This document details the design and fabrication of a MEMS electromagnetic relay capable of a 1 mm stroke, utilizing a cantilever beam and a vertical coil structure. The device, made from copper and nickel-iron permalloy, achieves a force of 1μN with a current requirement of 1.41 mA while being CMOS compatible to minimize costs. Key considerations for optimization include reducing fabrication complexity and ensuring correct alignment to improve production efficiency.

1. AN ELECTROMAGNETIC MEMS RELAY CAPABLE OF 1 MM STROKE
Andrew Boodhoo, Meiyang Cui, Antoine Galand, and Shu-Yu Lai
University of Pennsylvania, Philadelphia, Pennsylvania, USA
ABSTRACT
MEMS relays provide reliable and inexpensive solu-
tions to operate a circuit that requires a large current. We
provide the design and fabrication of a magnetic relay that
uses a space-saving coil capable of producing an actuation
force of 1 µN and a stroke of 1 mm. The method of
action is a cantilever beam that deflects in-plane with the
wafer. The device should require less than 1.5 mA to bring
the beam into contact with the opposing part, effectively
turning the switch on. The coil is made of copper and the
cantilever beam is made of a Ni/Fe permalloy. The relay
is fabricated with a sacrificial layer underneath the beam,
so that when it is removed, the beam can actuate. A layer
of SU-8 between the magnetic core and conductive coil is
used to provide electrical isolation. Our fabrication process
is CMOS compatible.
INTRODUCTION
Traditional mechanical relays are electromagnetically-
operated switches. In a relay, a small electric current is
used to operate a circuit that requires a much larger one.
This essentially bridges the gap to make small currents
activate larger ones.
A relay consists of four components: an electromag-
net, two circuits, and a power supply. The electromagnet is
typically a coil of wire.
The operation of a relay is shown in Figure 1. In this
diagram, power flows through the first circuit (1), which
activates the electromagnet. A magnetic field is generated,
which produces a force according to the right-hand rule.
The contact in the second circuit (2), which is normally
open, is attracted to the electromagnet, which completes
the circuit and turns it on. When the power is switched off,
the contact returns to the original position, thereby turning
off the circuit [1].
Figure 1: Simple operation of a relay [1].
There are two main types of relays: normally open
(NO) and normally closed (NC). In an NO relay, the second
circuit has an open loop and is switched off by default, and
then turns on when current flows through the first circuit.
In an NC relay, the second circuit is has a closed loop and
is switched on by default, and then turns off when current
flows through the circuit.
There are many applications that can take advantage
of the relay’s important feature of being able to control an
output circuit that has a much larger current than the input
circuit. One notable example is an electronic thermometer
circuit that can sense minute changes in temperature. A
relay can bridge this circuit with devices that require a
much larger current than would be produced from a small
temperature change, such as fans or air conditioning. This
system can be used in a computer server room to maintain
a constant temperature, ensuring that the servers do not
overheat.
Another notable example is a sensor circuit that can
detect when a sensor has been activated. A relay can bridge
this circuit with devices that require a much larger current
than would be produced from the tripping of a sensor, such
as alarms or security cameras. This system can be used in
home security to provide accurate real-time monitoring.
The motivation behind using MEMS relays rather than
traditional mechanical relays is that they share the same
advantages, but their size, cost, and switching time are
greatly improved. Furthermore, they can be combined with
smaller electronic components.
A study by Ruan and Shen used an electromagnetic
MEMS actuator with the coil designed in the plane of
the substrate, as shown in Figure 2. This design has a
drawback of covering a substantial area of the wafer,
especially for a large number of turns in the coil. This is
because every subsequent turn has a larger diameter than
the previous one. Surface area is a precious commodity on
a silicon wafer, so a device designed in this manner with a
large number of turns would cost a significant amount.
Figure 2: Top view of a latching relay. Planar coils are
integrated on the substrate underneath the cantilever [2].
