- Email: [email protected]
Residual stress gradients in electroplated nickel thin films
Microelectronic Engineering 134 (2015) 60–67
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Residual stress gradients in electroplated nickel thin films Yasin Kilinc a, Ugur Unal b,c, B. Erdem Alaca a,c,⇑ a
Department of Mechanical Engineering, Koç University, Istanbul, Turkey Department of Chemistry, Koç University, Istanbul, Turkey c Koç University Surface Science and Technology Center, Istanbul, Turkey b
a r t i c l e
i n f o
Article history: Received 1 September 2014 Accepted 31 January 2015 Available online 12 February 2015 Keywords: Electroplating Residual stress gradient Microelectromechanical systems (MEMS) Nickel thin films
a b s t r a c t Residual stress gradients in electroplated nickel films of 1 lm thickness are characterized for a wide range of current densities (1–20 mA/cm2) and electroplating temperatures (30–60 °C) in a nickel sulfamate bath. Although a variety of stress measurements is available, exploration of stress gradients remain unstudied at the scale of 1 lm. Stress gradients – unlike uniform stresses – can cause significant bending even in monolayered released structures. Moreover, examples of misinterpretation of wafer curvature data as a measure of stress gradients exist in the literature. Based on these motivations, monolayered Ni microcantilevers are employed in this work as mechanical transducers for the characterization of stress gradients within the nickel film. Experiments are supported with finite element simulations. Residual stress gradient is found to vary in the range of about 130 to 70 MPa/lm with the sign change indicating a transition from downward to upward deflection of the microcantilever. Thus, a window of electroplating parameters is established yielding zero residual stress gradients, i.e. straight cantilevers, without the use of any additive agents. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction Nickel has a wide range of applications in microelectromechanical systems (MEMS) due to its attractive mechanical and magnetic properties. Advances in electroplating and extensive employment of thick nickel layers in MEMS applications make electroplating a method of choice for the deposition of nickel. Submicron deposition thicknesses are reported as well [1]. In MEMS literature, nickel sulfamate-based nickel electroplating is mostly carried out in a temperature and current density range of 30–60 °C and 1–20 mA/ cm2, respectively, while the pH of the electrolyte remains in the range of 3.5–5.0. Among various MEMS applications one can mention Ni structures serving as micromechanical resonators with magnetic actuation in various biological and physical sensor studies [2–5]. Ni cantilevers fabricated by LIGA provide platforms for mechanical property measurements [6,7]. Ni layers are also utilized as micromechanical switches [1,8,9]. Especially in optical readout schemes, where large scatter of light reflected from severely bent microstructures results in lower signal amplitudes, initial bending of
⇑ Corresponding author at: Department of Mechanical Engineering, Koç University, Istanbul, Turkey. Tel.: +90 212 338 17 27; fax: +90 212 338 1548. E-mail address: [email protected] (B.E. Alaca). http://dx.doi.org/10.1016/j.mee.2015.01.042 0167-9317/Ó 2015 Elsevier B.V. All rights reserved.
the cantilever is undesired. Hence, deposition processes need to be tightly controlled to eliminate bending upon release. Unlike uniform residual stresses that lead to in-plane deformations, residual stress gradients are primarily responsible for out-ofplane bending [10]. In Fig. 1, one such cantilever is shown. Two scenarios for stress distribution along the cross-section A–B are possible. Depending on fabrication conditions that cause either increasing compressive or tensile stresses through film thickness from B to A, downward or upward deflection is obtained. In the literature, uniform residual stresses in thin films are studied extensively with various techniques [8,11–13]. However, residual stress gradients in thin metallic films remain mostly unstudied. Among the limited number of accounts one can mention the recent study on strain-gradient-free 4.5-lm-thick Ni cantilevers obtained through multiple electroplating steps [14]. A few other studies [10,11,15] determined residual stress gradients for metallic materials under some specific deposition conditions with fairly limited fabrication parameter range. Therefore a systematic stress gradient characterization as a function of fabrication parameters supported by a microstructural analysis would provide a much needed input in the field of electroplated nickel films to understand out-of-plane deformations encountered in surface micromachining. This work comprises such a study on stress gradients in Ni thin films, where the effect of uniform stresses is omitted in measurements.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
In the remainder of this work, a mechanical framework for the measurement of residual stress gradients will be discussed first followed by a description of employed experimental procedures using 1-lm-thick, 10-lm-wide and 100-lm-long cantilevers. Results are then presented leading to a useful set of deposition parameters for nickel sulfamate bath resulting in near-zero bending of Ni cantilevers. 2. Residual stress gradients and their characterization Residual stresses in microstructures can be modeled with the combination of a uniform stress (r) and a stress gradient across the thickness ðrrÞ. Uniform residual stresses are found in either tensile or compressive form leading to in-plane expansion or contraction of the microstructure after release. In contrast, the relief of a residual stress gradient induces a bending moment resulting in the formation of a curved microstructure with a strain gradient of opposite sign as shown in Fig. 2a. The residual stress gradient for a monolayer cantilever is given in Eq. (1) in terms of the modulus of elasticity (E) and tip deflection (d) of a cantilever of length L [10]. The assumption of linear stress distribution, i.e. constant stress gradient, is frequently employed in the literature [10,14,16–18], which is especially proper for very thin films such as the 1-lm-thick nickel film of this study:
rr ¼ E
2d
ð1Þ
L2
