Fabrication of SiCN ceramic MEMS using injectablepolymer-precursor technique

Li-Anne Liew*, Wenge Zhang, Victor M. Bright, Linan An,Martin L. Dunn, Rishi Raj

Department of Mechanical Engineering, NSF Center for Advanced Manufacturing and Packaging of Microwave,

Optical, and Digital Electronics (CAMPMODE), University of Colorado, Boulder, CO 80309-427, USA

Abstract

In this paper, a novel and cost-effective technology for the fabrication of high-temperature MEMS based on injectable polymer-derived

ceramics is described. Micro-molds are fabricated out of SU-8 photoresist using standard UV-photolithographic processes. Liquid-phase

polymers are then cast into the molds and converted into monolithic, fully-dense ceramics by thermal decomposition. The resultant ceramic,

based on the amorphous alloys of silicon, carbon and nitrogen, possess excellent mechanical and physical properties for high-temperature

applications. This capability for micro-casting is demonstrated in the fabrication of simple single-layered, high aspect ratio SiCN

microstructures. A polymer-based bonding technique for creating more complex three-dimensional structures is also presented.

# 2001 Published by Elsevier Science B.V.

Keywords: MEMS; High-temperature; Polymer; Ceramic

1. Introduction

MEMS for extreme temperature environments have

attracted much attention due to their many potential appli-

cations, such as optical MEMS for high power laser applica-

tions [1] and microcombustors [2] for MEMS power

sources. However, the high temperature environments

remain a signi®cant challenge to current MEMS technology.

The fabrication of MEMS for high-temperature applications

is a two-fold problem: selecting suitable refractory materials

and developing appropriate microfabrication techniques.

Traditional microfabrication processes, such as surface

micromachining and LIGA, rely on polysilicon and plas-

tic/nickel as structural materials, respectively. These mate-

rials cannot operate at high temperatures for extended

periods of time. At the same time, state-of-the-art ceramics

that are designed for high-temperature environments cannot

be easily processed using existing microfabrication techni-

ques. Moreover, CVD of SiC [3], a technique currently

under development for high-temperature MEMS, is time-

consuming and expensive. Also, the planar nature of

CVD prevents the fabrication of complex three-dimensional

structures. Therefore, the development of new materials and

appropriate microfabrication techniques for high-tempera-

ture MEMS is of both scienti®c and practical interest to the

MEMS community. Fig. 1 compares the aspect ratios

achieved and maximum operating temperature of tradition-

ally-microfabricated MEMS with that of the polymer-

derived SiCN MEMS that are described in this paper.

2. Injectable polymer derived ceramics

A novel fabrication technique for high-temperature

MEMS using an injectable polymer-derived SiCN ceramic

has been developed. This new technology is based on the

recently developed polymer-derived ceramics, which are

bulk ceramics fabricated by the thermal decomposition of

compacted crosslinked polymer powders [4]. The polymer-

derived ceramics are amorphous alloys of silicon, carbon,

and nitrogen (SiCN) which remain thermally stable up to

15008C. The compositions of the new ceramics can be

varied through the use of different polymer precursors,

and can be tailored to produce SiCN with excellent thermal

and mechanical properties. Table 1 compares the physical

properties of SiCN with those of Si and SiC. Young's

modulus, Poisson's ratio, and density are in the same range

as those of SiC and Si. The creep resistance of SiCN is

Sensors and Actuators A 89 (2001) 64±70

* Corresponding author. Tel.: �1-303-492-3842;

fax: �1-303-492-3498; URL: http://mems.colorado.edu.

E-mail address: li.liew@colorado.edu (L.-A. Liew).

0924-4247/01/$ ± see front matter # 2001 Published by Elsevier Science B.V.

PII: S 0 9 2 4 - 4 2 4 7 ( 0 0 ) 0 0 5 4 5 - 8

comparable to that of SiC and Si3N4 [5,6], while its oxida-

tion resistance exceeds that of the same materials [7]. In

addition, the thermal shock resistance of SiCN appears very

promising for high-temperature applications.

SiCN may be obtained from liquid- or powder-based

polymer precursors. However, the SiCN obtained from

the powder-route shows relatively low strength and hardness

due to the high porosity of powder-derived ceramics in

general (typically �10 vol.% porosity). In addition, pow-

der-processing cannot be easily integrated into existing

microfabrication techniques. Therefore, liquid polymer pre-

cursors were used to develop a novel micro-casting techni-

que for the fabrication of SiCN MEMS structures. Fig. 2

compares the microstructure, strength, and hardness of SiCN

samples obtained through powder- and casting-routes.