A study by Ahn and Allen used a two-layer coil
design,for an inductor which can be integrated into a relay,
as shown in Figure 3. While this has an advantage over
the design by Ruan and Shen in that it effectively doubles
the number of turns of the coil that can be fabricated in the
same area of wafer, it still has the disadvantage of having
subsequent increasing diameter turns. A large number of
turns in the electromagnetic coil are important because it
allows the circuit to achieve a high inductance and high
magnetic flux density [3]. We therefore designed our relay

2. to optimize the number of coil turns that can fit in the
smallest area of wafer while minimizing the complexity,
and therefore cost, of the fabrication sequence.
Figure 3: Schematic diagram of a spiral-type inductive
structure, which has a two-layer magnetic coil design [3].
DESIGN
Objectives
The main objective of our study is to design a method
to produce relays on a wafer that a vertical coil rather than
the horizontal ones that are classically used. We also had
specific requirements to meet when designing our device.
We had to generate a force of 1µN with a stroke of 1mm.
Structure of the device
When designing the MEMS relay, we chose to use
soft magnetic materials. In order to actuate the device, two
parts are thus required. First, a coil is used to generate the
magnetic flux. Second, the device itself must me made of a
magnetic material and will also act as the core of the coil.
Because of its high permeability and magnetic
properties, we chose to use permalloy for the core. This
material is nickel-iron alloy, consisting of 80% nickel and
20% iron. For fabrication purposes, the coil is made of two
materials; the top and bottom part are both made of copper
and the vertical parts are made of permalloy.
As shown in Figure 4, the relay is a cantilever beam
sitting on its side on the substrate. The beam can bend
towards the bottom part of the device. The insulating gap
prevents current from flowing through the structure and
through a circuit to be connected to the bottom part on
each two sides of said gap-the switch is off.
Figure 4: Structure of the relay. The coil is wrapped around
the left side of the device. The top red circle shows a small
gap between the cantilever beam and the wafer. This beam
is able to deflect, every other part of the design is anchored
on the substrate. The bottom red circle shows the gap serv-
ing as the insulator. The white arrow describes the deflec-
tion of the beam. The contacts are shown in blue.
However, when the beam deflects and makes contacts
with the bottom part, current will flow- the switch is now
on. The dimensions of the device will be discussed in
the next section, with the mathematical model. The gap
shown in the top red circle enables the deflection of the
beam by separating it from the wafer. Without the gap,
the beam would be fixed. If it is too small, such that the
beam rests on the wafer, high friction and adhesion will be
experienced, drastically reducing the device performances
and leading to wear.
Mathematical model
In this section, we discuss the dimensions of our relay
and the magnetic force needed to actuate the device.
First, we need a stroke of 1 mm, thus to be in the small
deflection regime, we need the beam to be at least 2.5 mm
long. We chose to make the device as small as possible and
thus decided to make a 2.5 mm long beam. We arbitrarily
chose an insulating gap g=1 µm, which enables us to input
lower currents.
The other dimensions have been determined using the
small deflection of a cantilever beam theory. The vertical
deflection is calculated both for a beam under a point load
at the tip (1) and a distributed load (2).
y =
FL3
3EI
(1)
y =
PL4
8EI
(2)
Where I =
bh3
12
(3)
Where PL and F are the weight of the beam. The di-
mensions are summarized in Table . The horizontal deflec-
tion is the stroke and corresponds to a point load at the tip of
1 µN. Using (1) with an adapted moment of inertia I = hb3
12
enables us to adjust the different dimensions to balance the
vertical deflection and the required load, which should be
around 1 µN. We found that the beam would only deflect
a dozen of nanometers towards the substrate under its own
weight.
Beam (µm) Tip (µm)
Length L=2500 l=20
Height h=100 100
Width b=1.9 100
Table 1: Summary of the key dimensions
With those values, we can now determine the current
required to provide the 1 µN force. Because the permeabil-
ity of permalloy is several orders of magnitude larger than
that of air, we can neglect the reluctance of the material.
We then have:
g+x
µ0hl
=
Ni
Φ
(4)
Where x is the stroke, N the number of turns and i the
current. The coenergy is thus:
W =
i2
2
µ0lN2h
g+x
(5)
Hence, the maximum force is given by (at x=0):
Fmax = −
µ0lh
2g2
(Ni)2
(6)
Which leads to N ∗ i = 0.0282 turns ∗ A Using the plot in
Figure 5, it is apparent that 15-20 turns is the suitable range.

