ABSTRACT. Thin-film silicon carbide is a promising new material for ... Poisson's ratio of 0.19 GPa were measured in this film ..... 44, no. 12, 3460-3466, 1964.
Mechanical Properties of 3C Thin-Film Silicon Carbide
1
2
3
1
Kamili M. Jackson , Guy F. Dirras , Richard L. Edwards , and William N. Sharpe, Jr. 1 The Johns Hopkins University, Department of Mechanical Engineering 3400 N. Charles St., Baltimore MD 21218 2 LPMTM-CNRS, Institut Galilee, University de Paris XIII Avenue J.B. Clément Villetaneuse, Fr. 93430 3 The Johns Hopkins University Applied Physics Laboratory 11100 Johns Hopkins Rd. Laurel, MD 20723
ABSTRACT Thin-film silicon carbide is a promising new material for microelectromechanical systems or (MEMS). In some ways it surpasses polysilicon by being able to resist high temperatures, radiation, and corrosive environments more effectively. Unfortunately, mechanical testing with this material has not been extensive. This research provides more data on the mechanical properties of 3C thin-film silicon carbide. Microsample tensile tests were performed on two materials to provide measurements of Young’s modulus, Poisson’s ratio, and strength. One material was a single crystal film with a thickness range of 0.5-1 µm. A Young’s modulus of 410 GPa, a strength of 1.1 GPa and a Poisson’s ratio of 0.19 GPa were measured in this film provided by Case Western Reserve University. The second material was a polycrystalline material provided by Massachusetts Institute of Technology. It had a thickness of 20-40 µm and showed a Young’s modulus of 430 GPa, a Poisson’s ratio of 0.24, and a strength of 0.5 GPa. For both materials the microstructure was observed and the elastic modulus was calculated in relation to the anisotropic elastic constants. These calculations agreed well with the measured results. INTRODUCTION The significant advancement of mechanical testing of thin films over the past ten years has enabled extensive mechanical testing of polysilicon, the dominant MEMS – material [1-5]. Test results include fracture strength, Young’s modulus, Poisson’s ratio, fatigue life, fracture toughness, temperature dependence and size dependence. As MEMS technology grows, the properties of polysilicon are becoming a limiting factor in some designs. In order to adapt to new challenges, new materials like silicon carbide are being considered. Thin-film silicon carbide is an attractive MEMS material because of it’s electrical and mechanical properties. Some advantages of silicon carbide over polysilicon include a larger band gap, stability at higher temperature, resistance to corrosive environments, and a higher stiffness. These characteristics make it an attractive material for high temperature sensors, micro-power applications and some biological uses. The potential of silicon carbide in MEMS is the subject of several articles [68]. A drawback to using thin-film silicon carbide in MEMS is
that knowledge of the mechanical properties is not readily available. The available studies are somewhat contradictory or do not account for microstructure [9-16]. However, in order to design MEMS effectively with silicon carbide, its mechanical properties must be well characterized. This research presents measured values for the elastic modulus, Poisson’s ratio, and strength of 3C thin-film silicon carbide. EXPERIMENTAL METHOD Specimens: To produce the data on mechanical properties of thin-film silicon carbide, microsample tension tests were performed on two types of thin-film silicon carbide. The materials were provided by Case Western Reserve University (CWRU) and Massachusetts Institute of Technology (MIT). Both providers used a one centimeter square specimen with a gage width of 600 µm and varying thickness. The specimen design can be seen in Figure 1. gripping ends
4mm
gage section
600 µm 10 mm
strain markers support strip
tensile direction
Figure 1. Schematic of the tensile specimen design Because silicon carbide is resistant to most chemicals, it is a difficult material to etch. To circumvent this problem, both MIT and CWRU used a molding technique. The material from CWRU was a single crystal film that was deposited epitaxially on a (100) oriented silicon substrate. However,
before the film is deposited, oxide was deposited to use as a mold. The oxide was then patterned into the negative pattern of the specimen. Deposition of the silicon carbide followed the oxide patterning. The silicon carbide was deposited using chemical vapor deposition with a combination of propane and silane. After the film was deposited, the wafer was planarized using chemical mechanical polishing. Metal markers for strain measurement were deposited using the lift-off technique. The wafers were then sent to the Applied Physics Laboratory (APL) where the specimens were released by bulk etching with KOH. The remaining oxide was removed with HF resulting in the finished specimen seen in Figure 2.