Fig. 1. Diagram of microstructure aspect ratio against maximum use

temperature for different MEMS materials and processes.

Table 1

Comparison of physical properties of SiCN, Si, and SiC

SiCN Si SiC

Density (g/cm3) 2.20 2.33 3.17

E modulus (GPa) 158 163 405

Poisson's ratio 0.18 0.22 0.14

CTE � 10ÿ6 (K) 0.5 2.5 3.8

Hardness (GPa) 15 11.2 30

Strength (MPa) 250 175 418

Toughness (MPa m1/2) 3.5 0.9 4±6

Fig. 2. (a) Scanning electron micrographs showing the microstructure of polymer-derived SiCN from powder-route and (b) from casting-route. This figure

shows that the material resulting from casting is fully dense in that no defects can be seen despite a higher magnification than that in (a). (c) Comparison of

strength and hardness (which are not directly related to each other) of samples obtained from powder- and casting-routes.

L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70 65

3. Microfabrication of SiCN MEMS

Fig. 3 outlines the process ¯ow in the fabrication of SiCN

MEMS. First, a mold is fabricated using standard photo-

lithographic techniques. The liquid polymer precursor is

then cast into the mold, and the mold and polymer precursor

are then heated or `̀ thermal-set'' at �2508C to solidify the

polymer. After thermal-setting, the polymer becomes a

transparent solid, and may be separated from the mold if

suitable techniques are used. After thermal-setting, the

polymer part is crosslinked by heating to �4008C under

isostatic pressure. After crosslinking, the polymer becomes

infusible, remaining transparent. In the ®nal stage (pyroly-

sis), the crosslinked polymer part is heat-treated at�10008Cto convert it to a monolithic ceramic part.

The commercially-available Ceraset (CerasetTM, Lanxide

Company, USA) was used as the starting polymer. A catalyst

(Dicumyl peroxide, Aldrich Chemical, Milwaukee WI,

USA) is added to lower the thermosetting temperature to

1408C, so that it is compatible with photoresist processing.

The mold was made using SU-8 photoresist [8] from Micro-

Chem Corp. SU-8 is a well-known negative, epoxy-type

photoresist based on EPON SU-8 resin. Its key features are

its ability to produce high-aspect ratio structures and low

optical absorption in the near-UV range, which leads to

straight vertical sidewalls. Other advantages of using SU-8

as the mold are: (1) standard UV-photolithography can

produce high-aspect ratio structures; (2) batch fabrication;

(3) the SU-8 photoresist decomposes during pyrolysis, enabling the release of the structures using the `̀ lost mold

technique''.

Fig. 4 outlines the micro-casting technique in more detail.

First, the photoresist is spun onto a substrate (a). Then, the

photoresist is patterned using standard UV-lithography and

developed, producing cavities of desired shapes (b). The

liquid precursor (Ceraset) is then cast into the cavities by

spinning, and the wafer is thermal-set in an oven for 20 min.

At the end of this step, the Ceraset is solid and there is a thin

layer of it covering the entire wafer (c). This thin top layer of

Ceraset is polished-off (d). The wafer is then crosslinked

under isostatic pressure (e), and during pyrolysis the SU-8

decomposes and the SiCN part no longer adheres to the

substrate (f). The ®nal result is a free-standing high aspect-

ratio microstructure (g).

Of course, this schematic shows only one structure being

fabricated; in reality, multiple structures are fabricated at

once on a wafer, which is one of the attractions to micro-

casting using photoresist molds.

Fig. 5(a) shows an SU-8 mold for a micro-gear. Fig. 5(b)

shows the same mold when ®lled with the polymer pre-

cursor. For small cavities (less than 20 mm wide), often the

liquid does not fully ®ll the cavity due to the surface tension

of the liquid and/or air bubbles in the mold cavity. This can

be solved by placing the ®lled mold in vacuum to remove

most of the air bubbles, and then pressing a thin ®lm of

Te¯on over the wafer to force the liquid into the cavities.

This ®lm is left on the mold during thermosetting, and alsoFig. 3. General processing steps in the fabrication of injectable polymer-

derived ceramics.

Fig. 4. Fabrication process for SiCN MEMS.

66 L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70

serves to minimize the thickness of the extra top layer of

Ceraset that needs to be polished-off.