Hence, the residual stress gradient can be determined by measuring process-dependent parameters of tip deflection and modulus of elasticity. While tip deflection can be measured by optical means, modulus of elasticity is determined through measuring the first-mode resonance frequency, f1, of a cantilever. In case of a monolayer, homogeneous cantilever with a rectangular uniform cross-section, the relation between E and f1 is given by Eq. (2). This relation remains unaltered by the presence of residual stress gradients:
f 1 ¼ 0:162
h
sffiffiffiffi E
L2
q
ð2Þ
In Eq. (2) two new parameters, q and h, density and thickness of a cantilever, respectively, are introduced. Being a relatively process-independent parameter, q is taken as the bulk mass density of 8908 kg/m3 throughout this work [19]. Fig. 2b provides a case study for the verification of Eq. (1), where a finite element analysis (FEA) of a cantilever is carried out. COMSOL Multiphysics 4.3 FEA software with 3d free-tetrahedral mesh elements is employed to study the release of a cantilever with a
61
stress gradient of 1 MPa/lm. The thickness, width and length of the cantilever are chosen as 100 100 1000 lm, respectively. Fig. 2b depicts the linear variation of strain along the midline of a cross-section within the cantilever. Strain gradient from the analytical treatment deviates from the slope of the strain plot in Fig. 2b by less than 0.5%. It is not uncommon to encounter wafer curvature measurements employed in the literature for the characterization of residual stress gradients. There are reports claiming wafer curvature measurements encompass the effect of stress gradients as well as uniform stresses [20]. However, use of wafer curvature data might be misleading, as there are cases where infinite radius of curvature, i.e. straight substrate, does not guarantee straight cantilevers upon release. Similarly, finite wafer curvature associated with thin film stresses might lead – in a seemingly counterintuitive fashion – to a straight cantilever. To elucidate the misinterpretation associated with wafer curvature measurements, let us consider a <1 0 0> Si substrate of a diameter (D) of 10 mm and a thickness (T) of 500 lm. Let us also assume that a Ni film with a thickness (h) of 10 lm is coated on the top surface of this substrate. If there is a stress gradient of 20 MPa/ lm in the Ni film (Fig. 3a), it leads to a negligible wafer curvature ð1=qÞ of 8.864 104 m1. This can be computed by incorporating the resulting moment due to stress gradients into the moment equilibrium in a similar fashion to Stoney formulation. The associated formulation is given in Fig. 3a as well, and invariant biaxial modulus of <1 0 0> Si (E/(1m)) is taken as 180.51 GPa [21]. If this curvature measurement is misinterpreted as an indication of uniform stress only, it would correspond to a stress level of less than 1 MPa. However, if one fabricates and releases a Ni cantilever of a length (L) of 100 lm, this seemingly harmless effect leads to a considerable tip deflection of 0.5 lm. On the contrary, one can consider the same substrate/thin film system, where the thin film is under a uniform tensile stress of 100 MPa (Fig. 3b). Using Stoney formulation given in Fig. 3b, the associated wafer curvature is predicted to be 0.133 m1. This measurement is 150 times larger than the curvature in the presence of the aforementioned stress gradient case. More interestingly, as all associated stresses are uniform, they do not lead to any permanent bending of the cantilever upon release. Hence, a larger curvature does not necessarily mean a bending and stiction risk for surface micromachining. Aforementioned loading scenarios are verified using COMSOL as described in connection with Fig. 2 as well. They support the claim that one has to release microstructures for a reliable characterization of stress gradients, and Stoney formulation alone is not an efficient way for the characterization of surface micromachining.
Fig. 1. The effect of residual stress gradient, rr, on bending of a monolayer cantilever. (a) With compressive stresses increasing with increasing distance from thin filmsubstrate interface, the cantilever bends towards the substrate. (b) If residual stresses become more tensile away from the interface, the cantilever bends upwards.
62
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 2. (a) Stress gradient ðrrÞ in a film is converted into a strain gradient ðrÞ following the release process. (b) The finite element simulation of a microcantilever with a predefined stress gradient also confirms the same strain gradient upon release.
3. Experimental procedures 3.1. Microfabrication The employed fabrication flow shown in Fig. 4 is based on a single-mask lithographic process on Si. 400 -diameter, <1 0 0> Si wafers are chosen with a resistivity of 0.5–1.0 X cm and a thickness of 525 ± 25 lm. The process consists of RF magnetron sputtering, photolithography, nickel electroplating and release of microcantilevers. An RF magnetron sputtering system (Vaksis PVD-Handy/ 2S) is employed to deposit a blanket coating of about 20-nm-thick Cr and 100-nm-thick Au on Si wafers as an adhesion promoter and
seed layer, respectively. Then photolithography is carried out using AZ 1514H positive photoresist by Microchemicals. After defining molds for electroplating by photolithography, 400 -diameter Si wafers are cut into 1 cm 1 cm dice with a scriber (OEG MR200) to perform electroplating for each die separately. Nickel electroplating is performed with a nickel sulfamate solution consisting of 600 g/l nickel sulfamate (Ni(SO3NH2)24H2O), 10 g/l nickel chloride (NiCl26H2O) and 40 g/l boric acid (H3BO3). The acidity (pH) of the solution is adjusted from 5.2 to 4.0 by adding approximately 10 g of sulfamic acid powder. The release of microcantilevers is achieved by subsequent wet etching of exposed Au and Cr layers by using the etchants of
Fig. 3. Two scenarios for wafer curvature measurement for a 10-lm-thick Ni coating on a 500-lm-thick, <1 0 0> Si substrate. (a) In the presence of a stress gradient of 20 MPa/lm in Ni, a negligible curvature is obtained. However, release of a 100-lm-long Ni cantilever leads to a considerable tip deflection of 0.5 lm. (b) In the presence of a uniform stress of 100 MPa, a 150-times larger curvature results. However, no tip deflection will be encountered as all stresses in Ni are uniform.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
63
Fig. 4. Fabrication flow is composed of (a) wafer cleaning, Cr and Au layer deposition by RF magnetron sputtering, (b) lithography and electroplating of nickel, (c) removal of the photoresist (PR) and release of microcantilevers, and finally (d) etching of Cr and Au layers from underneath cantilevers and drying.
