JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 6, NO. 4, DECEMBER 1997 303
Microfabrication of Oxidation-Sharpened
Silicon Tips on Silicon Nitride
Cantilevers for Atomic Force Microscopy
Albert Folch, Mark S. Wrighton, and Martin A. Schmidt
Abstract—We have developed a novel process for the micro-
fabrication of atomic force microscope (AFM) cantilevered tips
from silicon-on-insulator (SOI) wafers. The tip and cantilever
are made of crystalline silicon and low-stress silicon nitride,
respectively. This choice of materials allows us to sharpen the
tips by oxidation sharpening without affecting the cantilever. We
evaluated their performance in contact mode during imaging
of artificial nanostructures and compared them to commercially
available ones. The images acquired with our tips feature superior
resolution on those samples. 
Index Terms—Atomic force microscope, microfabricated can-
tilevers, microfabricated tips, silicon-on-insulator wafers.
nanometer scale. An AFM consists of a sharp cantilevered tip
that is scanned within nanometric proximity over a surface.
The forces between the tip and sample cause a minute can-
tilever deflection, which is sensed to obtain a topographical
map of the surface on a nanometer or atomic scale. Since its
invention, the AFM has been used not only to view surface
structures, but also to probe electrical, magnetic, van der
Waals, adhesion, and chemical interactions between the tip and
surface and to modify surfaces by means of those interactions
(see, for example,  and references therein).
The sharpness of the tip is often a fundamental resolution-
limiting parameter. When the tip and sample are in contact, the
area of contact increases with the radius of curvature of the tip
apex. During noncontact imaging, on the other hand, sharper
tips can be brought into closer proximity with the surface
before a jump-to-contact instability occurs . Furthermore,
the sample features appear widened or convoluted by the tip.
Microfabrication is necessary because it is desirable that
the cantilever have low stiffness for higher sensitivity and
high resonance frequency for better immunity to vibrations .
HE ATOMIC force microscope (AFM)  has become a
powerful tool for investigating surfaces on an atomic or
Manuscript received May 21, 1997; revised August 5, 1997. Subject Editor,
R. B. Marcus. This work was supported by the National Science Foundation
and the Direcci´ o General de la Recerca, Catalonia, Spain, and Motorola, which
supplied the SOI wafers.
A. Folch is with the Department of Chemistry and the Microsystems Tech-
nology Laboratory, Massachusetts Institute of Technology (MIT), Cambridge,
MA 02139 USA (e-mail: email@example.com).
M. S. Wrighton was with the Department of Chemistry, Massachusetts
Institute of Technology (MIT), Cambridge, MA 02139 USA. He is now with
Washington University, St. Louis, MO 63130 USA.
M. A. Schmidt is with the Microsystems Technology Laboratory, Massa-
chusetts Institute of Technology (MIT), Cambridge, MA 02139 USA.
Publisher Item Identifier S 1057-7157(97)08221-8.
However, silicon micromachining techniques used to make the
cantilevered tip do not generally ensure nanometric sharpness
of the tip. Marcus et al. ,  observed the anomalous
oxidation of crystalline Si at sharp edges and low temperatures
(900 –950 C) and used it to sharpen Si tips for field-emission
devices. By repeatedly growing SiO
stripping it with HF, they achieved
the tip apex. Recently, Itoh et al.  implemented the concept
of oxidation sharpening in AFM Si tips on SiO cantilevers.
Their process, however, is limited in the sense that the tip
cannot be arbitrarily sharpened because dissolution of the
oxide grown in the sharpening process attacks the cantilever.
Therefore, their cantilevers cannot be arbitrarily thin, which
rules out their process for the microfabrication of thin low-
compliance AFM cantilevers. The same limitation applies to
crystalline tips integrated onto crystalline Si cantilevers .
Here, we present in detail a simple process that allows
us to microfabricate silicon nitride cantilevers with integrated
Si tips, which can be arbitrarily sharpened without affecting
the cantilever. Since the cantilever material is deposited, the
thickness of the lever can be specified with great accuracy.
on the Si tips and
1-nm radii of curvature at
II. CANTILEVERED TIP FABRICATION
The process is sketched in Fig. 1. We start with a double-
side polished silicon-on-insulator (SOI) wafer.1The front side
of the SOI wafer consists of a 5- or 10- m-thick (100)-oriented
Si layer (which we call the “SOI layer”) on a 0.5–1- m-thick
buried SiO film, which is grown on a
oriented handle wafer. The dopant type and concentration are
not relevant to our process. The tip is created from the SOI
layer, as explained below. The periodicity of the pattern (and
therefore the size of the AFM cantilever chip) is 3500
3600m. The process consists of four major steps: 1) bulk
micromachining to create a silicon membrane; 2) plasma etch
to create a bilayer cantilever; 3) wet isotropic etch to create a
tip; and 4) oxidation sharpening of the tip.
