- Email: [email protected]
Development of a process modeling for residual stress assessment of multilayer thin film structure
Thin Solid Films 584 (2015) 146–153
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Development of a process modeling for residual stress assessment of multilayer thin film structure H.J. Wang, H.A. Deng, S.Y. Chiang, Y.F. Su, K.N. Chiang ⁎ Advanced Micro-system Packaging and Nano-Mechanics Research Lab. Dept. of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan 300
a r t i c l e
i n f o
Available online 19 January 2015 Keywords: MEMS devices Process modeling Stoney's equation Residual stress Intrinsic stress Aluminum thin film
a b s t r a c t The mechanical behavior of Microelectromechanical systems (MEMS) devices is significantly influenced by fabrication processes conditions, deposition steps and geometry, e.g., processing temperature, thickness of thin film layers, deposition rate, etc. These processes under different deposition temperature will generate residual stress/strain in thin film layers, which will affect device sensing stability, accuracy and robustness, therefore, predicting and minimizing the residual stress become a critical issue on thin film devices design. This research develops a simulation methodology using finite element analysis (FEA) with process modeling technology to analyze the thermal stress/strain of thin films varying on different process procedures in a selected MEMS microphone device. In general, the residual stress/strain of multi-layers thin film is combined of intrinsic and thermal stresses, the thermal induced mechanical stress can be obtained using FEA but intrinsic stress which contains many uncertainties is very difficult to be defined. The residual stress in thin film layer on each processing procedure can be obtained from Stoney's equation. A comparison of the experiment and simulation results showed that the combined thermal induced stress and dislocation induced intrinsic stress in aluminum thin film will be rearranged after first annealing. The intrinsic stress, however, will affect the final residual stress when thickness of aluminum film is under 1 μm. The residual stress and the warpage of MEMS microphone are predicted by using processing modeling technology. The polysilicon will warp downward if the diaphragm is subjected to compressive stress. However, the polysilicon film will have much less warpage when intrinsic stress is positive. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Microelectromechanical systems (MEMS) microphone (Fig. 1) has become one of the most popular technologies for electronic devices because of its small size and high integration. Although the fabrication process of MEMS is compatible with semiconductor, the residual stress and undesirable pre-deformation of thin films during deposition and etching process will lead to reliability and yield problems. In the past few decades, numerous researches have been widely discussed in MEMS technology, but few of them focused on process simulation. This study developed a methodology using finite element analysis (FEA) with process modeling technology to analyze the residual stress in a structure of capacitive MEMS microphone. The residual stress of thin film is composed of thermal and intrinsic stresses [1]. The intrinsic stress, which is caused by deposition temperature, lattice mismatch, recrystallization and chemical reactions, etc., and is a process sensitive property, should be obtained from experiments. Extrinsic stress is applied to films mainly caused by coefficient ⁎ Corresponding author. Tel.: +886 3 5742925; fax: +886 3 5745377. E-mail address: [email protected] (K.N. Chiang).
http://dx.doi.org/10.1016/j.tsf.2015.01.014 0040-6090/© 2015 Elsevier B.V. All rights reserved.
of thermal expansion (CTE) mismatch and can also be called thermal stress which can be predicted using process modeling technique by element birth and element death in ANSYS®. In addition, extrinsic stress would also be caused by wafer chucking and wafer flatness
Fig. 1. The sketch of MEMS microphone.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
147
Fig. 2. Cross-section of 1P4M MEMS microphone.
analyzes the residual stresses of polysilicon diaphragm and aluminum backplate through experiment and numerical analysis. Polysilicon, the material of diaphragm, is often used in MEMS structure because it can be fabricated by thin film deposition process. For MEMS microphone, the polysilicon diaphragm would bend after etching process if it is subjected to compressive stress. This would make resonant frequency and microphone sensitivity far from the original design. The residual stress can be controlled by changing deposition and annealing temperatures [9–12]. Setting tensile film stress after etching process will let the microphone have better performance. The objective of this work is to obtain the intrinsic stress of thin film and discuss the intrinsic effect on a microphone structure. 2. Theory and experiment
Fig. 3. The process flow of the multilayered structure.
variation [2]. The average residual stress of multilayered structure is determined from Stoney’s experiment which measures the curvature on blanket wafer after each process. The experiment results indicated that the residual stress of thin films reduced and became stable after several thermal cycles. The residual stress can be measured by wafer curvature measurements and then the thermal stress subtracted from it to estimate the intrinsic component. Several authors have studied intrinsic stress in thin film [3–7], but few [8] discussed intrinsic effects in simulation. For capacitive type MEMS microphone, the component consists of a flexible diaphragm and a rigid backplate with acoustic holes. These two electrode plates are separated by an air gap. This study
The first theoretical formula for the evaluation of stresses in a thin film on a thick substrate was developed by Stoney [13]. The effective residual stresses of aluminum and polysilicon thin films on silicon wafers have been determined by Stoney’s equation, shown as below 2
σf ¼
Es t s 6ð1−νs Þt f
1 1 − R R0
ð1Þ
where σf is thin film stress; E s and ts are Young’s modulus and thickness of substrate, respectively; ν s is the Poisson’s ratio; tf is the film thickness; R0 and R is the radius of curvature on the substrate before and after film deposition. It assumes that ts N N tf, the lateral dimensions of the film and substrate are significantly greater than the thickness, and the substrate is elastically isotropic. The above formula does not involve the material properties of the film, and the film stress was obtained by measuring the curvature before and after film deposit.