GE-8148 from Transene Inc. and CEP-200 from Microchrome Inc., respectively. This is followed by etching of Si substrate with a 45% KOH solution at 60 °C to a depth of approximately 30 lm. Final etching of Cr and Au layers from underneath released cantilevers provides monolayer nickel cantilevers. After the release of microcantilevers, each sample is immersed in water, acetone and IPA subsequently for at least 5 min and dried at room temperature on a flat surface to prevent stiction of microcantilevers. Further details regarding fabrication processes and equipment can be found elsewhere [2]. A wide electroplating operation range is chosen to investigate stress gradients: three current densities of 1, 5 and 20 mA/cm2 and three temperatures of 30, 45 and 60 °C. Thus there are 9 different cases studied separately with DC and pulse electroplating. Despite best efforts to reach uniform 1 lm nickel film thickness across samples, considerable film thickness non-uniformity – especially pronounced for the current density of 20 mA/cm2 – requires determination of exact cantilever thickness. Hence, 15 cantilevers across each sample are chosen to determine the exact film thickness. Three thickness measurements per cantilever are carried out using a surface profiler (Veeco, Dektak 8). The average value of the three thickness measurements is assigned as the thickness of the particular cantilever. Afterwards, tip deflections and resonance frequencies of these individual cantilevers are traced to quantify the corresponding stress gradient for each cantilever precisely. A DC power supply along with a multimeter are employed to perform the DC electroplating process. The pulse plating experiments are performed using a pe86cb type reverse pulse power
supply from Plating Electronic. In pulse electroplating, a squarewave pattern is chosen with a duty cycle of 0.5 at the frequency of 1 kHz. The mean current values are matched with DC electroplating values to ensure proper comparison between DC and pulse electroplating. Electroplating process is carried out using 1 cm 1 cm samples shown in Fig. 5. Contact areas for electroplating are visible at each corner of the die in Fig. 5b. Electroplating is performed using two alligator clips placed diagonally to provide symmetric film thickness distribution across the die. Cantilever anchors are aligned parallel to <1 1 0> to minimize the undercut of silicon supports during wet anisotropic etching of silicon. This alignment ensures definition of satisfactory and reproducible clamping conditions with minimal undercut for each cantilever. Thus, variation of cantilever properties due to effective length variation and poorly defined clamping conditions is substantially eliminated. 3.2. Static and dynamic measurements Deflections of released cantilevers are measured by utilizing white light interferometry (WLI) technique. WLI measurements performed with an optical profiler (Bruker ContourGT 3d) provide the resolution of deflection on the order of 1 nm. The modulus of elasticity of thin films is investigated by determining resonance frequencies of microcantilevers in air. Electromagnetic actuation with an electromagnetic coil and a set of permanent magnets [2] is carried out together with optical readout via laser doppler vibrometry (LDV) (Polytec OFV-2500 Vibrometer
Fig. 5. (a) Layout with 46 dice of 1 cm 1 cm dimensions. (b) Each die consists of 58 cantilevers along with letters and numbers for labeling each cantilever. All cantilevers are identical. (c) A close-up view of two cantilevers on a die. (d) This scanning electron micrograph verifies the complete release of the fabricated cantilever from base Si substrate. Tilted Si sidewall due to KOH etching can be observed.
64
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Table 1 A list of samples highlighting the matrix of process conditions utilized in this study. Die number A1 A2 A3 B1 B2 B3 C1 C2 C3
Deposition temperature (°C)
Deposition current density (mA/cm2)
Ni thickness (nm) DC plating
Ni thickness (nm) pulsed plating
30 30 30 45 45 45 60 60 60
1 5 20 1 5 20 1 5 20
969 ± 28 980 ± 90 1163 ± 211 931 ± 29 889 ± 103 1009 ± 135 1031 ± 34 1007 ± 82 1165 ± 171
1006 ± 37 900 ± 108 960 ± 259 1075 ± 20 1005 ± 123 1072 ± 281 976 ± 34 1062 ± 128 1186 ± 285
Controller) to determine resonance frequencies of microcantilevers whose deflection and thickness values were already recorded. Texture measurements are carried out with Bruker AXS Advance D8 with DaVinci system. CuKa source in point focus coupled with polycapillary optics is collimated to 1 mm with a circular collimator. A 1d detector (VANTEC-1) is used as a detector. Pole figures for Au and Ni are measured by recording diffraction by rotating the sample around the sample normal (0° < u < 360°) at each tilt angle (0° < v < 85°). MULTEX software by Bruker AXS is used to analyze pole figure measurements. Calculated pole figures with MULTEX are imported into MTEX [22], a MATLAB toolbox for quantitative texture analysis, for inverse pole figure calculations and plotting. Crystallite size of Ni films is estimated by using quantitative Rietveld analysis with TOPAS software supplied by Bruker-AXS by taking microstrain into account. Crystallite size is calculated by the software based on refinements realized by introducing Ni, Au and Si phases into the software. Furthermore, cross-sections of thin films are examined by scanning electron microscopy (Zeiss Ultra Plus Field Emission SEM with Bruker XFlash 5010 SDD Detector). 4. Results A set of nine different deposition conditions is utilized throughout the study as summarized in Table 1 and fabricated separately
Fig. 6. The current density of 20 mA/cm2 results in excessive thickness nonuniformity across die A3. The corners with electrical contacts are indicated with ‘‘C’’. All numbers are reported in nm.