Figure 2. Tensile specimen from CWRU glued in the grips In contrast to CWRU, MIT used a wafer etched with DRIE for the mold of the specimens. The wafers were etched at MIT and then sent to HyperTherm for silicon carbide deposition by chemical vapor deposition. The resulting material was polycrystalline. The wafer, coated with silicon carbide, was then sent to Johns Hopkins University (JHU) where it was diced at APL, polished at JHU, and had gold strain markers deposited with the lift-off technique at APL. The specimens were then sent back to JHU where they were bulked etched using XeF2. XeF2 was chosen since it only needed a 10 µm thick photoresist as an etch-stop when it etched through the wafer. Thus the lack of oxide on the back of the wafer was not a concern. A final specimen viewed from the bottom of the specimen can be seen in Figure 3.
Figure 3. Finished tensile specimen from MIT. The notches in the side strips made cutting them easier Testing: To perform the tensile tests, the specimens were attached to grips with adhesive. A 24hr curing cement was used for the thicker specimens from MIT while an ultraviolet
curing adhesive was used for the CWRU specimens. Careful alignment was performed under a microscope before curing the adhesive. Before the tensile tests were conducted, the support strips were cut with a rotary tool using a diamond-coated blade. A piezoelectric actuator was used to displace the specimen. Forces were recorded with a load cell that had a maximum of 1lb for the CWRU specimens and 25 lbs for the thicker MIT specimens. Strain was measured directly on the specimen with the Interferometric Strain Displacement Gage (ISDG). Details of the ISDG method, developed by W.N Sharpe, Jr., are given in [17]. Metal markers were placed on the specimens as the last step of fabrication with the lift-off technique. The position of the markers on the specimens can be seen schematically in Figure 1. Two sets of two parallel lines were placed perpendicular to each other with a space between the parallel lines of 250 µm. Aiming a laser at the metal lines created two sets of fringes. The movement of the fringes was sensed by photodiode arrays. This data was then recorded and used to calculate strain. Because two sets of markers were placed perpendicular to each other, Poisson’s ratio can also be calculated. Thus, one test gave data for Young’s modulus, Poisson’s ratio, and strength. Thickness measurements: After testing, the thickness of the specimens was measured by standing them up in a microscope. The MIT specimens were thick enough to be seen well in a 400x optical microscope. The CWRU specimens were mounted vertically on blocks and measured in the scanning electron microscope. The specimens were coated with a gold palladium alloy to increase conductivity and thus image quality. RESULTS and DISCUSSION: Using the method outlined above, results for the CWRU and MIT material were obtained. More results were obtained for the CWRU material because fabrication was more streamlined while the MIT fabrication methods needed more development. In addition, overall there were more results for strength than Young’s modulus or Poisson’s ratio where strain measurement was required. Because the strain measurement was highly dependent on the quality of the metal markers, measuring strain while the fabrication processes were being refined was a challenge. Strength measurements only required a final load and a crosssectional area that were relatively easier to measure. Since all four strain markers are essential for Poisson’s ratio measurements, only a few values are reported here. CWRU material: The results of testing this silicon carbide gave one value of Poisson’s ratio of 0.19. As stated above, the quality of the metal lines determined the quality of the strain data. In this case the platinum used for the strain markers did not adhere well to the silicon carbide. It was noted that this value was lower than typical metal values for Poisson’s ratio. However, this was expected judging from bulk values of silicon carbide and general trends in metals and ceramics. Of course more tests need to be performed before confidence can be put in this value. Fourteen measurements of elastic modulus resulted in an average value of 410 GPa with a standard deviation of 45 GPa and a coefficient of variation of 11%. Before any comparisons can be made, the orientation of the tensile axis
of this single crystal material must be known. In this case the tensile axis was along the direction. The tensile orientation was determined by the orientation of the specimens on the substrate in relation to the wafer flats as seen in Figure 4 and confirmed by x-ray analysis.