Table 1 Processing steps and fabrication parameters of the multilayered structure. Procedure
1
2
3
4
5
Film Process Process temperature(°C) Film thickness (μm) Structure
Field oxide (FOX) Atmospheric pressure CVD 980 0.39
Poly-Si Low pressure CVD 620 0.2
ILD PE TEOS 400 0.85
IMD PE CVD 400 1.0
Al (TiN-AlSi1%Cu0.5%-TiN) Sputter 25-160 0.495
Number of thermal cycle Peak temperature(°C)
2 1,000
2 1,100
1 450
1 450
3 450
148
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
Table 2 Averaged residual stresses (MPa) of multilayer structure with 8 Samples.
Oxide Polysilicon ILD IMD Al
Before annealing
After 1st Annealing
After 2nd Annealing
11.08 −11.87 −87.54 −97.79 125.83
10.63 −2.51 −82.23 −23.39 439.3
6.00 −2.71 ----
Fig. 6. Stress vs. temperature in a multilayered structure after the deposition of the Al film.
Fig. 4. 1 μm plasma-enhanced chemical vapor deposition (PECVD) oxide with 450 °C thermal cycle.
For multilayer thin film structure, the effective stress (σeff) of whole films on substrate and the new stress (σΝ) of top layer thin film can be obtained by Eqs. (2) and (3), shown as below: σ eff ¼
σN ¼
2 Es t s 1 1 − XN 6ð1−νs Þ i¼1 t i R R0
σ eff
XN
t− i¼1 i
XN−1 j¼1
σ jt j
tN
ð2Þ
ð3Þ
where N, i, and j are layer number, and tN, ti and tj are thickness of Nth , ith, and jth layer individually. σj is stress of jth layer. A chip size of 1.5 mm × 1.5 mm with 450 μm thickness MEMS microphone (Fig. 2) using 1P4M (one polysilicon andfour metal layers) process was designed by ITRI. The first diaphragm metal layer is aluminum, and the second metal layer is a backplate. The processing flow,
Fig. 5. PECVD oxide with 500 °C thermal hysteresis loop [14].
thickness, deposition and annealing temperatures of multilayered structure are shown in Fig. 3 and Table 1. These films were deposited on (100) silicon wafer, and the curvatures of the substrate were measured using TENCOR FLX-2320 after each processing step. The samples were annealed under 1,000 °C (FOX) or 1,100 °C (Polysilicon) for 2 hours one to three times after film depositions to stabilize and minimize the influence on the stress in the next deposition step [14]. Different deposition and annealing temperatures result in different stress behavior in the polysilicon film [10–12], ranging from compressive to tensile stress. As the polysilicon film is annealed under 1,100 °C, the residual stress is reduced or almost no stress [15]. It is well known that reducing the stress on film can improve the sensing sensitivity of MEMS microphones. 3. Experiment results and discussion The films were deposited on 6" (100) silicon wafer. The samples, after each deposition step, will go through one to three times of annealing process to stabilize and minimize the residual stresses. The curvature of the substrate was measured at the end of each processing step. The stress history can be obtained by tracking the curvature. We assumed that the stresses of the prior films will not vary in the next deposition process since the post annealing has been applied to all steps to stabilize the residual stress. The average residual stresses in the multilayered structure before and after post annealing are presented in Table 2.
Fig. 7. Stress vs. temperature in the 0.49 μm Al film.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
149
Fig. 8. Schematic diagram of 2D FEA model before etching.
Fig. 9. Schematic diagram of 2D FEA model after etching.
Table 2 shows the residual stresses of each layer. The annealing process will reduce the residual stress of each layer except the Al film. The upper bound annealing temperatures is set as 1,100 °C, it is due to the in-situ stresses of field oxide and polysilicon cannot be measured by our experiment facility above that temperature. Eight samples have been measured in this research. The stress as function of temperature during the 450 °C thermal cycle in IMD layer is shown in Fig. 4. If the material is linear elastic, the residual stress will return to the original value as temperature cools down to room temperature. The stress hysteresis means that the intrinsic stress changed after the thermal cycling. Chen et al. [14] indicated that there is almost no hysteresis in the second cycle, as shown in Fig. 5. Therefore, after annealing process the stress in each layer will become stable and will not be altered in the next deposition processes. Tensile thermal stress is produced during cooling process in aluminum thin film because the CTE of aluminum is greater than silicon substrate. Although the film was sputtered at room temperature, it will be heated under high energy ion bombardment. The
temperature is 80 °C (process power, 2 kW) for only aluminum film and 160 °C (process power, 5 kW) for titanium nitride (TiN) sputtering close to it. The deposition of TiN as barrier layer can
Table 3 Material properties of microphone films.
Silicon (100) Field oxide Polysilicon ILD IMD Al[58]
Young’s Modulus (GPa)
Poisson’s Ratio
Coefficient of Thermal Expansion (ppm/°C)
129 94 169 120 85 69.7-0.037 × T
0.28 0.25 0.22 0.2 0.25 0.33
2.6 0.5 2.3 2 2 23.3 + 1.7 × 10−2 × T
Deposition Temp. (°C)
980 620 400 400 160
Fig. 10. The nonlinear material property of aluminum.
Table 4 Temperature dependent yielding stress of aluminum thin film [8]. Temperature(°C) yielding stress (MPa)
25 195
150 160
200 140
300 60
380 10
475 10
150
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153 Table 5 Etching radius of each layer. Etching radius (μm) Silicon Field oxide Polysilicon ILD IMD Al Oxide SiN
410 419 -419 422 -419 419
avoid hillock in aluminum film [16]. Therefore, 250 Å and 700 Å TiN films will be deposited before and after the aluminum film, respectively. Hysteresis stresses history of three thermal cycles is
presented in Fig. 6. It can be seen that the film is performed as linear elastic in the beginning of the second and third cycles and then entered plastic zone after 200 °C, and it indicated that the hysteresis stress decreased at the third cycle. A single Al film deposit on an oxide blanket wafer (Fig. 7) was made to see the multilayers effect (Fig. 6) on residual stress. The curve shown, Al film will enter the plastic zone when temperature reaches 200 °C. The residual stress of aluminum film was stabilized after the second cycle. The residual increased from 26 MPa to 356 MPa for single layer Al film, and from 125 MPa to 439 MPa for deposition of Al film on multilayered structure. It is because that most of the residual stress came from the non-recoverable plastic deformation. Although there is a little difference between Figs. 6 and 7, we can still see a similar thermo-mechanical behavior of Al thin film.