Table 2 The comparison of cantilever tip deflection measurements between DC and pulse plated samples implies that there is no significant difference. Die number
DC Plating (lm)
Pulse plating (lm)
A1 A2 A3 B1 B2 B3 C1 C2 C3
4.53 ± 0.74 0.21 ± 0.84 1.40 ± 0.55 4.40 ± 0.39 2.65 ± 0.51 2.12 ± 0.75 0.83 ± 0.23 3.45 ± 0.33 2.03 ± 0.51
3.17 ± 0.59 1.36 ± 0.38 – 4.29 ± 1.10 2.94 ± 0.48 1.70 ± 0.66 0.56 ± 0.24 2.70 ± 0.79 1.71 ± 0.46
with DC and pulse electroplating. 15 cantilevers in each die are utilized for the determination of nickel film thickness, tip deflection and modulus of elasticity to characterize residual stress gradients. Nickel film thickness distribution is also demonstrated in Table 1. There is a noteworthy thickness variation for samples fabricated with a 20 mA/cm2 current density. This variation is reduced significantly in the case of 1 mA/cm2. As an example, the thickness variation for sample A3 of Table 1 is depicted via a contour plot in Fig. 6, where the distribution is observed to be symmetric with respect to the two contact points designated with C. Film thickness decreases toward the two contact points, while it increases toward the remaining two corners, thereby resulting in a saddle-type distribution throughout the die. This thickness distribution indicates the necessity to track individual cantilevers before and after release rather than assigning an average cantilever thickness to the whole die. Table 2 lists resulting tip deflections after release. Since DC and pulse plating lead to similar cantilever tip deflections for most of the cases, DC plated samples are chosen for subsequent stress gradient measurements. Upon release, cantilever tip deflections are found to be upwards or downwards depending on employed process parameters. Combinations of different current densities and electrolyte temperatures demonstrate a transition between upward and downward deflections as shown in Fig. 7. Adopting the sign convention in Fig. 2, deflections of upward bent cantilevers will be taken as positive. Moduli of elasticity of cantilevers are reported in Fig. 8. The mean modulus of elasticity of nickel is found to be in the range of 128.2–184.4 GPa, which is smaller than the modulus of elasticity of bulk nickel of 219 GPa [23]. Having the moduli of elasticity and cantilever tip deflections characterized, it becomes now possible to link stress gradients to electroplating parameters via Eq. (1). Thus as the prime outcome of this study, Fig. 9 summarizes the residual stress gradient variation within the current density and electrolyte temperature ranges of 1–20 mA/cm2 and 30–60 °C, respectively. The stress gradient is found to vary in the range of 126.5 to +65.6 MPa/lm. What is interesting about the results is the fact that obtaining straight
Fig. 7. A plot of cantilever tip deflections. The trend shows that obtaining a straight cantilever is possible for a certain combination of deposition parameters.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 8. The modulus of elasticity measurements with respect to electroplating temperature and current density.
Fig. 9. Stress gradients with respect to the electroplating temperature and current density. Numbers I–IV refer to blanked-coated Ni films for microstructure characterization experiments.
cantilevers after release is possible with a suitable choice of deposition parameters. For example, current densities of approximately 14 mA/cm2 and 5 mA/cm2 result in stress-gradient-free, straight beams for plating temperatures of 45 °C and 30 °C, respectively, without the use of any stress-reducing agents. Additional Ni film samples are electroplated with fabrication parameters designated with I–IV in Fig. 9 to evaluate the effects of the current density and electrolyte temperature on film internal structure by performing texture and grain size measurements. Diffractions from Si substrate, Au seed layer and Ni film are seen on the XRD patterns in Fig. 10. Crystallite size is estimated for the nickel films from I to IV as 30.9 nm, 46.1 nm, 53.2 nm and 46 nm, respectively.
5. Discussion As shown in Table 2, DC and pulse plating follow the same trend in terms of cantilever tip deflection. Similarity between DC- and pulse-plated cantilevers deposited with the same current density in terms of internal structure and residual stress is also
Fig. 10. XRD patterns of approximately 1-lm-thick films electroplated at conditions shown in Fig. 9. For better visualization of differences, an offset is introduced between plots in vertical direction.
65
emphasized in the literature [24]. Thus, stress gradient characterization is performed only for DC-plated cantilevers. Since properties of electroplated films are highly sensitive to deposition conditions such as current density, plating temperature, pH and chemistry of the electrolyte and film thickness, it is quite difficult to find an identical study in the literature for comparison of the results of an experiment. However, trends showing relationships between two independent parameters can still be extracted as explained below. The effect of electroplating conditions on the modulus of elasticity is investigated thoroughly in several studies [25]. Variation of modulus of elasticity was attributed to changes in the film internal structure by the introduction of non-metallic species into deposited films by hydrolysis with increasing electroplating temperature and shifting from the instantaneous nucleation and growth mechanism to progressive nucleation mechanism by increasing current density. According to the stress gradient results in Fig. 9, there is a transition from negative to positive stress gradient at electroplating temperatures of 30 and 45 °C. This transition is expected to be related to the film internal structure. Effects of the electrolyte temperature and current density on the film internal structure are investigated systematically by studying cross sections of electroplated films. As shown in Fig. 11a, cross-sections of electroplated nickel films are formed as follows: Coating and development of a photoresist (PR) layer on a silicon substrate with an RF-magnetron-sputtered Au seed layer continues with growth of nickel films electroplated at operating conditions shown in Fig. 9 with numbers I–IV. Then, the PR layer is removed to allow SEM examination of the cross section of deposited nickel films. Fig. 11b demonstrates SEM micrographs of the cross sections of 1-lm-thick films fabricated using designated conditions. In I, low current density results in a needle-like internal structure. However, the current density of 20 mA/cm2 and electroplating temperature of 30 °C lead to somewhat strongly textured structure as shown in III. The cross section shown in IV seems to have smoother surface compared with the surface shown in III. Blanket-coated nickel film samples employed for the SEM crosssection evaluation are also used for texture and grain size measurements. Several XRD studies [24,26] state that the texture of nickel films deposited on <1 1 1> oriented gold changes from <1 1 0> orientation at low current densities to <1 0 0> texture by increasing current density. Amblard et al. [26] explained this shift with a specific inhibition mechanism correlating to adsorbed hydrogen for a nickel sulfate based electrolyte. In this study, the crystallographic orientation of the polycrystalline Au seed layer evaluated before the texture of Ni films indicates that the preferred orientation is indeed in <1 1 1> direction with fiber texture and {1 1 1} planes parallel to substrate surface. As seen in Fig. 10, (1 1 1) peak of Ni overlaps with (2 0 0) peak of Au. Although Au layer shows strong <1 1 1> orientation and the contribution of its (2 0 0) peak at 2h = 44.5° might be low, (2 0 0), (2 2 0) and (3 1 1) peaks of Ni are employed for the estimation of the preferred orientation. All of the analyzed nickel films possess fiber texture in normal direction to the sample surface and corresponding planes are found to be parallel to the sample surface. Contrary to theoretical predictions for electroplated Ni films, all films (I–IV in Fig. 9) show a preferred orientation mainly in <5 1 1> crystal direction. A very weak <1 1 1> texture is found to develop for all films electroplated at high current density (II–IV in Fig. 9). In all cases, grain sizes of microstructures are on the order of a few tens of nanometers. Deposition at high current density (II–IV in Fig. 9) results in larger crystallite sizes of approximately 50 nm. However, the effect of deposition temperature on the crystallite size is not as significant as the current density. Dependence of the crystallite size on the current density can be explained by
66
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 11. (a) Preparation of cross-sections for scanning electron microscopy evaluation. (b) SEM micrographs of film cross-sections. Each letter corresponds to a specific fabrication condition given in Fig. 9.