In all photolithographic steps, positive photoresist was spun
onto the wafer with thicknesses ranging from 1 to 7 m, baked
at 90 C for 30–60 min, then exposed to UV light through a
contact chrome mask, and developed and baked at 120 C for
30–60 min. The patterns were transferred either onto silicon
nitride or onto Si with a SF -based plasma etch or onto SiO
with a buffered HF etch. Photoresist is always removed by
500- m-thick (100)-
1Courtesy of Motorola, Inc.
1057–7157/97$10.00 1997 IEEE
304 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 6, NO. 4, DECEMBER 1997
the microfabrication process (see text). In all drawings, the top corresponds
to the wafer front side.
Cross-sectional schematic representation of the wafer at each step of
immersing the wafers in a piranha solution (H SO :H O ::3:1)
for 10 min.
A. Bulk Micromachining
To avoid wafer-edge erosion during the bulk micromachin-
ing step, we confine the pattern to an inner circle
from the edge of the wafer. In addition, to help preserve the
integrity of the wafer throughout the process and to enable
handling of the wafer with vacuum wands, an area of
at the center of the wafer is not patterned.
First, a thin ( 430 ˚ A) stress-relief oxide (SRO) layer is
thermally grown and low-pressure chemical vapor deposition
(LPCVD) Si N ( 1500 ˚ A) deposited on it [Fig. 1(a)]. The
nitride/SRO layer is photolithographically patterned to expose
rectangular areas (975
side of the wafer. We use the wafer flat to align this mask
with the crystallographic axis of the wafer. These exposed
areas are etched in a 20% (% weight) potassium hydroxide and
water (KOH) aqueous solution at 80 C, while the front side is
protected from etching by the nitride/SRO layer. The measured
etch rate is
90 m/h. Since SiO etches
than Si(100) , the etch stops when the buried SiO layer
is reached. Thus, the formed pit (as observed from the back
side) has a flat bottom and (111)-oriented-inclined side walls
characteristic of the anisotropic KOH etch [Fig. 1(b)]. We then
proceed by selectively removing the nitride in a hot phosphoric
acid bath (175 C, 45 min) and by dissolving both the exposed
areas of the buried SiO layer and the front-side SRO layer in
buffered HF (5–10 min) [Fig. 1(c)]. At this point, the wafer
consists of an array of bare crystalline-Si membranes of a
thickness approximately equal to the starting SOI layer and
350 m 2000 m in size. The exact size depends on the
thickness of the starting wafer and does not affect our process.
2600 m) of Si on the back
100 times slower
B. Cantilever Pattern
The cantilever material, Si-rich nitride, is deposited on the
wafers in an LPCVD furnace at 775 C and 280 mTorr with
a mixture of SiH Cl
intrinsic stress of this material is ten times lower than that
of stoichiometric nitride (Si N ) . Si N films result in
in a 10:1 flow ratio. The
isotropic etch of Si (light gray) in a 100:1 solution of HNO?:HF, which
does not attack silicon nitride (dark gray). The pictures show the cantilever
(a) before, (b) during, and (c) after the Si etch.
SEM pictures depicting the formation of the cantilevered tip by
curved cantilever beams, whereas our Si-rich nitride films
do not, as shown below. Note that the membrane is coated
with nitride on both sides and that it is strong enough to
withstand the next two photolithography steps on its front side.
First, a pattern of squares of photoresist (10
m 20 m for the 5- or 10- m-thick SOI wafers,
respectively) is defined on the membrane and transferred onto
the front-side nitride layer with a high-power (250 W) SF
plasma etch [Fig. 1(d)]. Back-side illumination and an infrared
camera are used to align this mask onto the membrane. The
squares will later serve as a mask for the formation of the
tip, but we proceed first to the cantilever formation step.