Fig. 11. Y-displacement of 2D FEA model after etching.
Fig. 12. (a) von-Mises stress; (b) Y-displacement of the Al backplate along the radius within 420 μm before and after etching.
Fig. 13. (a) X-stress; (b) Y-displacement of the polysilicon diaphragm along the radius within 420 μm before and after etching.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
151
Table 6 Intrinsic stress of a multilayered structure.
Residual stress (MPa) Thermal stress (MPa) Intrinsic stress (MPa)
Field oxide
Polysilicon
ILD
IMD
Al
6.0 −250.95 256.95
−2.71 −38.18 35.47
−82.23 −33.39 −48.84
−23.39 −25.13 1.74
439 196.87 242.13
Fig. 16. Simulation results of average stress vs. temperature for aluminum film from 0.2 μm to 2.0 μm.
Fig. 14. Thermal cycle of thermal stress and stress with 242 MPa intrinsic stress of the Al film deposited at 160 °C.
4. Numerical analysis FEA is widely adopted by MEMS researches, such as for design and assessment of residual stress [17]. In this research, process modeling technology is carried out to simulate deposition and etching processes on films. Each material on deposition temperature is assumed at stress-free condition, as the temperature cools down to room temperature, for a simple bilayer material without considering the bending effect, the thermal stress can be expressed through the following equation: σf ¼
E 1−ν
α f −α s ðT 1 −T 0 Þ
ð4Þ
f
where σf is the film stress; αf and αs are the coefficient of thermal expansion of film and substrate respectively; T 1 is the deposition temperature; and T 0 is the room temperature. However, for a 3D structure such as wafer and MEMS devices, FEA has to be applied to get more accurate thermal stress. Figs. 8 and 9 show the schematics of two-dimensional model of a microphone before and after the etching process (radius of 419 μm), respectively. A two-dimensional axial symmetry model was used to
represent round shape multilayer films. The material properties are presented in Table 3, and the nonlinear material property of aluminum is shown in Fig. 10. Temperature dependent yielding stress of aluminum thin film [8] is shown in Table 4. The coefficient of thermal expansion of Al film is greater than silicon substrate, the stress of aluminum film changed to tensile stress when the temperature cooled down from deposition temperature, the compressive stress is observed in other layers. The etching radius is shown at Table 5. Fig. 11 shows the displacement in Y direction after the etching of the oxide films. Figs. 12 and 13 show the von-Mises/X stress and Y displacements before and after the etching process, respectively. The stresses of the films before the etching are mainly caused by CTE mismatch between films and substrate, and partially relaxed by the oxide films’ etching process. The aluminum backplate is under tensile stress, so its Y-displacement is small and not easy to deflect, which is good for MEMS microphone. The X direction stress of the polysilicon diaphragm changes from −36.67 MPa to almost stress free. The intrinsic stresses effect is also examined in this paper. Table 6 shows the residual stresses, intrinsic stress and thermal stress in multilayered structure after annealing. The residual stress is the sum of intrinsic stress and thermal stress. For complementary metal– oxide–semiconductor (CMOS) MEMS structure, the residual stress will influence its mechanical behavior. The intrinsic stress may drastically change the mechanical behavior, and should be included into FEA. The intrinsic effects on simulation, one can add stress to linear-elastic materials directly, but not for plastic materials. For plastic material like aluminum, Fig. 14 shows the thermal stress at various temperatures under thermal cycle loadings and the test sample, aluminum film, with 242 MPa intrinsic stress was deposited at 160 °C. The schematic diaphragm shows that these stresses will reach yielding strength when temperature heating to
Fig. 15. Schematic of the aluminum films with lattice mismatch effect.
152
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
shown in Fig. 17. These experimental data are consistent with the simulation results. Fig. 18 shows Y-displacement without and with intrinsic stress applied to the polysilicon diaphragm after etching process. The warpage of polysilicon diaphragm with 35.47 MPa intrinsic stress is −2.61 μm, which is smaller than the no-intrinsic stress case, the film becomes flat when the intrinsic stresses increase to 50 MPa and 75 MPa. From above simulation results, it is clear that designer can control the annealing and deposition temperatures to make diaphragm flat. 5. Conclusion
Fig. 17. Simulation and experimental results of average stress vs. temperature for 0.495 μm aluminum film.
175 °C. Because of the strain hardening effect in plastic zone, the stress of Al film with 242 MPa intrinsic stress increases by only 4.7 MPa when the temperature cooling back to room temperature. Therefore, the intrinsic stress effect of Al film can be ignored and is not considered in this research. Doerner et al. [18] indicated that yielding strength increases as the aluminum film thickness decreases. Nejhad et al. [19] assumed that the intrinsic stress are uniform distributed through the entire film. Thus one can adapt residual stresses by changing the deposition temperature. With these concepts, this research deposited the aluminum film by two steps. The first layer was deposited at 500 °C and set a thin layer of aluminum film about 0.2 μm to compensate the lattice mismatch effect and increase the interfacial strength to avoid crack. The second layer of aluminum thin film can be regarded as a bulk material without lattice mismatch. The schematic diaphragm is shown as Fig. 15. The average stresses in our simulation can be shown as function of temperature when the entire thickness of aluminum film from 0.2 μm to 2.0 μm (Fig. 16). At room temperature, the residual stress on wafer reduces as the increasing film thickness. It had good agreement with Senez’s [20] results. Comparing with the experimental curve, the calculated average stress on the aluminum thin film is
Fig. 18. Y-displacement without and with intrinsic stress applied to the polysilicon diaphragm.