inhibited nickel film growth as a result of hydrogen evolution on the cathode pronounced at low current densities [27,28]. Fritz et al. [24] analyzed cross-sections of electroplated thick (15 lm) films and found that columnar internal structure formed at low current densities turns into granular structure as employed current density increases. However, in this study, where we address much thinner films, there is no such transition and internal structure seems to be granular for all cases. Temporal evolution of stress gradients under room conditions is also studied by tracking tip displacements of various cantilevers 6 and 15 months after the release. Cantilevers were kept in closed containers under ambient conditions throughout this time. Creep effect due to residual stresses, thermal loadings due to mean ambient temperature shifts, hydrogen desorption from a deposited film with time are held responsible for temporal changes [29]. In this study, shifts in tip deflection are found to be negligible over a period of 15 months. This observation underlines employment of single-layer instead of multilayer cantilevers to obtain time-independent tip deflections which improve reliability of sensors. 6. Conclusions Nickel electroplating is a popular technique for the fabrication of micromechanical structures. With miniaturized structures finding more and more applications, a systematic stress gradient
characterization of micrometer-thick nickel films with electroplating parameters is needed. This study is intended to provide a guideline for recipe formation aiming at straight nickel microstructures. For this purpose microcantilevers are used as transducers for residual stress gradient measurement. Residual stress gradient is found to vary in the range of about 130 to 70 MPa/lm, where a transition from downward to upward deflection is observed. Hence, a window of electroplating parameters is established yielding zero residual stress gradients, i.e. straight cantilevers. For the deposition temperature of 30 °C, a current density of about 5 mA/ cm2 is needed, whereas a current density of approximately 14 mA/cm2 leads to straight cantilevers at 45 °C. No such condition was obtained at 60 °C. Furthermore, stress gradients are observed to exhibit negligible change 15 months after first measurements highlighting improved reliability of sensors made of monolayer cantilevers. Acknowledgements Authors thank Prof. Hakan Urey for access to resonance frequency measurement setup. Contributions by Erhan Ermek and Zuhal Tasdemir during characterization experiments are acknowledged. Discussions on electroplating and pulse plating by Prof. Levent Demirel, Dr. Annamaria Miko and Prof. Ozgur Birer are also gratefully acknowledged. Funding was provided by Tubitak under Grant no. 109E222 and 111E184.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
References [1] T. Itoh, S. Kawamura, K. Kataoka, T. Suga, Sens. Actuators, A 123–124 (2005) 490. [2] A. Ozturk, H.I. Ocakli, N. Ozber, H. Urey, I.H. Kavakli, B.E. Alaca, Photonics Technology Letters, IEEE 20/23 (2008) 1905. [3] E. Timurdogan, N. Ozber, S. Nargul, S. Yavuz, M.S. Kilic, I.H. Kavakli, H. Urey, B.E. Alaca, Biosens. Bioelectron. 26 (1) (2010) 195. [4] O. Cakmak, C. Elbuken, E. Ermek, A. Mostafazadeh, I. Baris, B. Erdem Alaca, I.H. Kavakli, H. Urey, Methods 63 (3) (2013) 225. [5] E. Uzunlar, B. Beykal, K. Ehrlich, D. Sanli, A. Jonáš, B.E. Alaca, A. Kiraz, H. Urey, C. Erkey, J. Supercrit. Fluids 81 (2013) 254. [6] H. Majjad, S. Basrour, P. Delobelle, M. Schmidt, Sens. Actuators, A 74 (1–3) (1999) 148. [7] J. Zhang, J. Mater. Process. Technol. 123 (3) (2002) 329. [8] Y.-H. Zhang, G.-F. Ding, Y.-L. Cai, H. Wang, B. Cai, Mater. Charact. 57 (2) (2006) 121. [9] Y.B. Wu, G.F. Ding, J. Wang, C.C. Zhang, H. Wang, Microelectron. Eng. 87 (11) (2010) 2035. [10] C.W. Baek, Y.-K. Kim, Y. Ahn, Y.-H. Kim, Sens. Actuators A 117 (2005) 17. [11] Z.M. Zhou, Y. Zhou, H. Mao, J. Mater. Sci. Technol. 25 (1) (2009) 77. [12] C. Genzel, Int. Centre Diffr. Data 44 (2001). [13] A. Atkinson, S.C. Jain, J. Raman Spectrosc. 30 (10) (1999) 885.
67
[14] H.-H. Yang, M.-H. Seo, C.-H. Han, J.-B. Yoon, J. Micromech. Microeng. 23 (5) (2013) 055016. [15] S. He, J.S. Chang, L. Li, H. Ho, Sens. Actuators, A 154 (1) (2009) 149. [16] A.K. Chinthakindi, D. Bhusari, B.P. Dusch, J. Musolf, B.A. Willemsen, E. Prophet, M. Roberson, P.A. Kohl, J. Electrochem. Soc. 149 (8) (2002) H139. [17] W. Fang, J.A. Wickert, J. Micromech. Microeng. 6 (3) (1996) 301. [18] V. Mulloni, F. Giacomozzi, B. Margesin, Sens. Actuators, A 162 (1) (2010) 93. [19] T. Fritz, M. Griepentrog, W. Mokwa, U. Schnakenberg, Electrochim. Acta 48 (20) (2003) 3029. [20] J. Luo, M. Pritschow, A. Flewitt, S. Spearing, N. Fleck, W. Milne, J. Electrochem. Soc. 153 (10) (2006) D155. [21] W.A. Brantley, J. Appl. Phys. 44 (1) (1973) 534. [22] F. Bachmann, R. Hielscher, H. Schaeben, Solid State Phenom. 160 (2010) 63. [23] W.N. Sharpe Jr., N. William, Mechanical Properties of MEMS Material, CRC Press, 2002. [24] T. Fritz, W. Mokwa, U. Schnakenberg, Electrochim. Acta 47 (1) (2001) 55. [25] J.K. Luo, A.J. Flewitt, S.M. Spearing, N.A. Fleck, W.I. Milne, Mater. Lett. 58 (17– 18) (2004) 2306. [26] J. Amblard, I. Epelboin, M. Froment, G. Maurin, J. Appl. Electrochem. 9 (2) (1979) 233. [27] M. Vogelaere, V. Sommer, H. Sprigborn, U. Mishelsen-Mohammadein, Electrochim. Acta 47 (2001) 55. [28] K.C. Chan, N.S. Qu, D. Zhu, Text. Microstruct. 30 (1-2) (1997) 43. [29] W. Fang, H.-C. Tsai, C.-Y. Lo, Sens. Actuators, A 77 (1) (1999) 21.