The shape of the cantilever is defined on the front side of
the membrane through a thick ( 6
(aligned onto the nitride square such that it is fully covered
by the photoresist with its edge
of the cantilever). A low-power (50 W) SF /CCl plasma etch
(10:1 SF :CCl flow-rate ratio) is used to clear the membrane
in the exposed Si areas [Fig. 1(e)]. The CCl gas serves to
selectively deposit a protective layer of carbonaceous polymer
on the side walls as the etch proceeds, reducing undercuts and
producing straight side walls . Thus, a bilayer (nitride/Si)
cantilever with a square nitride pattern is formed. A scanning
electron microscope (SEM) picture of the device at this point
of the process is shown in Fig. 2(a).
m) photoresist pattern
5 m away from the edge
C. Tip Formation
The tips are formed by wet etch in a solution of 100:1
HNO :HF solution, which etches Si ( 0.5
silicon nitride , . This etch isotropically undercuts a
nitride square to form a tip. Hence, the tip mask is designed
to be at least twice as large as the thickness of the starting
SOI layer. In a typical 5- m-deep Si etch, we measure a
m/min), but not
FOLCH et al.: MICROFABRICATION OF NITRIDE CANTILEVERS FOR MICROSCOPY305
Fig. 3.SEM picture of a tip (a) before and (b) after a set of two oxidation-sharpening steps (see text).
50-˚ A (negligible for our purposes) Si-rich nitride etch. An
SEM picture of an incomplete tip is shown in Fig. 2(b). When
the back-side nitride layer is reached, the front-side 10- m
10- m nitride square has been undercut to a degree such
that it is either supported by a thin ( 1
removed [Figs. 1(f) and 2(c)]. A finished tip typically has a
8- m 8- m base and is
starting thickness of the SOI layer. The etch rate ( 0.5–1.5
m/min) decreases with time, probably due to depletion of HF
in the solution ( 1 L), and is not uniform across the wafer.
Therefore, most of the finished tips on the wafer, such as the
one in Fig. 3(a), end in a submicron-sized platform or wedge
or in a broken stem, and they must be sharpened in order to
be suitable for AFM imaging.
To compensate for the nonuniformity of the etch, we de-
signed our mask so that chips contain three cantilevers, each
with slightly different (1
m) square sizes for the tip mask.
Thus, at least one cantilevered tip on each chip was viable for
imaging. We chose the wet etch for its simplicity and because
it serves to demonstrate the powerful idea that a Si tip can be
sharpened without affecting the nitride lever. Recently, Boisen
et al.  obtained high yields in the formation of Si tips of
various aspect ratios by reactive ion etch with SiO masks.
Our preliminary results, not shown here, similarly demonstrate
a 95% yield in the formation of Si tips by pure SF plasma
etch with nitride masks. This dry etch can straightforwardly
substitute the HNO :HF etch in our process to allow for better
control over the tip shape.
m) column or it is
1 m lower than the
D. Oxidation Sharpening
We have followed well-known recipes for oxidation sharp-
ening , . We usually grow
950 C and strip it in buffered HF. In accordance with Marcus
et al., we observe that if the wedge or platform at the apex
of a tip after the HNO :HF etch is wider than a certain value
( 500 nm in our case), then the apex develops into two or four
microtips, respectively, after oxidation sharpening. In our case,
this occurs for
50% of the tips due to the poor reliability and
nonuniformity of the HNO :HF etch process. However, all tips
with an apex narrower than 500 nm before the oxidation step(s)
500 nm of thermal oxide at
develop a single microtip and, after two oxidation-sharpening
steps, become indistinguishable from each other under SEM
inspection. Since the radius of curvature at the tip apex is in
most cases smaller than 10 nm, the resolution of our SEM does
not allow us to measure it directly. Instead, as shown below,
we tested the tips in AFM contact mode and demonstrated their
superior sharpness when compared to commercially available
ones. Each cycle of oxidation and dissolution results in a
decrease in tip size. Typically, tips made from wafers with
a 10- m-thick SOI layer are
6–7 m tall after the two
III. AFM TIP TESTING
We have successfully imaged a variety of samples with
AFM using our tips on 1- m-thick cantilevers ( 30-kHz res-
onance frequency) in contact mode. Thicker and stiffer ( 100
kHz) cantilevers must be used for operation in noncontact
In AFM, resolution is a sample-dependent concept: the
ability to resolve atomic spacings on graphite or mica, for
example, does not give information on the sharpness of the
tip. Well-defined artificial nanostructures, such as KOH-etched
100-nm-deep pyramidal nanopits on Si(100) (see Fig. 4), are
our preferred choice for evaluating lever performance and tip
sharpness simultaneously. Tip-sample convolution, manifested
in the form of features repeated over each nanostructure,
allows us to infer qualitative information on the tip shape.
