Adhesion is linked with surface forces like capillary pressure and is thus detrimental at the nanoscale where body forces are negligible. It can lead to instant failure during fabrication and operation but it can also lead to overtime failure because of induced friction and wear. However, when it is possible, coating a device with hydrophobic materials reduces drastically that mechanism. Understanding how adhesion works is crucial to design new systems and to enable new technologies. Two models (JKR and DMT) are studied in this paper and model adhesion in different cases. Photolithography and particularly the release step must be carefully designed to prevent contamination and stiction. Materials must be chosen and designed wisely to prevent adhesion failure during operation but lubricants can be used to reduce its impact as well as the impact of friction and wear.

1. Adhesion in MEMS: fabrication and operation issues
Antoine Galand
University of Pennsylvania, Philadelphia, Pennsylvania, USA
Adhesion is linked with surface forces like capillary pressure and is thus detrimental at the
nanoscale where body forces are negligible. It can lead to instant failure during fabrication and
operation but it can also lead to overtime failure because of induced friction and wear. However,
when it is possible, coating a device with hydrophobic materials reduces drastically that mech-
anism. Understanding how adhesion works is crucial to design new systems and to enable new
technologies. Two models (JKR and DMT) are studied in this paper and model adhesion in differ-
ent cases. Photolithography and particularly the release step must be carefully designed to prevent
contamination and stiction. Materials must be chosen and designed wisely to prevent adhesion
failure during operation but lubricants can be used to reduce its impact as well as the impact of
friction and wear.
Introduction
Microelectromechnical systems (MEMS) exhibit a wide range of applications. In fact, from ac-
celerometers in smartphones or in cars to trigger the Airbag system to smart sensors and drugs,
micro- and nanotechnology are changing the way we interact with our environment. For examples,
nanotechnology in sensors drives the industry forward, decreasing fabrication costs, response time
and increasing the reliability and number of different tests on a single tiny platform. Yet we are
only at the beginning of a revolution that holds as much potential as the Digital Revolution did.
However, both fabrication and reliable operations of MEMS are limited by our current tech-
nology. Although nanotribology has been widely used to finely tune properties of materials: in-
creasing roughness to increase hydrophobicity, it plays a major role in the limitation of designs
and application: adhesion, friction and wear are three issues impacting the industry. In this paper,
I focus on adhesion during the fabrication and the operation of MEMS and on the failure mecha-
nisms it leads to.
Adhesion at the small scale has many sources and applications. It explains why a gecko can
walk on a wall but also why certain parts in MEMS will stick to one another. Understanding this
phenomenon is crucial to design MEMS and their fabrication protocols.

2. Adhesion in MEMS: fabrication and operation issues 2
1 The origin of adhesion
Adhesion can be explained by different models depending on the situation and particularly on
the environment and the motion involved. Van der Waals forces can be used to describe most
interactions between two particles and can thus model the interaction between two surfaces.For
example, those weak interactions lead to surface tension, which leads to capillary pressure.
1.1 Surface forces at the small scale
At the macroscale, interactions are dominated by body forces like pressure or gravity, however they
scale poorly and as the dimensions of a systems are shrunk to the microscale, surface forces are
predominant. In fact, as a device is reduced from 1 mm to 1 micron, the area gets a million times
smaller but the volume gets 1 billion times smaller. At this scale, the body forces thus become
negligible. Friction and surface tension are about a thousand times larger than inertial or magnetic
forces [1].