The mechanical behavior of thin film is affected by different fabrication procedures such as depositing, annealing temperature and etching processes. These processes will induce different range of residual stress on thin film and affect the device performance. A FEA using process modeling technology is successfully developed in this research for thin film design, the simulation result is also validated with experimental data. The residual stresses, intrinsic stress and thermal stress in multilayered structure after annealing is obtained in this work. The residual stress is the sum of intrinsic stress and thermal stress. The intrinsic stress may drastically change the mechanical behavior, and should be included into FEA. However, for Al film, intrinsic effect is small and can be ignored. The aluminum backplate designed in this research is under tensile stress, so its Y-displacement is small and not easy to deflect, which is good for MEMS microphone. Acknowledgment The authors would like to thank the Advanced Micro-system Packaging and Nano-Mechanics Research Laboratory, ITRI, and National Science Council of Taiwan (NSC102-2221-E-007-038-MY3) for supporting this research, and also would like to thank Professor S. Liang, Georgia Institute of Technology, for his fruitful discussion on the residual stress behavior of Al material after high temperature annealing. References [1] T.L. Chou, S.Y. Yang, K.N. Chiang, Overview and applicability of residual stress estimation of film-substrate structure, Thin Solid Films 519 (2011) 7883. [2] J. Gong, P. Vukkadala, J. Sinha, K.T. Turner, Determining local residual stresses from high resolution wafer geometry measurements, J. Vac. Sci. Technol. B 31 (2013) 051205. [3] W. Buckel, Internal Stresses, J. Vac. Sci. Technol. 6 (1969) 606. [4] M.F. Doerner, W.D. Nix, Stresses and deformation processes in thin films on substrates, Crit. Rev. Solid State Mater. Sci. 14 (1988) 225. [5] K.S. Chen, Techniques in residual stress measurement for MEMS and their applications, in: C.T. Leondes (Ed.), MEMS/NEMS, Springer, US, 2006, p. 1252. [6] P.B. Ghate, L.H. Hall, Internal stresses in multilayered structures, J. Electrochem. Soc. 119 (1972) 491. [7] M. Malakouti, M. Ameri, P. Malekzadeh, Dynamic viscoelastic incremental-layerwise finite element method for multilayeredstructure analysis based on the relaxation approach, J. Mech. 30 (2014) 593. [8] J. Lee, A.S. Mack, Finite element simulation of a stress history during the manufacturing process of thin film stacks in VLSI structures, IEEE Trans. Semicond. Manuf. 11 (1998) 458. [9] A. Torkkeli, O. Rusanen, J. Saarilahti, H. Seppa, H. Sipola, J. Hietanen, Capacitive microphone with low-stress polysilicon membrane and high-stress polysilicon backplate, Sens. Actuators A85 (2000) 116. [10] J. Miao, R. Lin, L. Chen, Q. Zou, S.Y. Lim, S.H. Seah, Design considerations in micromachined silicon microphones, Microelectron. J. 33 (2002) 21. [11] X. Zhang, T.Y. Zhang, M. Wong, Y. Zohar, Rapid thermal annealing of polysilicon thin films, J. Microelectromech. Syst. 7 (1998) 356. [12] H. Guckel, D.W. Buurn, C.C.G. Visser, H.A.C. Tilmans, D. Deroo, Fine-grained polysilicon films with built-in tensile strain, IEEE Trans. Electron Devices 35 (1988) 800. [13] G.G. Stoney, The tension of metallic films deposited by electrolysis, Proc. R. Soc. London, Ser. A82 (1909) 172. [14] K.S. Chen, X. Zhang, S.Y. Lin, Intrinsic stress generation and relaxation of plasmaenhanced chemical vapor deposited oxide during deposition and subsequent thermal cycling, Thin Solid Films 434 (2003) 190. [15] D. Maier-Schneider, A. Köprülülü, S.B. Holm, E. Obermeier, Elastic properties and microstructure of LPCVD polysilicon films, J. Micromech. Microeng. 6 (1996) 436.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153 [16] D.S. Gardner, T.L. Michalka, K.C. Saraswat, T.W. Barbee Jr., J.P. McVittie, J.D. Meindl, Layered and homogeneous films of aluminum and aluminum/silicon with titanium and tungsten for multilevel interconnects, IEEE Trans. Electron Devices 32 (1985) 174. [17] W.C. Chuang, D.T.W. Lin, Y.C. Hu, H.L. Lee, C.H. Cheng, P.Z. Chang, N.B. Quyen, A method integrating optimal algorithm and FEM on CMOS residual stress, J. Mech. 30 (2014) 123. [18] M.F. Doerner, D.S. Gardner, W.D. Nix, Plastic properties of thin films on substrates as measured by submicron indentation hardness and substrate curvature techniques, J. Mater. Res. 1 (1986) 845.
153
[19] M.N.G. Nejhad, C. Pan, H. Feng, Intrinsic strain modeling and residual stress analysis for thin-film processing of layered structures, J. Electron. Packag. 125 (2003) 4. [20] V. Senez, T. Hoffmann, P. Le Duc, F. Murray, Mechanical analysis of interconnected structures using process simulation, J. Appl. Phys. 93 (2003) 6039.