3. In fact, the number of turns are also restricted by the fab-
rication process: we need to fit the turns in the device and
each turn increases the fabrication cost. Taking these fac-
tors into account and to have a minimal current, we chose
N=20 turns and i=1.41 mA. Using this number of turns en-
ables a coil cross-section of 10x10 µm2. Thus, current den-
sity is then about 14 A*mm−2 which is smaller than the
maximum current density of copper.
However, we calculated that a deflection of 1 µm of
the beam would require a force of 1nN. Given the geom-
etry and the 1 mm gap, this requires 45 mA. This current
density is about 10 times that of copper. We will need to
pulse the current into the coil in order to deflect the beam
enough to be able to use a more reasonable current.
Figure 5: Current versus number of turns.
FABRICATION
A fabrication process of this magnetic actuator is
demonstrated in figures 6, 7 and 8. The following de-
scriptions are used to reveal specific details for better
understanding. The design idea of the process focuses on
two points [4]. First, it is CMOS compatible so each step
will be less than 400◦C. Second, simplify the process steps
to decrease the overall cost without compromising the
performance.
Bottom coil structure
Figure 6: Illustration of the fabrication of the bottom part
of the coil.
We start with a silicon wafer < 111 > with a Si3N4 thin
layer on the surface. Si3N4 films can be obtained by many
techniques: by CVD, by chemical transport, by direct ni-
tridation of a silicon substrate. The CVD method can grow
uniform and adherent layers without defects and impuri-
ties and the process uses comparatively simple and cheap
equipment. Si3N4 with thickness of 1500 ˚A for insulation
is good enough for this device. Then, we use PECVD to de-
posit the sacrificial SiO2 layer on the top of the Si3N4 with
a thickness of 3000 ˚A. On this substrate, positive photore-
sist (Microchemical AZ 701 MiR [5]) with a thickness
of 8000 ˚A is spin-coated at 5000 rpm on the surface, then
patterned and developed to open an area for the sacrificial
layer, see Fig 6.a.
This process was designed for three purposes. First,
to leave space for the cantilever horizontal movement for
actuation function. Second, to prevent the cantilever beam
from friction in the vertical dimension. Third, to create the
electrical isolation gap in the sidewall.
Then, the same positive photoresist is spun on the sub-
strate, patterned and developed to create the bottom coil
structure, see Fig 6 b to f. Then a 2000 ˚A thick electroplat-
ing Cu seed layer is sputtered directly onto this photore-
sist. By removing the photoresist with AZ Kwik-Strip ,
the seed layer, which does not have direct contact with the
wafer, dissolves away with the photo resist. Here, it is im-
portant to note that the seed layer on the top of the photo
resist cannot be alternated. Putting the seed layer before the
photoresist might cause short circuit. Next, SU-8 (Micro-
chemical SU-8-5) is spun on the wafer at 3000 rpm to build
an electroplating mold for the bottom conducting coils.
The electroplating mold is then filled with permalloy
using standard electroplating techniques and the nickel-iron
electroplating bath described in Table 2. A thin layer of
1µm of SU-8-2 is spun again at 3000 rpm and used as in-
sulating material to separate the bottom coil from the mag-
netic core.
Component Quantity (g∗L−1)
NiSO4 · 6H2O 200
FeSO4 · 7H2O 8
NiCl2 · 6H2O 5
H3BO3 25
Saccharin 3
pH 2.5-3.0
Temperature (◦C) 25-30
Current density (mA∗cm−2) 13
Table 2: Composition of the permalloy electroplating solu-
tions [6].
Component Quantity
CuSO4 · 6H2O 1200 g∗L−1
H2SO4 100 mL∗L−1
Table 3: Composition of the copper electroplating solu-
tions.