1.2 Surface tension and capillary pressure
Surface tension and contact angle dictate how a liquid surface behaves. For example, if the surface
is hydrophilic, a drop of water will wet the surface and spread, if it is hydrophobic, the drop will
tend to form a sphere and will not spread. This phenomenon leads to the capillary pressure:
Pcap = γ
1
R1
+
1
R2
(1)
Where γ is the surface tension, R1 and R2 are the two curvature radii. This pressure explains
the formation of a meniscus and can drive a fluid, for example in a column, it leads to a capillary
rise of
h = −
2γcos(θ)
gρr
(2)
Where θ is the contact angle. This pressure has also an impact on two close surfaces. If water,
for example, is present between two glass slides the capillary pressure can have two effects. If
both surfaces are hydrophilic, the pressure will bring them closer together: adhesion. If they are
hydrophobic, the pressure will tend to separate them. The pressure in both situation is:
Pcap =
2γcos(θ)
d
(3)
Where d is the gap between the slides, θ < 90◦ for hydrophilic surfaces. This is a first example
of how surface tension induces adhesion between two solid surfaces. In this example, liquid water
was present between the two surfaces to enable this force. As will be discussed later, in a wet
environment water can condense and roughness drives this condensation. However adhesion also
happens in a dry environment.[2]

3. Adhesion in MEMS: fabrication and operation issues 3
1.3 Dry environment
1.3.1 Case study: a sphere on a flat
The sphere on a flat is a common model to explain the interaction between two material using the
Derjaguin approximation: F(D)sphere/flat = 2πRW(D), where D is the distance, R is the radius of
the sphere and W is the interaction energy. Using this approximation the force can be expressed
as:
Ladh = 2πRWAB (4)
Where WAB is the work of adhesion between materials A and B. The work of adhesion is
defined as the reversible work per unit area to separate two materials in contact equilibrium to
infinity in vacuum. Let’s consider a polystyrene sphere (R=0.5 microns) sitting on a PTFE flat:
WPTFE−PS = 2
√
γPTFEγPS (5)
And, Ladh = 0.15µN >> 5x10−15 (gravity). In fact, using the same equations we find that for
the same sphere the radius would need to be larger than 2.7 mm for gravity to overcome adhesion.
If we consider atomically sharp asperities: R=1nm and Ladh=0.3nN, which corresponds to a
tensile stress of 1Gpa which is 100 times larger than the yield stress of bulk PTFE![2]
1.3.2 Adhesive deformation
This last value show adhesive forces can induce deformation as they can exceed the yield stress
of a material. Two theories explain this deformation for different regimes: the Johnson-Kendall-
Roberts (JKR) theory and the Derjaguin-Muller-Toporov (DMT) theory.
Both theories aim to extend the Hertz model to factor in the adhesive forces. According to JKR,
short range adhesive forces lead to a tensile stress that can be added to the Hertz compressive stress.
According to DMT, the tensile stress derived earlier can readily be added to the Hertz compressive
stress. JKR only accounts for perfectly smooth surface and materials with low elasticity with short
range forces, high adhesion, compliant materials and large radii. In the DMT model, the materials
are assumed rigid with long range forces, low adhesion and small radii. Hertz, JKR and DMT
predict a real radius for the contact area of:
rHertz =
3R
4E∗
L
1/3
(6)
rJKR =
3R
4E∗
L+3πRW + 6πRWL+(3πRW)2
1/3
(7)
rDMT =
3R
4E∗
(L+Ladh)
1/3
(8)
Where L is the load, E∗ is the reduced modulus and W is the work of adhesion. Those equations
show the deformation of the sphere that flattens upon contact. This normally happens, as predicted
by Hertz, but when the adhesive forces are factored in, this deformation is larger. The DMT model
is useful for steel, diamond, silicon atomic force microscopy (AFM) probes and similar materials.
JKR can be used to model insects (like a fly walking on a ceiling), geckos, polymers and other
similar configurations.[2]

4. Adhesion in MEMS: fabrication and operation issues 4
1.4 Wet environment
In wet environments, as mentioned, water can condense. In fact, the capillary pressure shifts the
vapor pressure enabling the condensation of water even at low humidity. In that case, capillary
pressure between the two contacting surfaces will drive them closer, if they are hydrophilic and the
total adhesive force is:
Ladh = Lcap +Lsolid−solid = 2πRγL(cos(θ1)+cos(θ2))+2πRWAB (9)
Ladh = 4πR(γLcos(θ)+γSL) = 4πRγS (10)
This derivation assumes a sphere on a flat of the same material. It shows the effect condensation
has on adhesion as Lcap [2].
2 Adhesion in fabrication
2.1 Photolithography: etching the resist
Photolithography is the most used fabrication technique. Photoresists, masks and light exposure
are used together to pattern a wafer. The photoresist is coated on the wafer and pattern using light.
