is anisotropic and varies from 0.80 to
0.95 MPa m1/2. MEMS, like microelectronics,
use chemical vapor deposition to deposit
polysilicon films. A variety of microstruc-
tures can be produced, depending on the
deposition parameters and postdeposition
heat treatments.5 Micron-scale specimens
are fabricated, a sharp pre-crack is produced
by indentation, and the specimen is loaded
using an integrated MEMS electrostatic
actuator.6 The actual indent takes place on
the substrate, away from the specimen,
and the crack propagates into the speci-
men; in this way, the indentation-induced
stresses and damage are avoided. The test
structure is shown in Figure 1, and the re-
sults are presented in Table I.
On-Chip Testing
of Mechanical
Properties of
MEMS Devices
Note that single-crystal silicon specimens
can also be fabricated using silicon-on-
H. Kahn, A.H. Heuer, and R. Ballarini
The field of microelectromechanical
ingly difficult. The ability of MEMS struc-
tures to be reliably released from their
substrates, actuated, and then to move in
a controlled fashion permits a tremendous
range of experiments; essentially, any meas-
urement taken at the macroscale can be
analogously repeated at the microscale. In
addition, the batch fabrication inherent to
MEMS processing creates a multitude of
identical specimens that can generate data
with statistical significance.
Consider the effect of microstructure on
the fracture toughness, Kcrit, of silicon, which
is the most widely used material for MEMS
devices. Bulk specimens have typically
been tested by machining a centimeter-
scale bend bar, producing a sharp “pre-
crack” by indenting or preloading, and
then loading in tension or in bending. In
this manner, single-crystal3 as well as
polycrystalline4 silicon (polysilicon) have
been tested. In the latter case, the direc-
tionally solidified polysilicon grain size
(ꢀ1 mm) was much larger than the flaw
size; furthermore, the fracture toughness
systems (MEMS) involves the interaction
of the physical environment with electri-
cal signals through the use of microbatch-
fabricated devices. MEMS is a growing
technology, and commercial MEMS prod-
ucts are becoming commonplace.
To optimize MEMS designs, materials
properties must be thoroughly character-
ized and controlled. Device designers need
to know the allowable strain limits as well
as Young’s modulus, fracture strength and
fracture toughness, thermal conductivity,
and other key properties. For extrinsic ma-
terials properties, the statistical scatter is
critical for determining reliability safety
factors. While these values are generally
known for bulk materials, it is not clear
that they will be valid for materials with
the characteristic microstructures and mor-
phologies that result from MEMS proc-
essing. Also, as devices shrink, the ratios
of surface area to volume become so large
that surface properties, such as adhesion
energy and friction, can control the per-
formance (or failure) of MEMS devices.
(For more on the tribology challenges of
MEMS fabrication, see the article by de
Boer and Mayer in this issue.)
Figure 1. MEMS device used for
measuring fracture toughness of
polysilicon. The diamond-shaped
Vickers indent near the specimen is
artificially outlined in black. The area
around the indent is shown in higher
magnification in the inset.
Table I: FractureToughness (Kcrit) Results for MEMS-Fabricated Si Specimens
with Varying Microstructures.
While it is obvious that the mechanical
properties of MEMS materials must be
measured to enhance device performance,
it is equally true that the fabrication and
operating techniques developed for MEMS
can be exploited for performing basic
materials-properties investigations. Previ-
ous efforts aimed at determining mechani-
cal properties at micron-sized scales have
been severely limited. Thin films on sub-
strates can be studied using nanoindenta-
tion and microscratching with diamond
indenters.1 Freestanding films can also be
tested with specially designed equipment,2
although specimen handling is exceed-
Undoped
Multilayer
Polysilicon
Undoped
Polysilicon
(equiaxed
p-doped
Polysilicon
(equiaxed
Undoped
Polysilicon
(columnar
Undoped
(nine alternating
Amorphous grain diameter grain diameter grain diameter columnar and
Silicon
ꢀ0.1 ꢀm)
ꢀ0.7 ꢀm)
ꢀ1.0 ꢀm)
equiaxed layers)
Average Kcrit
1.0
1.0
1.1
1.0
0.9
(MPa m1/2
)
Standard deviation
0.3
8
0.1
12
0.2
7
0.1
6
0.1
11
(MPa m1/2
)
Number of tests
Note: All film thicknesses are 2–4 ꢀm.
300
MRS BULLETIN/APRIL 2001