One Specimen
The average strength was also lower than expected. Examinations of the types of flaws present were conducted to further explain the data. Analysis of the broken specimens led to several observations. First, it should be noted that there were two batches of material received. The fracture surfaces of the first batch were not examined. The second batch of material had a second material on some of the specimens that could not be removed by any method. A specimen with this material can be seen in Figure 6.
extra material
directions Figure 4. Wafer of specimens showing their orientation For single crystals the elastic modulus must be calculated to account for the anisotropy of the material. The Young’s modulus of a cubic material on the (100) plane in the direction was calculated in equation (1) in terms of the single crystal anisotropic elastic constants, S11, S12, and S44. E110=2/(S11+S12+S44/2)
(1)
The single crystal elastic constants have not been measured; however they have been modeled by various researchers using a variety of techniques [18-21]. Using these values and equation (1) a range of 419-517 GPa was tabulated for E110. It can be seen that the experimental results stated above agreed with this calculation. Thirty-five measurements of strength with an average of 1.10 GPa and a standard deviation of 0.5 GPa were made. The standard deviation was quite high which highlights the large scatter in results. The results of the Weibull analysis can be seen in Figure 5.
Figure 6. Specimen showing extra material that formed during processing It is believed that it was a result of a reaction of the silicon carbide and oxide mold. Because the material was as resistant to chemicals as silicon carbide, it may have been silicon or carbon rich silicon carbide. The specimens with a minimum of the extra material were tested. Through examination of the fracture surfaces, effects of this extra material were seen. Figure 7 shows a SEM photograph of the fracture surface of a specimen with a fracture strength of 0.95 GPa. Extraneous Material
The rougher surface is the extra material
Initiation flaw
The smoother surface is silicon carbide
1
0.9
0.8
Weibull Modulus = 2.5 Failure Probability
0.7
0.6
0.5
0.4
0.3
0.2
Figure 7. Fracure surface of a CWRU specimen
0.1
0 0
0.5
1
1.5
2
2.5
Strength GPa
Figure 5. Weibull distribution for the CWRU material Though scatter was expected for this ceramic, Weibull analysis showed a Weibull modulus of 2.5, which was on the low side for ceramic values.
It can be seen that the flaw began in the left corner of the extra material on the specimen. It can also be seen that the surface of the extra material is very rough which can serve as a source of flaws. Further scrutiny of the fracture surfaces also shows that the similar flaws can be seen in a specimen where the extra
material is not present. A micrograph of the fracture surface of a specimen with a fracture strength of 0.93 GPa is seen in Figure 8. This implies that there are other factors, perhaps in the fabrication, which may have decreased the fracture strength.
Figure 11. Fracture Type 2: Average Strength - 0.75 GPa, most of the specimen is intact. Initiation flaw
Figure 8. Fracture initiation in silicon carbide These figures show that though the extra material may have contributed to a lower fracture stress, the absence of this material did not guarantee a higher fracture stress. Other factors must be considered. A general examination of the fractures led to other observations. The specimens tended to crack along the directions. No cracks were seen o along the 100 direction which would be 45 from the tensile axis. An example of this behavior can be seen in Figure 9.
Figure 12. Fracture Type 3: Average Strength - 0.99 GPa, about half of the specimen is intact.
Tensile Direction
Figure 9. Fracture of CWRU specimen Another observation was that there were several types of fracture that each depended on a different fracture strength. These can be seen in Figures 10-13.
Figure 13. Fracture Type 4: Average Strength - 1.6 GPa most or the entire specimen is gone Fractures where more material disintegrated after fracture had higher fracture strengths. In contrast, lower fracture strengths left more material intact after fracture. This phenomenon may be due to more elastic strain energy in the specimen at higher strengths. The trend also suggests that flaws causing failure at lower strengths were large distinct flaws that traveled quickly across the specimen and did not branch.
Figure 10. Fracture Type 1: Average Strength - 0.51 GPa, all of the specimen is intact.