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Development of a process modeling for residual stress assessment of multilayer thin film structure H.J. Wang, H.A. Deng, S.Y. Chiang, Y.F. Su, K.N. Chiang ⁎ Advanced Micro-system Packaging and Nano-Mechanics Research Lab. Dept. of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan 300
a r t i c l e
i n f o
Available online 19 January 2015 Keywords: MEMS devices Process modeling Stoney's equation Residual stress Intrinsic stress Aluminum thin film
a b s t r a c t The mechanical behavior of Microelectromechanical systems (MEMS) devices is significantly influenced by fabrication processes conditions, deposition steps and geometry, e.g., processing temperature, thickness of thin film layers, deposition rate, etc. These processes under different deposition temperature will generate residual stress/strain in thin film layers, which will affect device sensing stability, accuracy and robustness, therefore, predicting and minimizing the residual stress become a critical issue on thin film devices design. This research develops a simulation methodology using finite element analysis (FEA) with process modeling technology to analyze the thermal stress/strain of thin films varying on different process procedures in a selected MEMS microphone device. In general, the residual stress/strain of multi-layers thin film is combined of intrinsic and thermal stresses, the thermal induced mechanical stress can be obtained using FEA but intrinsic stress which contains many uncertainties is very difficult to be defined. The residual stress in thin film layer on each processing procedure can be obtained from Stoney's equation. A comparison of the experiment and simulation results showed that the combined thermal induced stress and dislocation induced intrinsic stress in aluminum thin film will be rearranged after first annealing. The intrinsic stress, however, will affect the final residual stress when thickness of aluminum film is under 1 μm. The residual stress and the warpage of MEMS microphone are predicted by using processing modeling technology. The polysilicon will warp downward if the diaphragm is subjected to compressive stress. However, the polysilicon film will have much less warpage when intrinsic stress is positive. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Microelectromechanical systems (MEMS) microphone (Fig. 1) has become one of the most popular technologies for electronic devices because of its small size and high integration. Although the fabrication process of MEMS is compatible with semiconductor, the residual stress and undesirable pre-deformation of thin films during deposition and etching process will lead to reliability and yield problems. In the past few decades, numerous researches have been widely discussed in MEMS technology, but few of them focused on process simulation. This study developed a methodology using finite element analysis (FEA) with process modeling technology to analyze the residual stress in a structure of capacitive MEMS microphone. The residual stress of thin film is composed of thermal and intrinsic stresses [1]. The intrinsic stress, which is caused by deposition temperature, lattice mismatch, recrystallization and chemical reactions, etc., and is a process sensitive property, should be obtained from experiments. Extrinsic stress is applied to films mainly caused by coefficient ⁎ Corresponding author. Tel.: +886 3 5742925; fax: +886 3 5745377. E-mail address: [email protected] (K.N. Chiang).
http://dx.doi.org/10.1016/j.tsf.2015.01.014 0040-6090/© 2015 Elsevier B.V. All rights reserved.
of thermal expansion (CTE) mismatch and can also be called thermal stress which can be predicted using process modeling technique by element birth and element death in ANSYS®. In addition, extrinsic stress would also be caused by wafer chucking and wafer flatness
Fig. 1. The sketch of MEMS microphone.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
147
Fig. 2. Cross-section of 1P4M MEMS microphone.
analyzes the residual stresses of polysilicon diaphragm and aluminum backplate through experiment and numerical analysis. Polysilicon, the material of diaphragm, is often used in MEMS structure because it can be fabricated by thin film deposition process. For MEMS microphone, the polysilicon diaphragm would bend after etching process if it is subjected to compressive stress. This would make resonant frequency and microphone sensitivity far from the original design. The residual stress can be controlled by changing deposition and annealing temperatures [9–12]. Setting tensile film stress after etching process will let the microphone have better performance. The objective of this work is to obtain the intrinsic stress of thin film and discuss the intrinsic effect on a microphone structure. 2. Theory and experiment
Fig. 3. The process flow of the multilayered structure.
variation [2]. The average residual stress of multilayered structure is determined from Stoney’s experiment which measures the curvature on blanket wafer after each process. The experiment results indicated that the residual stress of thin films reduced and became stable after several thermal cycles. The residual stress can be measured by wafer curvature measurements and then the thermal stress subtracted from it to estimate the intrinsic component. Several authors have studied intrinsic stress in thin film [3–7], but few [8] discussed intrinsic effects in simulation. For capacitive type MEMS microphone, the component consists of a flexible diaphragm and a rigid backplate with acoustic holes. These two electrode plates are separated by an air gap. This study
The first theoretical formula for the evaluation of stresses in a thin film on a thick substrate was developed by Stoney [13]. The effective residual stresses of aluminum and polysilicon thin films on silicon wafers have been determined by Stoney’s equation, shown as below 2
σf ¼
Es t s 6ð1−νs Þt f
1 1 − R R0
ð1Þ
where σf is thin film stress; E s and ts are Young’s modulus and thickness of substrate, respectively; ν s is the Poisson’s ratio; tf is the film thickness; R0 and R is the radius of curvature on the substrate before and after film deposition. It assumes that ts N N tf, the lateral dimensions of the film and substrate are significantly greater than the thickness, and the substrate is elastically isotropic. The above formula does not involve the material properties of the film, and the film stress was obtained by measuring the curvature before and after film deposit.