Fig. 4(a) shows an AFM image of the pits (bottom) and an
SEM image of the commercial tip2(top) used to acquire the
AFM image. All the pyramids appear to have a flat inclined
top and end in a ridge rather than a vertex. By comparison, the
AFM image acquired with one of our oxidation-sharpened tips
[Fig. 4(b)] features sharply pointed pyramids and pyramids
with ridges oriented in either crystallographic direction, which
is to be expected if the square mask for the KOH etch is not
perfectly square . The same differences between the two tips
2We used a Dimension 3000 by Digital Instruments at the C.M.S.E. (MIT)
and the contact-mode tips (nominal force constant of 0.06 N/m) supplied with
it. The force constant of our cantilevers was computed from data in  and
well-known formulas for beam deflection.
306 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 6, NO. 4, DECEMBER 1997 Download full-text
?100-nm-deep and 150-nm-side pyramidal pits on Si. The AFM tip irregularities are attributed to wear during scanning. Imaging parameters were identical
for both images: 10–9N constant force, 4-Hz scan rate, and 600-nm scan range. For clarity, the AFM images are plotted in three-dimensional simulated
illumination and inverted color scale, which causes the pits to appear as pyramids. Our AFM tips clearly feature less convolution.
SEM images of (top) (a) a commercial tip and (b) one of our tips after acquiring the corresponding (bottom) AFM contact-mode images of
were observed in other sample spots. The tip in Fig. 4(b) was
not our sharpest, but one of the sharpest ones that survived
scanning with similar imaging conditions [10
force in both images (see the second footnote)]. Tips with a
smaller radius of curvature at the apex apply more pressure for
the same applied force and are thus more prone to fracture.
Note the debris or particles on both tips not present before
scanning and presumably due to wear during imaging.
In summary, we have described in detail a novel process for
microfabricating AFM low-stress silicon nitride levers with
oxidation-sharpened Si tips. The novelty of our process is
that both the sharpening procedure and the tip-formation step
leave the cantilevers intact. Alternatively, one could implement
the cantilever formation step at the end of the process to
increase the tip viability yield in combination with a dry tip
etch or to create Si tips on other low-stress materials. We
have favorably compared our tips with the most widely used
commercial tips during AFM imaging of demanding samples
such as nanofabricated pyramidal pits.
We thank P. Tierney and T. Tyson (M.T.L.) for making
the chrome masks, B. Alamariu (M.T.L.) for the low-stress
nitride depositions, and A. Franke (N.S.L., MIT) for making
the nanopits. We are grateful to L. Parameswaran, O. Hurtado,
and L. Cordes (M.T.L.) for their insight and fruitful comments.
 G. Binnig, C. F. Quate, and Ch. Gerber, Phys. Rev. Lett., vol. 56, no.
9, p. 930, 1986.
 D. Sarid, Scanning Force Microscopy.
 J. N. Israelachvili, Intermolecular and Surface Forces.
 T. R. Albrecht, S. Akamine, T. E. Carver, and C. F. Quate, J. Vac. Sci.
Technol., vol. A 8, no. 4, p. 3386, 1990.
 R. B. Marcus, T. S. Ravi, T. Gmitter, K. Chin, D. Liu, W. J. Orvis, R.
D. Ciarlo, C. E. Hunt, and J. Trujillo, Appl. Phys. Lett., vol. 56, no. 3,
p. 236, 1990.
 T. S. Ravi, R. B. Marcus, and D. Liu, J. Vac. Sci. Technol., vol. B 9,
no. 6, p. 2733, 1991.
 J. Itoh, Y. Tohma, S. Kanemaru, and K. Shimizu, J. Vac. Sci. Technol.,
vol. B 13, no. 2, p. 331, 1995.
 J. Brugger, R. A. Buser, and N. F. de Rooij, Sens. Actuators, vol. A
34, p. 193, 1992.
 K. E. Petersen, “Silicon as a mechanical material,” Proc. IEEE, vol. 70,
no. 5, p. 420, 1982.
 B. Alamariu, private communication.
 S. Wolf and R. N. Tauber, Silicon Processing for the VLSI Era (Volume
1: Process Technology). Sunset Beach, CA: Lattice, 1986.
 H. Robbins and B. Schwartz, J. Electrochem. Soc., vol. 106, no. 6, p.
, J. Electrochem. Soc., vol. 107, no. 2, p. 108, 1960.
 A. Boisen, O. Hansens, and S. Bouwstra, J. Micromech. Microeng., vol.
6, p. 58, 1996.
New York: Oxford Univ. Press,
Albert Folch, photograph and biography not available at the time of publi-
Mark S. Wrighton, photograph and biography not available at the time of
Martin A. Schmidt, photograph and biography not available at the time of