4. Crosslinking and pyrolysis

The key to the success of the casting-route is the applica-

tion of isostatic pressure during crosslinking. The heat-

treatment during crosslinking generates gaseous by-pro-

ducts. This out-gassing may cause the formation of micro-

cracks which disintegrate the sample [4]. In the casting

approach, the applied isostatic pressure provides an obstacle

to the nucleation of bubbles and microcracks. The gaseous

by-products would need to build up suf®cient pressure to

overcome the applied pressure before bubbles and micro-

cracks can nucleate. Due to the open structure of the

polymer, the gases will diffuse out of the sample before

they can build up a pressure that is high enough to overcome

the applied pressure. Thus, transparent and defect-free cross-

linked polymers can be obtained by the application of

isostatic pressure during crosslinking. The fact that no

defects can be seen in the ®nal ceramics indicates that

the infusible polymer network has enough strength to with-

stand the further heat-treatment during pyrolysis.

The crosslinked samples are then pyrolyzed at 10008C for

8 h to convert the polymer into a ceramic. Very low heating/

cooling rates are used: the heating rate is about 18C/min to

4008C and 258C/h to 10008C; the cooling rate is 18C/min to

room temperature. Higher heating rates result in the produc-

tion of most of the gaseous by-products at a narrow tem-

perature range (H2 at 600±7008C, and CH4 at 6008C), which

could generate defects in the pyrolyzed sample and degrade

the mechanical strength of the structures. Examples of

microstructures fabricated from this casting process are

shown in Fig. 6.

5. Demolding

Initially during pyrolysis, the SU-8 photoresist mostly

decomposed but left a thick, hard residue on the walls of the

samples, as can be seen in Fig. 6 and also in Fig. 7(a). These

SU-8 by-products, being chemically resistant to etchants and

strongly attached to the SiCN, could not be easily removed

through either chemical or physical means. Attempts to

burn-off the residue at 10008C in an oxygen atmosphere

were also not successful. The by-products are the result of

reactions between the liquid Ceraset and the polymer mold

at the interface, producing unknown compounds. Many

attempts were made to de-mold the parts before pyrolysis

in order to prevent the reactions.

One drawback of using SU-8 is that it is not easily

removed, even when using the solvent provided by the

manufacturer. It was found that the SU-8 could be stripped

if the solvent was heated to about 908C. However, before

pyrolysis, the Ceraset is highly reactive and chemically

Fig. 5. (a) Photoresist mold (200 mm diameter and 75 mm thick); (b) mold filled with Ceraset.

Fig. 6. (a) Gear (200 mm diameter and 45 mm thick) made from the mold shown in Fig. 8; (b) cantilever beam (137 mm thick); (c) tensile test sample (142 mm

thick).

L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70 67

reacts with many polymers. Various photoresists that could

be easily stripped were experimented with as alternatives to

SU-8, but the highly reactive nature of Ceraset still proved to

be the limiting factor. This problem was ®nally solved

through a combination of an extended bake and extended

UV-exposure of the SU-8, following the mold fabrication

but prior to casting. This ensured that the SU-8, being a

branched polymer, was completely crosslinked before the

Ceraset was cast into it. Results were encouraging. SU-8

thus remained the material of choice for mold fabrication,

and the `̀ lost mold'' technique was pursued following this

post-fabrication treatment of the mold prior to casting.

Fig. 7(b) shows the marked improvement in surface quality

following this treatment.

Despite the success of the post-fabrication treatment of

the mold in producing SiCN parts that were clean and

residue-free, it is still advantageous to de-mold prior to

pyrolysis, especially if polymer-based bonding (decribed

later in this paper) is to be used to fabricate multi-layer

structures.

6. Size limitations

We have studied the size range of samples that can be

successfully cast. It has been found that the samples will

crack at the crosslinking and pyrolysis stages if the length

exceeds a certain critical value, which is a function of the

sample thickness, being larger for thicker samples. Ideally,

there should not be any limit on the sample size, which is

proven by casting in macro-sized Te¯on molds that have

been conventionally machined. In this case, the thermally-

set samples are released from the Te¯on mold and cross-

linked and pyrolyzed as free-standing samples. The diameter

of these samples can be up to 10 mm (which is limited by the

size of our furnace). The reason for the observed scale limit

for the micro-cast samples is explained with the aid of Fig. 8.

The Ceraset experiences volume shrinkages of 5 and 25%

during crosslinking and pyrolysis, respectively. However, at

the same time the Ceraset adhesion to the silicon substrate

results in a tensile stress, s, in Ceraset structure. This tensile

stress in turn induces a shear stress, t, at the silicon/Ceraset

Fig. 7. (a) SiCN cantilever beam encased in SU-8 residue; (b) SiCN gear tooth (75 m thick), obtained from SU-8 mold that had undergone post-fabrication

treatment to prevent residue formation.