MIT material: There were fewer results for the MIT material because of the extensive individualized fabrication that was done for each specimen. The tests conducted resulted in two values for Poisson’s ratio of 0.21 and 0.26. Like the
CWRU material, it was seen that the Poisson’s ratio was lower than that of metals. Unlike the CWRU specimens, the gold markers used as strain markers adhered very well to the silicon carbide. However, a residue covered the gold markers after etching that could not be removed. It is now believed that this was a flourocarbon that can be removed by plasma etching. Four results for elastic modulus gave an average of 430 GPa. In order fully evaluate this result and compare it to other data, the microstructure must be considered. Examination of the grains showed a columnar grain structure with a grain size of 200nm. Micrographs of the grains can be seen in Figures 14 and 15.
that there were other orientations present but over 95% of the grains had a orientation. Using the single crystal constants the elastic modulus on the (111) plane was calculated as E111=2/(S11+S12+S44/2)
(2)
The Young’s modulus was the same in every direction on the (111) plane and was coincidentally the same as the equation for the CWRU specimens. Using the available single crystal elastic constants, the range of E111 was tabulated as 419-517 GPa. This range compared well with the measured values. Eleven strength measurements were made which resulted in an average value of 0.5 GPa with a standard deviation of 0.2 GPa. As with the CWRU results, there was a lot of scatter in the results. This can be seen from the Weibull analysis that resulted in a Weibull modulus of 2.3 and the Weibull distribution plotted below in Figure 17. 1
Weibull Modulus = 2.3
0.9
Figure 14. Cross-sectional view of columnar grains in the MIT silicon carbide
Probability of Failure
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Strength GPa
Figure 17. Weibull Distribution for the MIT material The lower strength of the MIT material could be due to a number of factors including grain boundaries, a larger crosssectional area, and side roughness due to the DRIE mold. A SEM micrograph of the side walls is shown in Figure 18. Figure 15. Plan view showing 200nm grains in MIT material The results of the x-ray analysis can be seen in Figure 16.
XRD Texture Results
intensity
Figure 18. Sidewall of an MIT tensile specimen angle
Figure 16. X-ray Diffraction results of the MIT material The analysis showed that the grains have a random in-plane texture with a out-of-plane texture. It can be seen
CONCLUSON: Results from tensile testing and microstructural studies of two types of silicon carbide have been reported. Elastic modulus data of 410 GPa for the single crystal material and 430 GPa for the polycrystalline material agreed well with calculated values. The strength results showed an average
of 1.1 GPa for the single crystal material and 0.5 GPa for the polycrystalline material. Differences in the strength may have been due to specimen size, presence of grains, initial flaws, and fabrication effects. More examinations of fracture surfaces are needed to explain these results further. Poisson’s ratio measurements were very few and resulted in values of 0.19 for the single crystal material and 0.24 for the polycrystalline material. Though they seem to be in the correct range more data needs to be collected to verify these values. ACKNOWLEDGEMENTS: Special thanks must be given to Chris Zorman and his colleagues at CWRU as well as Mark Spearing and his colleagues at MIT for supplying the materials. This research was conducted under DARPA grant number F 30602-99-20553. REFERENCES: 1. Ding, J.N., Meng, Y.G., and Wen, S.Z., “Size Effect on the Mechanical Properties of Microfabricated Polysilicon Thin Films”, Journal of Materials Research, vol.16, no.8, p. 22232228, 2001.
2. Kahn, H., Tayebi, N., Ballarini, R., Mullen, R.L., Heuer,
10. Serre, C., Perez-Rodriguez, A., Romano-Rodriguez, A., Morante, J., Esteve, J., and Acero, M., “Test Microstructures for Measurement of SiC Thin Film Mechanical Properties” Journal of Micromechanics and Microengineering, vol. 9, 1999. 11. Li, X. and Bhushan, B., “Micro/nanomechanical Characterization of Ceramic Films for Microdevices”, Thin Solid Films, vol. 340, 1999. 12. Sriram, S. and Bhushan, B., “Micro/nanotribogical Studies of Polysilicon and SiC Films for MEMS Applications,” Wear, vol. 217, 251-261, 1998. 13. Zorman, C., Roy, S., Wu, C., Fleishman, A., and Mehregany, M., “Characterization of Polycrystalline Silicon Carbide Films Grown by Atmospheric Pressure Chemical Vapor Deposition on Polycrystalline Silicon,” Journal of Materials Research, vol. 13, no. 2, 406-412, 1998. 14. Mehregany, M., Tong, L., Matus, L., and Larkin, D., “Internal Stress and Elastic Modulus Measurements on Micromachined 3C-SiC Thin Films,” IEEE Transactions on Electron Devices, vol. 44, no. 1, 1997.