After development, the remaining resist is used for different purposes. It can be used to selectively
deposit a metal layer or to etch a pattern for example. It can also be used as a sacrificial layer to
build cantilevers or other structures over a gap, like membranes. This kind of structures will be
discussed later in this paper. The resist is then stripped or left on the wafer as a structural part.
Adhesion plays a major role in this process as the resist must remain on the substrate with the
desired thickness, both before exposure and after development. Hence, the polymer must have
good interactions with the substrate - usually silicon (Si) or silica (SiO2). This desired character-
istic of a good photoresist implies that it should be removed carefully, not to damage the device
or not to leave debris. The main method is to etch the resist away. However, because of adhesive
forces some residue can be left on the wafer. Hence, each lithography step can lead to a loss in
properties because of resist residue [3].
2.2 Graphene transfer
Graphene sheets are commonly grown on copper using chemical vapor deposition (CVD). The
graphene is then coated with polymethyl methacrylate (PMMA), like a wafer is coated with a
resist. The copper is then removed using bubble separation in an hydroxyde (OH–
) bath. The
graphene is then transferred on a silicon wafer and the PMMA is stripped, like a resist. The
PMMA adheres strongly to graphene and studies show a massive drop in graphene properties after
this step because of residue [4].
2.3 The last step: release
When all layers have been deposited and patterned, and before the encasing, the release step is the
last part of the microfabrication. At this point, each photolithography step has likely left residue
on the device, but it is now ready. The release step is the removal of every remaining sacrificial

5. Adhesion in MEMS: fabrication and operation issues 5
layers to free moving and suspended structures like membranes and cantilever. Those structures are
used as sensors, valves, resonators, relays, capacitors among other things and are thus extremely
useful and widely spread. One of the most commonly used material as a sacrificial layer is SiO2
and is etched using HF. The device is then rinsed as HF is highly toxic and other particles could
contaminate the substrate. Water has traditionally been used to rinse the structure.
The device is now a wet environment and when it is dried, droplets of water will form meniscus
on walls and will tend to collapse suspended structures on the substrate, leading to a confined wet
environment in which water will condense leading to permanent adhesion or stiction [5].
Figure 1: The release of a cantilever beam followed by a water rinse. Water, in blue, is sticking and
condensing, pulling on the beam (blue arrow)[5]
3 Adhesion in operation of MEMS
After the release step, either in normal operation or when hit against a surface, the free suspended
structures can collapse on the substrate. Usually, the restoring force is enough to overcome the
adhesive force, but sometimes it will lead to permanent failure due to stiction. The mechanism
is simple, the adhesive force, determined by JKR or DMT for a dry environment or DMT and
capillary pressure in a wet one, is sticking the two surfaces together. On the other hand the restoring
force, behaving like a spring, is pulling them apart. The larger force determines the behavior.
3.1 The example of Texas Instruments
One famous case of adhesion failure and its solution is the Texas Instruments (TI) Digital Micro-
mirror Devices (DMD). During operation mirrors eventually come in contact with a mechanical
stop. At the macro-scale, this simple design is used almost everywhere and nothing happens. How-
ever, at the micro- and nanoscale, the device is operating in a wet environment. When the mirror
yoke is in contact with the stop, water will start condensing, leading to the formation of a meniscus
and capillary pressure, drastically increasing the adhesive force. The mirror is now stuck on the

6. Adhesion in MEMS: fabrication and operation issues 6
mechanical stop and cannot be operated again. [2]
3.2 Electromagnetic cantilever relay
Figure 2: Electromagnetic relay, the wafer is in the plane of the page. The structure is in gray, the coil is
warped around the left part, the black material is an insulating material.
Let’s consider the relay depicted in Figure 2. This is an horizontal cantilever sitting on a wafer
in the plane of the page. The main material used for magnetism purposes is permalloy, a nickel and
iron alloy. The copper coil can induce an electromagnetic field to deflect the cantilever downwards
until the tip contacts the structure. When contact is made, current can flow through the structure
and a connecting circuit, but it cannot otherwise because of the insulator. When the coiled is turned
off, the cantilever must return to its unbent state to stop the current flow in the connected circuit.