Fig. 8. Induced stresses during heat-treatment of Ceraset on silicon.

68 L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70

interface. Assuming a disk-shaped sample, the tensile and

shear stresses are related by

s � r

2ht

where s increases with the progress of heat-treatment. Let tc

and sc be the critical values for which the sample will peel-

off from the substrate and fracture, respectively. Then, if the

ratio r/h is too large (which corresponds to thin samples),

then sc will be reached first and sample will crack. If

the sample is thick, then tc will be reached first and

sample will peel-off from the substrate without cracking.

The problem can be prevented by using crosslinked Ceraset

as the substrate, because it will shrink together with the

microstructures.

7. Polymer-based bonding

Another advantage of using a liquid precursor is the

ability to use it as an adhesive layer to bond two or more

SiCN parts together. This `̀ polymer-based bonding'' tech-

nique thus may be used as a means of fabricating monolithic

multi-layer structures, whose ®nal thicknesses might not

otherwise be achievable due to the size limitations described

above.

The basic bonding process is shown schematically in

Fig. 9. First, two thermal-set or crosslinked solid parts

are fabricated by casting (a). The same liquid polymer is

then spread on the desired location in the same way that glue

is spread on components to be attached. The liquid nature of

the polymer precursor allows atomic level contact at the

interfaces between the adhesive layer and the solid polymer

parts. The thermal-set or crosslinked components are then

aligned and held together (b). The assembled structure is

then thermal-set again to solidify the adhesive layer, and

then crosslinked under isostatic pressure (c) to establish

chemical bonds between the polymer in the adhesive layer

and the polymer in the original components. Therefore, after

crosslinking, the whole structure becomes a solid piece of

polymer without a noticeable interface. This bonded struc-

ture is identical to a single cast piece. The sample is then

pyrolyzed (d). Further examination of parts produced by this

method indicated that there are no observable defects at the

bonding interface.

8. Conclusion

A novel micro-casting process based on injectable poly-

mer-derived SiCN has been developed. The resulting cera-

mic structures exhibit excellent mechanical and thermal

properties, making this process very promising for fabricat-

ing MEMS for high-temperature applications. SU-8 photo-

resist molds enable the use of low-cost microfabrication

facilities. A polymer-based bonding technique can be used

whereby SiCN parts are bonded together to produce com-

plex, three-dimensional monolithic ceramic components.

Compared to the existing MEMS materials and processes,

the new injectable polymer-precursor technique substan-

tially enhances the manufacturability of MEMS for high-

temperature applications.

Acknowledgements

This work is supported by the Defense Advanced

Research Projects Agency (DARPA) and U.S. Air Force

under contract # F30602-99-2-0543. Also, thanks to Mr.

Tsali Cross for developing the polymer-based bonding

technique.

References

[1] D.M. Burns, V.M. Bright, Optical power induced damage to

microelectromechanical mirrors, Sens. Actuators A 70 (1999) 6±14.

[2] C. Rossi, D. Esteve, C. Mingues, Pyrotechnic actuator: a new

generation of Si integrated actuator, Sens. Actuators A 74 (1999)

211±215.

[3] C. Jacob, S. Nishino, P. Pirouz, C.A. Zorman, A.J. Fleischman, M.

Mehregany, Structural and electrical characterization of single-crystal

4-inch CVD grown 3C-SiC films, Silicon Carbide Related Mater. 142

(1996) 429±432.

[4] R. Riedel, G. Passing, H. Schonfelder, R.J. Brook, Synthesis of dense

silicon-based ceramics at low temperature, Nature 355 (1992) 714±

716.

[5] L. An, R. Riedel, C. Konetachny, H.J. Kleebe, R. Raj, Newtonian

viscosity of amorphous silicon carbonitride at high temperature, J.

Am. Ceram. Soc. 81 (1998) 1349±1352.

Fig. 9. Illustration of process steps for polymer-based bonding of SiCN

parts: (a) two solid polymer parts; (b) bonding parts together with a layer

of liquid polymer; (c) crosslinked structure, and (d) ceramic multi-layer

component.

L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70 69

[6] R. Riedel, L.M. Ruwisch, L. An, R. Raj, Amorphous silicoboron

carbonitride ceramics with anomalously high resistance to creep, J.

Am. Ceram. Soc. 81 (1998) 3341±3344.

[7] R. Raj, L. An, S. Shah, R. Riedel, Oxidation kinetics of amorphous

silicon carbonitride ceramics, J. Am. Ceram. Soc., accepted for

publication, November 2000.