Table 1 Processing steps and fabrication parameters of the multilayered structure. Procedure
1
2
3
4
5
Film Process Process temperature(°C) Film thickness (μm) Structure
Field oxide (FOX) Atmospheric pressure CVD 980 0.39
Poly-Si Low pressure CVD 620 0.2
ILD PE TEOS 400 0.85
IMD PE CVD 400 1.0
Al (TiN-AlSi1%Cu0.5%-TiN) Sputter 25-160 0.495
Number of thermal cycle Peak temperature(°C)
2 1,000
2 1,100
1 450
1 450
3 450
148
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
Table 2 Averaged residual stresses (MPa) of multilayer structure with 8 Samples.
Oxide Polysilicon ILD IMD Al
Before annealing
After 1st Annealing
After 2nd Annealing
11.08 −11.87 −87.54 −97.79 125.83
10.63 −2.51 −82.23 −23.39 439.3
6.00 −2.71 ----
Fig. 6. Stress vs. temperature in a multilayered structure after the deposition of the Al film.
Fig. 4. 1 μm plasma-enhanced chemical vapor deposition (PECVD) oxide with 450 °C thermal cycle.
For multilayer thin film structure, the effective stress (σeff) of whole films on substrate and the new stress (σΝ) of top layer thin film can be obtained by Eqs. (2) and (3), shown as below: σ eff ¼
σN ¼
2 Es t s 1 1 − XN 6ð1−νs Þ i¼1 t i R R0
σ eff
XN
t− i¼1 i
XN−1 j¼1
σ jt j
tN
ð2Þ
ð3Þ
where N, i, and j are layer number, and tN, ti and tj are thickness of Nth , ith, and jth layer individually. σj is stress of jth layer. A chip size of 1.5 mm × 1.5 mm with 450 μm thickness MEMS microphone (Fig. 2) using 1P4M (one polysilicon andfour metal layers) process was designed by ITRI. The first diaphragm metal layer is aluminum, and the second metal layer is a backplate. The processing flow,
Fig. 5. PECVD oxide with 500 °C thermal hysteresis loop [14].
thickness, deposition and annealing temperatures of multilayered structure are shown in Fig. 3 and Table 1. These films were deposited on (100) silicon wafer, and the curvatures of the substrate were measured using TENCOR FLX-2320 after each processing step. The samples were annealed under 1,000 °C (FOX) or 1,100 °C (Polysilicon) for 2 hours one to three times after film depositions to stabilize and minimize the influence on the stress in the next deposition step [14]. Different deposition and annealing temperatures result in different stress behavior in the polysilicon film [10–12], ranging from compressive to tensile stress. As the polysilicon film is annealed under 1,100 °C, the residual stress is reduced or almost no stress [15]. It is well known that reducing the stress on film can improve the sensing sensitivity of MEMS microphones. 3. Experiment results and discussion The films were deposited on 6" (100) silicon wafer. The samples, after each deposition step, will go through one to three times of annealing process to stabilize and minimize the residual stresses. The curvature of the substrate was measured at the end of each processing step. The stress history can be obtained by tracking the curvature. We assumed that the stresses of the prior films will not vary in the next deposition process since the post annealing has been applied to all steps to stabilize the residual stress. The average residual stresses in the multilayered structure before and after post annealing are presented in Table 2.
Fig. 7. Stress vs. temperature in the 0.49 μm Al film.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
149
Fig. 8. Schematic diagram of 2D FEA model before etching.
Fig. 9. Schematic diagram of 2D FEA model after etching.
Table 2 shows the residual stresses of each layer. The annealing process will reduce the residual stress of each layer except the Al film. The upper bound annealing temperatures is set as 1,100 °C, it is due to the in-situ stresses of field oxide and polysilicon cannot be measured by our experiment facility above that temperature. Eight samples have been measured in this research. The stress as function of temperature during the 450 °C thermal cycle in IMD layer is shown in Fig. 4. If the material is linear elastic, the residual stress will return to the original value as temperature cools down to room temperature. The stress hysteresis means that the intrinsic stress changed after the thermal cycling. Chen et al. [14] indicated that there is almost no hysteresis in the second cycle, as shown in Fig. 5. Therefore, after annealing process the stress in each layer will become stable and will not be altered in the next deposition processes. Tensile thermal stress is produced during cooling process in aluminum thin film because the CTE of aluminum is greater than silicon substrate. Although the film was sputtered at room temperature, it will be heated under high energy ion bombardment. The
temperature is 80 °C (process power, 2 kW) for only aluminum film and 160 °C (process power, 5 kW) for titanium nitride (TiN) sputtering close to it. The deposition of TiN as barrier layer can
Table 3 Material properties of microphone films.
Silicon (100) Field oxide Polysilicon ILD IMD Al[58]
Young’s Modulus (GPa)
Poisson’s Ratio
Coefficient of Thermal Expansion (ppm/°C)
129 94 169 120 85 69.7-0.037 × T
0.28 0.25 0.22 0.2 0.25 0.33
2.6 0.5 2.3 2 2 23.3 + 1.7 × 10−2 × T
Deposition Temp. (°C)
980 620 400 400 160
Fig. 10. The nonlinear material property of aluminum.
Table 4 Temperature dependent yielding stress of aluminum thin film [8]. Temperature(°C) yielding stress (MPa)
25 195
150 160
200 140
300 60
380 10
475 10
150
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153 Table 5 Etching radius of each layer. Etching radius (μm) Silicon Field oxide Polysilicon ILD IMD Al Oxide SiN
410 419 -419 422 -419 419
avoid hillock in aluminum film [16]. Therefore, 250 Å and 700 Å TiN films will be deposited before and after the aluminum film, respectively. Hysteresis stresses history of three thermal cycles is
presented in Fig. 6. It can be seen that the film is performed as linear elastic in the beginning of the second and third cycles and then entered plastic zone after 200 °C, and it indicated that the hysteresis stress decreased at the third cycle. A single Al film deposit on an oxide blanket wafer (Fig. 7) was made to see the multilayers effect (Fig. 6) on residual stress. The curve shown, Al film will enter the plastic zone when temperature reaches 200 °C. The residual stress of aluminum film was stabilized after the second cycle. The residual increased from 26 MPa to 356 MPa for single layer Al film, and from 125 MPa to 439 MPa for deposition of Al film on multilayered structure. It is because that most of the residual stress came from the non-recoverable plastic deformation. Although there is a little difference between Figs. 6 and 7, we can still see a similar thermo-mechanical behavior of Al thin film.