However, depending on the geometry of the tip, the same process as for the TI example can happen
leading to failure.
3.3 Friction and wear
Adhesion can lead to permanent failure instantly but can also lead to the wearing of a system. In
fact, wear is mostly a consequence of adhesion and friction. It can be useful at the macroscale,
to polish a surface with sand paper for example, but like fatigue it can lead to catastrophic failure
over time. Wear drives the smoothing out of materials, the removal of materials and generally the
alteration of the shape and properties of the system.
Moreover adhesion can increase friction too, this is known as adhesive friction. When two
surfaces are in contact, their asperities interact. It can lead to elastic or plastic deformation of the
asperities. Friction is highly dependent on the real area of contact and as discussed previously,
adhesion increases that area.
The wear mechanism involving adhesion is fairly simple. Adhesive forces and plastic flow
deform the surface or even pull out particles from the asperity tips. It leads to the formation of
debris on the substrate, which increase both adhesion and friction. This mechanism is extremely
sensitive to coating. In fact, very clean metallic surfaces undergo high adhesive forces [2].

7. Adhesion in MEMS: fabrication and operation issues 7
4 Solutions to prevent adhesion failure in fabrication
One key parameter in adhesion if the hydrophobicity of the surface. As discussed, hydrophilic
surfaces lead to adhesive forces and hydrophobic tend to lead to less adhesion or even repulsion.
One simple solution, if it does not change the device property is thus to coat the exposed surfaces
(or those which can impact another) with an hydrophobic material. Another trick is to use the resist
as a structural part of the device, this is especially true for SU-8. In fact, SU-8 is difficult to remove
once exposed and is used to build thick layers. It also has good mechanical properties. Hence, if
a device need a wall and some lithography steps require SU-8, it can be interesting not to strip the
SU-8 and use it for the wall [6].
4.1 Removal of residue in lithography
Usually, after the removal of the resist, the fabrication continues with a new lithography step or
deposition. However, techniques are being developed to clean residues after the removal of a resist.
They can be heavy and expensive. Thus, as resist residues are not always a real problem, it may be
better to not follow these protocols. However, when residues are a problem, those steps should be
considered. Some cleaning processes involve the use of active oxygen and cleaning fluid [3].
In the case of 2.5D and 3D fabrication using the Bosh process (or deep reactive ion etching),
after removal of the resist, a fluoropolymer residue usually coats a sidewall and is difficult to strip.
This residue must be removed after each step. A novel cleaning method is proposed by Pollard et
al. to strip the resist and the residue in one step. This method involves a combination soak and
high pressure sprays, coupled with a megasonic rinse [7].
4.2 Using the transfer polymer as the substrate
I am currently working on a hydration sensor using CVD grown graphene on copper as described
before. However, after the graphene growth our fabrication process is different. We do not coat
the graphene with PMMA but we deposit a custom polymer. This deposition step is followed by a
bubble separation and the device is ready. In fact, the polymer used instead of PMMA can also be
used as the substrate. Hence, there is no need to transfer the device to silicon and strip the polymer.
Therefore, there is no residue on the graphene and the graphene properties are good.
I believe this idea could be used with other polymers that could be used both as the separation
material and the final substrate. Such a fabrication is cost effective, quick and clean. However
it is only possible, because we do not need to process the graphene further, our sensor uses the
graphene, bare or functionalized, and silver paint to make the leads. Other devices might require
a few lithography steps afterwards to make a transistor for example, in which case the silicon
substrate might be the best choice.
4.3 Supercritical CO2 after the release
As discussed previously, after the removal of the sacrificial layers, the device must be rinsed to
remove any solvent and contaminants. If the surfaces must be hydrophilic, water might lead to the
adhesion mechanism described in Section 2.3. Supercritical CO2 is used in many industries. One
major breakthrough in microfabrication was the use of supercritical CO2 to dry released devices.

