Table 1.

Comparison of non-volatile relays.

Reference	20	13	14	15	16	Our device
Architecture	Out-of-plane See-saw type beam with torsional hinges	In-plane straight cantilevered beam with two gates.	In-plane straight cantilevered fin with two gates.	In-plane straight cantilever with two gates.	In-plane straight cantilevered beam with two gates.	In-plane curved beam with quad gate architecture.
Footprint	500 μm × 25 μm	~3 μm × 2 μm	2 μm beam	0.5 μm beam	1.2 μm beam	~5 μm × 10 μm
Actuation airgap (nm)	450	40	80	25	30	120
Pull-in/programming voltage	3.1 V	1.05 V	10 V	17 V	3–4 V	Various depending on hinge type and offset. 1.6 V for serpentine hinge with 1.2 μm offset.
Reprogramme-ming voltage	N/A	1.95 & 2 V	~12–15 V	25 V	3–4 V	Vrep = Vpo = 7–8.7 V for device with Vpr = 3.5 V
Non-volatile cycles	0	2 at room temp.	11 at 50 °C	1 at room temp.	1 at room temp.	42 at 200 °C and 20 at room temp.
Contact material	Ni on Cr–Au	Al	Si doped with As	Si doped with As	Al	Ti
Switching time	Not available.	Not available.	Not available.	Not available.	Not available.	Prog.: 511 ns; and 954 ns for straight, serpentine hinges; reprog.:  ~1.5 × prog. time (simulated)
Comment	Stiction recovery in first cycle only. Long beam (0.5 mm) and large gate area to achieve 3 V pull-in.	Reliability not mentioned. Low stiffness of Al (Young’s modulus  ~1/3rd that of Si) and small airgap contribute to low pull-in.	Pull-in increases with cycling, failure through microwelding. Similar footprint but much higher device layer thickness (3.5 μm) for a much larger gate area than our device (which has a 300 nm device layer thickness).	Reliability, failure mechanism not discussed.	Complicated approach using charge trapping to alter the pull-in voltage. Reliability and failure mechanism not discussed. Cannot be used at high temperatures due to requirement for charge trapping.	Failure mode is contact resistance increasing. Mechanical failure not observed until experiments were halted, or high overdrive is applied, causing beam to collapse.