Fig. 11. Y-displacement of 2D FEA model after etching.
Fig. 12. (a) von-Mises stress; (b) Y-displacement of the Al backplate along the radius within 420 μm before and after etching.
Fig. 13. (a) X-stress; (b) Y-displacement of the polysilicon diaphragm along the radius within 420 μm before and after etching.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
151
Table 6 Intrinsic stress of a multilayered structure.
Residual stress (MPa) Thermal stress (MPa) Intrinsic stress (MPa)
Field oxide
Polysilicon
ILD
IMD
Al
6.0 −250.95 256.95
−2.71 −38.18 35.47
−82.23 −33.39 −48.84
−23.39 −25.13 1.74
439 196.87 242.13
Fig. 16. Simulation results of average stress vs. temperature for aluminum film from 0.2 μm to 2.0 μm.
Fig. 14. Thermal cycle of thermal stress and stress with 242 MPa intrinsic stress of the Al film deposited at 160 °C.
4. Numerical analysis FEA is widely adopted by MEMS researches, such as for design and assessment of residual stress [17]. In this research, process modeling technology is carried out to simulate deposition and etching processes on films. Each material on deposition temperature is assumed at stress-free condition, as the temperature cools down to room temperature, for a simple bilayer material without considering the bending effect, the thermal stress can be expressed through the following equation: σf ¼
E 1−ν
α f −α s ðT 1 −T 0 Þ
ð4Þ
f
where σf is the film stress; αf and αs are the coefficient of thermal expansion of film and substrate respectively; T 1 is the deposition temperature; and T 0 is the room temperature. However, for a 3D structure such as wafer and MEMS devices, FEA has to be applied to get more accurate thermal stress. Figs. 8 and 9 show the schematics of two-dimensional model of a microphone before and after the etching process (radius of 419 μm), respectively. A two-dimensional axial symmetry model was used to
represent round shape multilayer films. The material properties are presented in Table 3, and the nonlinear material property of aluminum is shown in Fig. 10. Temperature dependent yielding stress of aluminum thin film [8] is shown in Table 4. The coefficient of thermal expansion of Al film is greater than silicon substrate, the stress of aluminum film changed to tensile stress when the temperature cooled down from deposition temperature, the compressive stress is observed in other layers. The etching radius is shown at Table 5. Fig. 11 shows the displacement in Y direction after the etching of the oxide films. Figs. 12 and 13 show the von-Mises/X stress and Y displacements before and after the etching process, respectively. The stresses of the films before the etching are mainly caused by CTE mismatch between films and substrate, and partially relaxed by the oxide films’ etching process. The aluminum backplate is under tensile stress, so its Y-displacement is small and not easy to deflect, which is good for MEMS microphone. The X direction stress of the polysilicon diaphragm changes from −36.67 MPa to almost stress free. The intrinsic stresses effect is also examined in this paper. Table 6 shows the residual stresses, intrinsic stress and thermal stress in multilayered structure after annealing. The residual stress is the sum of intrinsic stress and thermal stress. For complementary metal– oxide–semiconductor (CMOS) MEMS structure, the residual stress will influence its mechanical behavior. The intrinsic stress may drastically change the mechanical behavior, and should be included into FEA. The intrinsic effects on simulation, one can add stress to linear-elastic materials directly, but not for plastic materials. For plastic material like aluminum, Fig. 14 shows the thermal stress at various temperatures under thermal cycle loadings and the test sample, aluminum film, with 242 MPa intrinsic stress was deposited at 160 °C. The schematic diaphragm shows that these stresses will reach yielding strength when temperature heating to
Fig. 15. Schematic of the aluminum films with lattice mismatch effect.
152
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153
shown in Fig. 17. These experimental data are consistent with the simulation results. Fig. 18 shows Y-displacement without and with intrinsic stress applied to the polysilicon diaphragm after etching process. The warpage of polysilicon diaphragm with 35.47 MPa intrinsic stress is −2.61 μm, which is smaller than the no-intrinsic stress case, the film becomes flat when the intrinsic stresses increase to 50 MPa and 75 MPa. From above simulation results, it is clear that designer can control the annealing and deposition temperatures to make diaphragm flat. 5. Conclusion
Fig. 17. Simulation and experimental results of average stress vs. temperature for 0.495 μm aluminum film.
175 °C. Because of the strain hardening effect in plastic zone, the stress of Al film with 242 MPa intrinsic stress increases by only 4.7 MPa when the temperature cooling back to room temperature. Therefore, the intrinsic stress effect of Al film can be ignored and is not considered in this research. Doerner et al. [18] indicated that yielding strength increases as the aluminum film thickness decreases. Nejhad et al. [19] assumed that the intrinsic stress are uniform distributed through the entire film. Thus one can adapt residual stresses by changing the deposition temperature. With these concepts, this research deposited the aluminum film by two steps. The first layer was deposited at 500 °C and set a thin layer of aluminum film about 0.2 μm to compensate the lattice mismatch effect and increase the interfacial strength to avoid crack. The second layer of aluminum thin film can be regarded as a bulk material without lattice mismatch. The schematic diaphragm is shown as Fig. 15. The average stresses in our simulation can be shown as function of temperature when the entire thickness of aluminum film from 0.2 μm to 2.0 μm (Fig. 16). At room temperature, the residual stress on wafer reduces as the increasing film thickness. It had good agreement with Senez’s [20] results. Comparing with the experimental curve, the calculated average stress on the aluminum thin film is
Fig. 18. Y-displacement without and with intrinsic stress applied to the polysilicon diaphragm.
