maps.)
Again, pronounced hysteresis is observed at the lowest
temperature, but it is almost gone at 1.3 K, while the up-
per switching current, I+
c≈1µA, is not much suppressed
compared to the lowest temperature. Further optimisa-
tion may result in suppressing the hysteresis at the base
temperature. However, even the present device could be
placed at the still plate of the dilution refrigerator to effi-
ciently rectify a range of currents in the 0.1−1µA range.