A.H., “Wafer-level Strength and Fracture Toughness Testing of Surface-Micromachined MEMS Devices”, Materials Research Society Symposium Proceedings, vol.605, 25-30, 2000.
15. Su, C. M., Wuttig, M., Fekade, A., and Spencer, M., “Elastic and Anelastic Properties of Chemical Vapor Deposited Epitaxial 3C-SiC”, Journal of Applied Physics, vol. 77, no. 11, 5611-5615, 1995.
3. Bagdahn, J, and Sharpe, W.N., “Reliability of Polysilicon th Under Long-Term Cyclic Loading,” 15 IEEE International Conference on MEMS, 447-450, 2002.
16. Tong, L., Mehregany, M., Matus, L., “Mechanical Properties of 3C Silicon Carbide” Applied Physics Letters, vol. 60, no. 24, 2992-2994, 1992.
4. Sharpe, W.N., Jr., Jackson, K.M., Hemker, K.J., and Xie, Z., “Effect of Specimen Size on Young’s Modulus and Fracture Strength of Polysilicon”, Journal of Microelectromechanical Systems,” vol. 10, no. 3, 317-326, 2001.
17. Sharpe, W.N., “A Potential Optical Standard for Resistance Strain Gages,” Journal of Testing and Evaluation, vol.26, no.5 p. 481-8.
5. Sharpe, W.N., Jr., Yuan, B., Vaidyanathan, R., and Edwards, R.L., “Measurements of Young's Modulus, th Poisson's Ratio, and Tensile Strength of Polysilicon”, 10 IEEE Conference on MEMS, 424-9, 1997.
18. Lambrecht, W.R.L., Segall, B., Methfessel, M., van Schilfgaarde, M., “Calculated Elastic Constants and Deformation Potentials of Cubic SiC,” Physical Review B, vol. 44, 3685-3694, 1991.
6. Mehregany, M., Zorman, C., Rajan, N., and Wu, C., “Silicon Carbide MEMS for Harsh Environments,” Proceedings of the IEEE, vol. 86, no. 8, 1594-1610, 1998.
19. Karch, K., Pavone,P., Windl, W., Strauch, D., and Bechstedt,F., “Ab Inito Calculation of Structural, Lattice Dynamical, and Thermal Properties of Cubic Silicon Carbide,” International Journal of Quantum Chemistry, vol. 56, 801-817, 1995.
7. Davis, R., “Silicon Carbide: The Premier Paradigm for Structural and Microelectronic Device Applications in Severe Environments,” Materials Research Society Symposium Proceedings, vol. 410, 365-376, 1996.
20.Tolpygo, K. B., “Optical, Elastic, and Piezoelectric Properties of Ionic and Valence Crystals with the ZnS Type Lattice,” Soviet Physics Solid State, vol 2, no. 10, 23672376, 1961.
8. Muller, G., Krotz, G., and Niemann, E., “SiC for Sensors and High Temperature Electronics,” Sensors and Actuators A, vol. 43, 259-268.
21. Slack, G. A., “Thermal Conductivity of Pure and Impure Silicon, Silicon Carbide, and Diamond,” Journal of Applied Physics, vol. 44, no. 12, 3460-3466, 1964.
9. Roy, S., Zorman, C., and Mehregany, M., “The Mechanical Properties of Polycrystalline Silicon Carbide Films Determined Using Bulk Micromachined Diaphragms”, Materials Research Society Symposium Proceedings, vol. 657, EE9.5.1-EE9.5.6, 2001.