Magnetic core and vertical coil
Another layer of photoresist is spun, patterned
and developed to open electroplating windows for the
sidewall, the cantilever beam and the center magnetic
core. The 2nd seed layer consisting of 200 ˚A of chromium,
4000 ˚A of copper, and 400 ˚A of chromium is deposited
using electron beam evaporation or sputtering which will
only cover the open window for the magnetic structure.
After removing the photoresist, the thick layer SU-8 of 100
µm is spin-coated (as above) at 3000 rpm, and hardbaked

4. Figure 7: Illustration of the fabrication of the magnetic
core and the vertical part of the coil.
at 300◦C for 30 min. In order to insulate the bottom
magnetic core from the conductor coil, a thin layer of SU-8
is spun and developed into the mold for the following
electroplating step of Ni/Fe Permalloy. We are using the
permalloy for both the core and the vertical coil structure.
Since we are using DC current for the actuation, two
different materials for the coil will not affect the final
performance. The fabrication process is simplified greatly
without forming separate molds for the Cu coil structure.
Then, as mentioned in the previous process, a thin layer of
SU-8 is used for insulation between the core and the top
coils.
Top coil and structure release
Figure 8: Illustration of the fabrication the top part of the
coil and the release of the structure.
The same process is used to open windows for the top
coil. The top coil need to be perfectly aligned with the
vertical coil. Then the photoresist is removed and thick
SU-8 mold is created for the electroplating process. The
top conductor is Cu. Then we use SU-8 to fully cover
the coil structure and open two windows, since we need
the current to flow into the coil structure to generate the
magnetic force. The last step is to release our cantilever
beam structure and to create the room for the beam to move
in plane. We immerse the device into 49% HF solution and
the SiO2 sacrificial layer will be completely removed.
FUTURE CONSIDERATIONS
There are some considerations that we should keep in
mind for optimizing our production process. The first is
that our fabrication process is complicated with many steps
and masks. This makes it quite error-prone and therefore
may affect our percent yield of successful devices during
batch fabrication.
The second consideration is that all turns of the coil
must be connected in order for current to flow. Therefore,
we need perfect mask alignment between fabrication steps
and we may need to use optical proximity correction to get
small feature sizes. This is one of the reasons that we did
not choose more than 20 turns for our coil.
The third consideration is that the device has an
irregular shape. This may make it difficult to optimize the
number of structures that can fit on a single wafer and their
orientation for mass production.
CONCLUSIONS
This paper presents the design of a novel MEMS
electromagnetic relay. The design consists of a cantilever
that actuates horizontally with respect to the substrate
and a coil wrapped around the cantilever sidewall. The
cantilever is made of a nickel-iron permalloy, and the coil
is made of copper. The device produces an electromagnetic
force of 1 µm with a stroke of 1 mm.
In order to minimize the cost of the device while
also minimizing the required current, we determined the
optimal number of turns of the coil to be 20. This requires a
current of 1.41 mA. The fabrication process to produce the
device is CMOS compatible. It is also designed to decrease
the overall cost without compromising the performance.
To optimize the production process, we should keep in
mind the complicated fabrication process, the connections
between the vertical sidewalls and top and bottom of the
coil, and the irregular shape of the device.
REFERENCES
[1] C. Woodford, “How do relays work?.” http://www.
explainthatstuff.com/howrelayswork.html,
2009.
[2] M. Ruan and J. Shen, “Latching micromagnetic re-
lays,” Microelectromech, Syst, vol. 10, pp. 511–517,
2001.
[3] C. Ahn and M. Allen, “A planar micromachined spiral
inductor for integrated magnetic microactuator applica-
tions,” Micromech. Microeng, vol. 3, pp. 37–44, 1993.
[4] C. H. Ahn and M. Allen, “A new toroidal-meander type
integrated inductor with a multilevel meander mag-
netic core,” Magnetics, IEEE Transactions on, vol. 30,
pp. 73–79, Jan 1994.
[5] MicroChemicals gmbh, “Dry etching with photoresist
masks,” 2013.
[6] J. Y. Park and M. G. Allen, “Development of magnetic
materials and processing techniques applicable to in-
tegrated micromagnetic devices,” Journal of Microme-
chanics and Microengineering, vol. 8, no. 4, p. 307,
1998.