[8] H. Lorenz, M. Despont, N. Fahri, J. Brugger, P. Vettiger, P. Renaud,

High-aspect-ratio, ultrathick, negative-tone-near-UV-photoresist and

its applications for MEMS, Sens. Actuators A 64 (1998) 33±39.

Biographies

Li-Anne Liew received her BS and MS degrees in Mechanical Engineering

from the University of Colorado at Boulder in 1998 and 2000, respectively.

She is currently a doctoral research assistant in the Department of

Mechanical Engineering at the University of Colorado, Center for

Advanced Manufacturing and Packaging of Microwave, Optical and

Digital Electronics (CAMPMODE). Her research interests are in the

design and packaging of MEMS sensors, MEMS for biomedical

applications, and the design and fabrication of MEMS for high-

temperature environments.

Wenge Zhang received his BSME degree from Dalian University of

Technology in Dalian, China, in 1982 and his MS degree in Mechanical

Engineering from the University of Colorado, Boulder, in 1995. He joined

the University in 1991 and is currently a research associate for the Center

for Advanced Manufacturing and Packaging for Microwave, Optical and

Digital Electronics, at the University of Colorado. His research interests

include low-cost prototyping and thermal management of MCMs,

thermosonic flip-chip bonding and optoelectronics packaging. He has 14

years of research experience in mechanical design, computer control

systems and logic circuit design, and 5 years of research experience in

thermosonic flip-chip bonding, MCM substrate fabrication, and optoelec-

tronics packaging.

Dr. Victor M. Bright is an Associate Professor of Mechanical Engineering

and the Director of the MEMS R&D Laboratory, University of Colorado at

Boulder. Prior to joining the University, he was an Associate Professor and

the Director of the Microelectronics Research Laboratory in the

Department of Electrical and Computer Engineering, Air Force Institute

of Technology, Wright-Patterson Air Force Base, Ohio (June 1992±

December 1997). Prof. Bright's research includes MEMS, silicon

micromachining, microsensors, microactuators, MEMS self-assembly,

MEMS packaging, opto-electronics, and semiconductor device physics.

Dr. Bright received the following awards in the area of MEMS: Best

paper of the MCM'98 Ð International Conference and Exhibition on

Multichip Modules and High Density Packaging, 1998; R.F. Bunshah Best

Paper Award at the 1996 International Conference on Metallurgical

Coatings and Thin Films. Dr. Bright has authored and co-authored more

than 70 papers in the areas of MEMS. He is a member of IEEE, ASME,

and SPIE. He serves on the Executive Committee for ASME MEMS

Sub-Division.

Dr. Linan An obtained his PhD degree from Lehigh University in the field

of materials science and engineering. His research includes processing of

ceramic materials, microstructural design of ceramics, high temperature

behavior, oxidation and corrosion of ceramic materials, and mechanisms of

fracture in brittle solids.

Dr. Martin L. Dunn is an Associate Professor of Mechanical Engineering.

He received the PhD in Mechanical Engineering in 1992 from the

University of Washington. Prior to joining the faculty at the University of

Colorado, he was a postdoctoral appointee in a solid mechanics group at

Sanida National Laboratories, and a design engineer in a transducers group

at the Boeing commercial Airplane Company. Professor Dunn's research

interests include (i) the micromechanical behavior (fracture and deforma-

tion) of materials and structures for microelectromechanical systems

applications, (ii) the micromechanics and physics of heterogeneous media,

including defects, polycrystals, and composites, with emphasis on

piezoelectric solids, and (iii) the acoustic characterization of material

microstructure. His research is based on a strong coupling between

theoretical and experimental efforts. Professor Dunn has published over 60

articles in refereed journals on these subjects and has been the principal or

co-principal investigator on grants and contracts in these areas from NSF,

DOE, ONR, NIST, DARPA, and Sandia National Laboratories.

Dr. Rishi Raj prior to joining C.U. Boulder in 1996, had served on the

Faculty of Materials Science and Engineering at Cornell University for 21

years. He received the PhD in Applied Physics from Harvard University in

1970. His research is in processing of high-temperature metals,

composites, and ceramics. His most significant contributions have been

in understanding how interfaces control high-temperature processing and

mechanical behavior. He has over 200 publications of which 150 are in

refereed journals, and include 14 U.S. Patents. He has been a Guggenheim

Fellow, an Alenxander von Humboldt Senior Scientist Awardee, and a

Matthias Scholar at Los Alamos National Laboratory.

70 L.-A. Liew et al. / Sensors and Actuators A 89 (2001) 64±70