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Residual stress gradients in electroplated nickel thin films Yasin Kilinc a, Ugur Unal b,c, B. Erdem Alaca a,c,⇑ a
Department of Mechanical Engineering, Koç University, Istanbul, Turkey Department of Chemistry, Koç University, Istanbul, Turkey c Koç University Surface Science and Technology Center, Istanbul, Turkey b
a r t i c l e
i n f o
Article history: Received 1 September 2014 Accepted 31 January 2015 Available online 12 February 2015 Keywords: Electroplating Residual stress gradient Microelectromechanical systems (MEMS) Nickel thin films
a b s t r a c t Residual stress gradients in electroplated nickel films of 1 lm thickness are characterized for a wide range of current densities (1–20 mA/cm2) and electroplating temperatures (30–60 °C) in a nickel sulfamate bath. Although a variety of stress measurements is available, exploration of stress gradients remain unstudied at the scale of 1 lm. Stress gradients – unlike uniform stresses – can cause significant bending even in monolayered released structures. Moreover, examples of misinterpretation of wafer curvature data as a measure of stress gradients exist in the literature. Based on these motivations, monolayered Ni microcantilevers are employed in this work as mechanical transducers for the characterization of stress gradients within the nickel film. Experiments are supported with finite element simulations. Residual stress gradient is found to vary in the range of about 130 to 70 MPa/lm with the sign change indicating a transition from downward to upward deflection of the microcantilever. Thus, a window of electroplating parameters is established yielding zero residual stress gradients, i.e. straight cantilevers, without the use of any additive agents. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction Nickel has a wide range of applications in microelectromechanical systems (MEMS) due to its attractive mechanical and magnetic properties. Advances in electroplating and extensive employment of thick nickel layers in MEMS applications make electroplating a method of choice for the deposition of nickel. Submicron deposition thicknesses are reported as well [1]. In MEMS literature, nickel sulfamate-based nickel electroplating is mostly carried out in a temperature and current density range of 30–60 °C and 1–20 mA/ cm2, respectively, while the pH of the electrolyte remains in the range of 3.5–5.0. Among various MEMS applications one can mention Ni structures serving as micromechanical resonators with magnetic actuation in various biological and physical sensor studies [2–5]. Ni cantilevers fabricated by LIGA provide platforms for mechanical property measurements [6,7]. Ni layers are also utilized as micromechanical switches [1,8,9]. Especially in optical readout schemes, where large scatter of light reflected from severely bent microstructures results in lower signal amplitudes, initial bending of
⇑ Corresponding author at: Department of Mechanical Engineering, Koç University, Istanbul, Turkey. Tel.: +90 212 338 17 27; fax: +90 212 338 1548. E-mail address: [email protected] (B.E. Alaca). http://dx.doi.org/10.1016/j.mee.2015.01.042 0167-9317/Ó 2015 Elsevier B.V. All rights reserved.
the cantilever is undesired. Hence, deposition processes need to be tightly controlled to eliminate bending upon release. Unlike uniform residual stresses that lead to in-plane deformations, residual stress gradients are primarily responsible for out-ofplane bending [10]. In Fig. 1, one such cantilever is shown. Two scenarios for stress distribution along the cross-section A–B are possible. Depending on fabrication conditions that cause either increasing compressive or tensile stresses through film thickness from B to A, downward or upward deflection is obtained. In the literature, uniform residual stresses in thin films are studied extensively with various techniques [8,11–13]. However, residual stress gradients in thin metallic films remain mostly unstudied. Among the limited number of accounts one can mention the recent study on strain-gradient-free 4.5-lm-thick Ni cantilevers obtained through multiple electroplating steps [14]. A few other studies [10,11,15] determined residual stress gradients for metallic materials under some specific deposition conditions with fairly limited fabrication parameter range. Therefore a systematic stress gradient characterization as a function of fabrication parameters supported by a microstructural analysis would provide a much needed input in the field of electroplated nickel films to understand out-of-plane deformations encountered in surface micromachining. This work comprises such a study on stress gradients in Ni thin films, where the effect of uniform stresses is omitted in measurements.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
In the remainder of this work, a mechanical framework for the measurement of residual stress gradients will be discussed first followed by a description of employed experimental procedures using 1-lm-thick, 10-lm-wide and 100-lm-long cantilevers. Results are then presented leading to a useful set of deposition parameters for nickel sulfamate bath resulting in near-zero bending of Ni cantilevers. 2. Residual stress gradients and their characterization Residual stresses in microstructures can be modeled with the combination of a uniform stress (r) and a stress gradient across the thickness ðrrÞ. Uniform residual stresses are found in either tensile or compressive form leading to in-plane expansion or contraction of the microstructure after release. In contrast, the relief of a residual stress gradient induces a bending moment resulting in the formation of a curved microstructure with a strain gradient of opposite sign as shown in Fig. 2a. The residual stress gradient for a monolayer cantilever is given in Eq. (1) in terms of the modulus of elasticity (E) and tip deflection (d) of a cantilever of length L [10]. The assumption of linear stress distribution, i.e. constant stress gradient, is frequently employed in the literature [10,14,16–18], which is especially proper for very thin films such as the 1-lm-thick nickel film of this study:
rr ¼ E
2d
ð1Þ
L2
Hence, the residual stress gradient can be determined by measuring process-dependent parameters of tip deflection and modulus of elasticity. While tip deflection can be measured by optical means, modulus of elasticity is determined through measuring the first-mode resonance frequency, f1, of a cantilever. In case of a monolayer, homogeneous cantilever with a rectangular uniform cross-section, the relation between E and f1 is given by Eq. (2). This relation remains unaltered by the presence of residual stress gradients:
f 1 ¼ 0:162
h
sffiffiffiffi E
L2
q
ð2Þ
In Eq. (2) two new parameters, q and h, density and thickness of a cantilever, respectively, are introduced. Being a relatively process-independent parameter, q is taken as the bulk mass density of 8908 kg/m3 throughout this work [19]. Fig. 2b provides a case study for the verification of Eq. (1), where a finite element analysis (FEA) of a cantilever is carried out. COMSOL Multiphysics 4.3 FEA software with 3d free-tetrahedral mesh elements is employed to study the release of a cantilever with a
61
stress gradient of 1 MPa/lm. The thickness, width and length of the cantilever are chosen as 100 100 1000 lm, respectively. Fig. 2b depicts the linear variation of strain along the midline of a cross-section within the cantilever. Strain gradient from the analytical treatment deviates from the slope of the strain plot in Fig. 2b by less than 0.5%. It is not uncommon to encounter wafer curvature measurements employed in the literature for the characterization of residual stress gradients. There are reports claiming wafer curvature measurements encompass the effect of stress gradients as well as uniform stresses [20]. However, use of wafer curvature data might be misleading, as there are cases where infinite radius of curvature, i.e. straight substrate, does not guarantee straight cantilevers upon release. Similarly, finite wafer curvature associated with thin film stresses might lead – in a seemingly counterintuitive fashion – to a straight cantilever. To elucidate the misinterpretation associated with wafer curvature measurements, let us consider a <1 0 0> Si substrate of a diameter (D) of 10 mm and a thickness (T) of 500 lm. Let us also assume that a Ni film with a thickness (h) of 10 lm is coated on the top surface of this substrate. If there is a stress gradient of 20 MPa/ lm in the Ni film (Fig. 3a), it leads to a negligible wafer curvature ð1=qÞ of 8.864 104 m1. This can be computed by incorporating the resulting moment due to stress gradients into the moment equilibrium in a similar fashion to Stoney formulation. The associated formulation is given in Fig. 3a as well, and invariant biaxial modulus of <1 0 0> Si (E/(1m)) is taken as 180.51 GPa [21]. If this curvature measurement is misinterpreted as an indication of uniform stress only, it would correspond to a stress level of less than 1 MPa. However, if one fabricates and releases a Ni cantilever of a length (L) of 100 lm, this seemingly harmless effect leads to a considerable tip deflection of 0.5 lm. On the contrary, one can consider the same substrate/thin film system, where the thin film is under a uniform tensile stress of 100 MPa (Fig. 3b). Using Stoney formulation given in Fig. 3b, the associated wafer curvature is predicted to be 0.133 m1. This measurement is 150 times larger than the curvature in the presence of the aforementioned stress gradient case. More interestingly, as all associated stresses are uniform, they do not lead to any permanent bending of the cantilever upon release. Hence, a larger curvature does not necessarily mean a bending and stiction risk for surface micromachining. Aforementioned loading scenarios are verified using COMSOL as described in connection with Fig. 2 as well. They support the claim that one has to release microstructures for a reliable characterization of stress gradients, and Stoney formulation alone is not an efficient way for the characterization of surface micromachining.
Fig. 1. The effect of residual stress gradient, rr, on bending of a monolayer cantilever. (a) With compressive stresses increasing with increasing distance from thin filmsubstrate interface, the cantilever bends towards the substrate. (b) If residual stresses become more tensile away from the interface, the cantilever bends upwards.
62
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 2. (a) Stress gradient ðrrÞ in a film is converted into a strain gradient ðrÞ following the release process. (b) The finite element simulation of a microcantilever with a predefined stress gradient also confirms the same strain gradient upon release.
3. Experimental procedures 3.1. Microfabrication The employed fabrication flow shown in Fig. 4 is based on a single-mask lithographic process on Si. 400 -diameter, <1 0 0> Si wafers are chosen with a resistivity of 0.5–1.0 X cm and a thickness of 525 ± 25 lm. The process consists of RF magnetron sputtering, photolithography, nickel electroplating and release of microcantilevers. An RF magnetron sputtering system (Vaksis PVD-Handy/ 2S) is employed to deposit a blanket coating of about 20-nm-thick Cr and 100-nm-thick Au on Si wafers as an adhesion promoter and
seed layer, respectively. Then photolithography is carried out using AZ 1514H positive photoresist by Microchemicals. After defining molds for electroplating by photolithography, 400 -diameter Si wafers are cut into 1 cm 1 cm dice with a scriber (OEG MR200) to perform electroplating for each die separately. Nickel electroplating is performed with a nickel sulfamate solution consisting of 600 g/l nickel sulfamate (Ni(SO3NH2)24H2O), 10 g/l nickel chloride (NiCl26H2O) and 40 g/l boric acid (H3BO3). The acidity (pH) of the solution is adjusted from 5.2 to 4.0 by adding approximately 10 g of sulfamic acid powder. The release of microcantilevers is achieved by subsequent wet etching of exposed Au and Cr layers by using the etchants of
Fig. 3. Two scenarios for wafer curvature measurement for a 10-lm-thick Ni coating on a 500-lm-thick, <1 0 0> Si substrate. (a) In the presence of a stress gradient of 20 MPa/lm in Ni, a negligible curvature is obtained. However, release of a 100-lm-long Ni cantilever leads to a considerable tip deflection of 0.5 lm. (b) In the presence of a uniform stress of 100 MPa, a 150-times larger curvature results. However, no tip deflection will be encountered as all stresses in Ni are uniform.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
63
Fig. 4. Fabrication flow is composed of (a) wafer cleaning, Cr and Au layer deposition by RF magnetron sputtering, (b) lithography and electroplating of nickel, (c) removal of the photoresist (PR) and release of microcantilevers, and finally (d) etching of Cr and Au layers from underneath cantilevers and drying.