8. Adhesion in MEMS: fabrication and operation issues 8
Water is still used to rinse the structure but is then dried with supercritical CO2 instead of a simple
air dry or nitrogen flush. Supercritical CO2 has the advantages of behaving like a gas and a liquid
at the same time and can be easily liquefied or ”vaporized”. In the supercritical phase the fluid has
no surface tension like in the gas phase, but will flow like a liquid and incorporate contaminants
and water [5]. CO2 is commonly used because of its good supercritical point, but it might be
interesting to look for another solution as CO2 is bad for the environment.
5 Designing to prevent adhesion during operation
A good system design is one that prevents adhesion altogether. At this scale, the design is often
controlled by the available fabrication technologies but that means that a system can also be
designed to avoid adhesion during its fabrication -at least for the release step. As mentioned in
the previous section, hydrophobicity is a great way to prevent adhesion but not all devices can be
made hydrophobic.
5.1 The TI solution
Faced with the adhesion failure of their DMD, TI studied the mechanisms of adhesion and re-
designed their devices to prevent it. They first enclosed the whole system in a dry environment to
minimize the meniscus formation. The yoke and stop were both coated with hydrophobic ”anti-
stick” material minimizing Van der Waals forces. The best trick they used was energy storage.
They added miniature springs between the yoke and the mechanical stop. When a mirror comes to
a rest on its stop, its spring contracts and stores mechanical energy. When the mirror is actuated
the other way, the spring provides the extra energy required to overcome adhesion by expanding.
In fact, energy storage is one of the most used techniques to overcome such situations or to
actuate systems. For example, a parallel plate capacitor can be used to tune a gap by actuating the
plates. However, the electrostatic force between those plates can only bring them closer together
and cannot pull them apart. Springs are commonly used to store energy when the plates get closer
and to pull them apart by releasing this energy.
5.2 Tuning the geometry of the tip
This concept of energy storage is important in the case of the relay described in 3.2. The magnetic
force will bring the tip in contact with the wall but cannot pull it apart. However a cantilever
beam behaves like a spring. During its deflection towards the wall, the beam stores energy and
when the magnetic force is turned off, the beam will go back to its original position, if it deformed
elastically. This kind of structure -cantilever beam- is used extensively at the microscale in large
deflection behaviors. They can be used as relays like this one, as valve if the tip closes a channel,
as sensors, accelerometers and capacitor comb-drives for example.
The beam should always be able to go back to its natural state. It is thus important to understand
the friction mechanism to prevent its failure. In the case of the relay, the only force balancing
adhesion is the spring restoring force, if it’s too low it cannot be pulled back. However, in that
application, the tip geometry does not matter, only a small point of contact is required, though an
atomically asperity might lead to arcing. The tip can thus be designed carefully to prevent adhesion

9. Adhesion in MEMS: fabrication and operation issues 9
- with a small enough radius for example.
For the valve, there are two cases. First, the fluid is trying to push though the hole against the
tip, in that case, the combined restoring force and pressure should overcome adhesion. However if
the fluid is trying to push though the hole along the tip, its pressure can actually enhance adhesion.
The tip or the whole system must be redesigned to account for that pressure.
5.3 Lubricants
As explained in section 3.3, the coating of a surface influences the friction and wear behavior of
a system. Two clean metallic surfaces in contact will experience very high adhesion. However,
the two same surfaces coated with the right lubricant will experience significantly lower adhesion,
friction and wear. Lubricants are widely used at the macroscale - during both fabrication and oper-
ation. For example, they are used during machining to prevent heat, in bearings to enable motion
and so on.
Lubricants can also be used in MEMS during operation. Organic coatings are used to reduce
friction and adhesion in MEMS but two major breakthrough are the alcohol adsorption lubrification
and the use of ionic liquids. The first one uses alcohol vapor to coat a system with a replenishing
lubricant that can be used when no electric contact is required [8]. The second one uses ionic liq-
uids as lubricants. This is particularly interesting in applications where electric contact is required,
as the ionic fluid are conductors [9].