The mechanical behavior of thin film is affected by different fabrication procedures such as depositing, annealing temperature and etching processes. These processes will induce different range of residual stress on thin film and affect the device performance. A FEA using process modeling technology is successfully developed in this research for thin film design, the simulation result is also validated with experimental data. The residual stresses, intrinsic stress and thermal stress in multilayered structure after annealing is obtained in this work. The residual stress is the sum of intrinsic stress and thermal stress. The intrinsic stress may drastically change the mechanical behavior, and should be included into FEA. However, for Al film, intrinsic effect is small and can be ignored. The aluminum backplate designed in this research is under tensile stress, so its Y-displacement is small and not easy to deflect, which is good for MEMS microphone. Acknowledgment The authors would like to thank the Advanced Micro-system Packaging and Nano-Mechanics Research Laboratory, ITRI, and National Science Council of Taiwan (NSC102-2221-E-007-038-MY3) for supporting this research, and also would like to thank Professor S. Liang, Georgia Institute of Technology, for his fruitful discussion on the residual stress behavior of Al material after high temperature annealing. References [1] T.L. Chou, S.Y. Yang, K.N. Chiang, Overview and applicability of residual stress estimation of film-substrate structure, Thin Solid Films 519 (2011) 7883. [2] J. Gong, P. Vukkadala, J. Sinha, K.T. Turner, Determining local residual stresses from high resolution wafer geometry measurements, J. Vac. Sci. Technol. B 31 (2013) 051205. [3] W. Buckel, Internal Stresses, J. Vac. Sci. Technol. 6 (1969) 606. [4] M.F. Doerner, W.D. Nix, Stresses and deformation processes in thin films on substrates, Crit. Rev. Solid State Mater. Sci. 14 (1988) 225. [5] K.S. Chen, Techniques in residual stress measurement for MEMS and their applications, in: C.T. Leondes (Ed.), MEMS/NEMS, Springer, US, 2006, p. 1252. [6] P.B. Ghate, L.H. Hall, Internal stresses in multilayered structures, J. Electrochem. Soc. 119 (1972) 491. [7] M. Malakouti, M. Ameri, P. Malekzadeh, Dynamic viscoelastic incremental-layerwise finite element method for multilayeredstructure analysis based on the relaxation approach, J. Mech. 30 (2014) 593. [8] J. Lee, A.S. Mack, Finite element simulation of a stress history during the manufacturing process of thin film stacks in VLSI structures, IEEE Trans. Semicond. Manuf. 11 (1998) 458. [9] A. Torkkeli, O. Rusanen, J. Saarilahti, H. Seppa, H. Sipola, J. Hietanen, Capacitive microphone with low-stress polysilicon membrane and high-stress polysilicon backplate, Sens. Actuators A85 (2000) 116. [10] J. Miao, R. Lin, L. Chen, Q. Zou, S.Y. Lim, S.H. Seah, Design considerations in micromachined silicon microphones, Microelectron. J. 33 (2002) 21. [11] X. Zhang, T.Y. Zhang, M. Wong, Y. Zohar, Rapid thermal annealing of polysilicon thin films, J. Microelectromech. Syst. 7 (1998) 356. [12] H. Guckel, D.W. Buurn, C.C.G. Visser, H.A.C. Tilmans, D. Deroo, Fine-grained polysilicon films with built-in tensile strain, IEEE Trans. Electron Devices 35 (1988) 800. [13] G.G. Stoney, The tension of metallic films deposited by electrolysis, Proc. R. Soc. London, Ser. A82 (1909) 172. [14] K.S. Chen, X. Zhang, S.Y. Lin, Intrinsic stress generation and relaxation of plasmaenhanced chemical vapor deposited oxide during deposition and subsequent thermal cycling, Thin Solid Films 434 (2003) 190. [15] D. Maier-Schneider, A. Köprülülü, S.B. Holm, E. Obermeier, Elastic properties and microstructure of LPCVD polysilicon films, J. Micromech. Microeng. 6 (1996) 436.
H.J. Wang et al. / Thin Solid Films 584 (2015) 146–153 [16] D.S. Gardner, T.L. Michalka, K.C. Saraswat, T.W. Barbee Jr., J.P. McVittie, J.D. Meindl, Layered and homogeneous films of aluminum and aluminum/silicon with titanium and tungsten for multilevel interconnects, IEEE Trans. Electron Devices 32 (1985) 174. [17] W.C. Chuang, D.T.W. Lin, Y.C. Hu, H.L. Lee, C.H. Cheng, P.Z. Chang, N.B. Quyen, A method integrating optimal algorithm and FEM on CMOS residual stress, J. Mech. 30 (2014) 123. [18] M.F. Doerner, D.S. Gardner, W.D. Nix, Plastic properties of thin films on substrates as measured by submicron indentation hardness and substrate curvature techniques, J. Mater. Res. 1 (1986) 845.
153
[19] M.N.G. Nejhad, C. Pan, H. Feng, Intrinsic strain modeling and residual stress analysis for thin-film processing of layered structures, J. Electron. Packag. 125 (2003) 4. [20] V. Senez, T. Hoffmann, P. Le Duc, F. Murray, Mechanical analysis of interconnected structures using process simulation, J. Appl. Phys. 93 (2003) 6039.