Conclusion
Adhesion is detrimental at the micro- and at the nanoscale. As surface forces prevail at this scale,
surface tension, electrostatic interactions and Van der Waals forces are significantly larger than
body forces like gravity or pressure. Depending on the materials and the geometry adhesion can be
explained by different models. Two of which are the JKR and the DMT models. The JKR is great
to explain short range effects and high adhesion in the case of the gecko for example. The DMT
theory is best suited for long range effect, low adhesion and stiffer materials like AFM probes. In
both cases, adhesion increases the contact surface and deforms the impacting materials. In a wet
environment, a shift in vapor pressure enables condensation of water even at low humidity. This
water will form meniscus and the capillary pressure will drastically increase adhesion. However,
for hydrophobic surfaces these phenomena are not as readily observed. Adhesion can drive the
removal of materials and the formation of debris leading to higher wear rates and friction. Both
wear and friction are tribology issues that can lead to failure. The wear in MEMS with rubbing
surfaces can lead to failure within seconds. Adhesion can lead to failure by disabling the operation
of a system when two surfaces impact each other, this is an ”instant” failure but can lead to a failure
overtime through induced friction and wear.
During fabrication, adhesion leads to the contamination of wafers with residues after photore-
sist removal, contamination of graphene after removal of PMMA. Those residues can drastically
impact the performance of a system and may need to be removed. The release step is crucial as it
is the last one before enclosing the device and is the one in which free and suspended structures
are released. Rinsing is required after this step, but if water is used and dried using conventional
methods, meniscus will form driving the condensation of water and adhesion failure. However

10. Adhesion in MEMS: fabrication and operation issues 10
supercritical CO2 can be used to dry the device.
During operation, when two surfaces come into contact adhesion can occur. This was the case
in TI’s DMD where mirrors come to a rest on a mechanical stop. This device was failing be-
cause the mirror would stick to the stop. However by enclosing it in a dry environment, using
hydrophobic coatings and by storing energy using springs, they managed to overcome this failure
mechanism. Geometry is a key parameter that can be tuned in some devices to prevent adhesion.
Finally lubricants can be used to reduce adhesion, friction and wear.
References
[1] W. Zhang, G. Meng, and H. Li, “Electrostatic micromotor and its reliability,” Microelectronics
Reliability, vol. 45, no. 7, pp. 1230–1242, 2005.
[2] C. M. Mate, Tribology on the Small Scale: A Bottom Up Approach to Friction, Lubrication,
and Wear. Oxford: Oxford University Press, 2007.
[3] M. Chen, Y. Huang, W. LIAO, H. Hsiaw, and C. Shen, “Method and equipment for removing
photoresist residue after dry etch,” Sept. 11 2014. US Patent App. 13/785,172.
[4] Y.-C. Lin, C.-C. Lu, C.-H. Yeh, C. Jin, K. Suenaga, and P.-W. Chiu, “Graphene annealing:
How clean can it be?,” Nano Letters, vol. 12, no. 1, pp. 414–419, 2012. PMID: 22149394.
[5] R. Maboudian, “Adhesion and friction issues associated with reliable operation of mems,”
MRS Bulletin, vol. 23, no. 6, pp. 47–51, 1998.
[6] P. Abgrall, V. Conedera, H. Camon, A.-M. Gue, and N.-T. Nguyen, “Su-8 as a structural mate-
rial for labs-on-chips and microelectromechanical systems,” Electrophoresis, vol. 28, no. 24,
pp. 4539–4551, 2007.
[7] K. Pollard, M. Guo, R. Peters, M. Phenis, L. Mauer, J. Taddei, R. Youssef, and J. Clark,
“Efficient tsv resist and residue removal in 3dic,” Additional Conferences (Device Packaging,
HiTEC, HiTEN, & CICMT), vol. 2014, no. DPC, pp. 001435–001469, 2014.
[8] D. B. Asay, M. T. Dugger, J. A. Ohlhausen, and S. H. Kim, “Macro- to nanoscale wear
prevention via molecular adsorption,” Langmuir, vol. 24, pp. 155–159, 01/01 2008. doi:
10.1021/la702598g.
[9] M. Palacio and B. Bhushan, “Ultrathin wear-resistant ionic liquid films for novel mems/nems
applications,” Advanced Materials, vol. 20, no. 3, pp. 1194–1198, 2008.