GE-8148 from Transene Inc. and CEP-200 from Microchrome Inc., respectively. This is followed by etching of Si substrate with a 45% KOH solution at 60 °C to a depth of approximately 30 lm. Final etching of Cr and Au layers from underneath released cantilevers provides monolayer nickel cantilevers. After the release of microcantilevers, each sample is immersed in water, acetone and IPA subsequently for at least 5 min and dried at room temperature on a flat surface to prevent stiction of microcantilevers. Further details regarding fabrication processes and equipment can be found elsewhere [2]. A wide electroplating operation range is chosen to investigate stress gradients: three current densities of 1, 5 and 20 mA/cm2 and three temperatures of 30, 45 and 60 °C. Thus there are 9 different cases studied separately with DC and pulse electroplating. Despite best efforts to reach uniform 1 lm nickel film thickness across samples, considerable film thickness non-uniformity – especially pronounced for the current density of 20 mA/cm2 – requires determination of exact cantilever thickness. Hence, 15 cantilevers across each sample are chosen to determine the exact film thickness. Three thickness measurements per cantilever are carried out using a surface profiler (Veeco, Dektak 8). The average value of the three thickness measurements is assigned as the thickness of the particular cantilever. Afterwards, tip deflections and resonance frequencies of these individual cantilevers are traced to quantify the corresponding stress gradient for each cantilever precisely. A DC power supply along with a multimeter are employed to perform the DC electroplating process. The pulse plating experiments are performed using a pe86cb type reverse pulse power
supply from Plating Electronic. In pulse electroplating, a squarewave pattern is chosen with a duty cycle of 0.5 at the frequency of 1 kHz. The mean current values are matched with DC electroplating values to ensure proper comparison between DC and pulse electroplating. Electroplating process is carried out using 1 cm 1 cm samples shown in Fig. 5. Contact areas for electroplating are visible at each corner of the die in Fig. 5b. Electroplating is performed using two alligator clips placed diagonally to provide symmetric film thickness distribution across the die. Cantilever anchors are aligned parallel to <1 1 0> to minimize the undercut of silicon supports during wet anisotropic etching of silicon. This alignment ensures definition of satisfactory and reproducible clamping conditions with minimal undercut for each cantilever. Thus, variation of cantilever properties due to effective length variation and poorly defined clamping conditions is substantially eliminated. 3.2. Static and dynamic measurements Deflections of released cantilevers are measured by utilizing white light interferometry (WLI) technique. WLI measurements performed with an optical profiler (Bruker ContourGT 3d) provide the resolution of deflection on the order of 1 nm. The modulus of elasticity of thin films is investigated by determining resonance frequencies of microcantilevers in air. Electromagnetic actuation with an electromagnetic coil and a set of permanent magnets [2] is carried out together with optical readout via laser doppler vibrometry (LDV) (Polytec OFV-2500 Vibrometer
Fig. 5. (a) Layout with 46 dice of 1 cm 1 cm dimensions. (b) Each die consists of 58 cantilevers along with letters and numbers for labeling each cantilever. All cantilevers are identical. (c) A close-up view of two cantilevers on a die. (d) This scanning electron micrograph verifies the complete release of the fabricated cantilever from base Si substrate. Tilted Si sidewall due to KOH etching can be observed.
64
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Table 1 A list of samples highlighting the matrix of process conditions utilized in this study. Die number A1 A2 A3 B1 B2 B3 C1 C2 C3
Deposition temperature (°C)
Deposition current density (mA/cm2)
Ni thickness (nm) DC plating
Ni thickness (nm) pulsed plating
30 30 30 45 45 45 60 60 60
1 5 20 1 5 20 1 5 20
969 ± 28 980 ± 90 1163 ± 211 931 ± 29 889 ± 103 1009 ± 135 1031 ± 34 1007 ± 82 1165 ± 171
1006 ± 37 900 ± 108 960 ± 259 1075 ± 20 1005 ± 123 1072 ± 281 976 ± 34 1062 ± 128 1186 ± 285
Controller) to determine resonance frequencies of microcantilevers whose deflection and thickness values were already recorded. Texture measurements are carried out with Bruker AXS Advance D8 with DaVinci system. CuKa source in point focus coupled with polycapillary optics is collimated to 1 mm with a circular collimator. A 1d detector (VANTEC-1) is used as a detector. Pole figures for Au and Ni are measured by recording diffraction by rotating the sample around the sample normal (0° < u < 360°) at each tilt angle (0° < v < 85°). MULTEX software by Bruker AXS is used to analyze pole figure measurements. Calculated pole figures with MULTEX are imported into MTEX [22], a MATLAB toolbox for quantitative texture analysis, for inverse pole figure calculations and plotting. Crystallite size of Ni films is estimated by using quantitative Rietveld analysis with TOPAS software supplied by Bruker-AXS by taking microstrain into account. Crystallite size is calculated by the software based on refinements realized by introducing Ni, Au and Si phases into the software. Furthermore, cross-sections of thin films are examined by scanning electron microscopy (Zeiss Ultra Plus Field Emission SEM with Bruker XFlash 5010 SDD Detector). 4. Results A set of nine different deposition conditions is utilized throughout the study as summarized in Table 1 and fabricated separately
Fig. 6. The current density of 20 mA/cm2 results in excessive thickness nonuniformity across die A3. The corners with electrical contacts are indicated with ‘‘C’’. All numbers are reported in nm.
Table 2 The comparison of cantilever tip deflection measurements between DC and pulse plated samples implies that there is no significant difference. Die number
DC Plating (lm)
Pulse plating (lm)
A1 A2 A3 B1 B2 B3 C1 C2 C3
4.53 ± 0.74 0.21 ± 0.84 1.40 ± 0.55 4.40 ± 0.39 2.65 ± 0.51 2.12 ± 0.75 0.83 ± 0.23 3.45 ± 0.33 2.03 ± 0.51
3.17 ± 0.59 1.36 ± 0.38 – 4.29 ± 1.10 2.94 ± 0.48 1.70 ± 0.66 0.56 ± 0.24 2.70 ± 0.79 1.71 ± 0.46
with DC and pulse electroplating. 15 cantilevers in each die are utilized for the determination of nickel film thickness, tip deflection and modulus of elasticity to characterize residual stress gradients. Nickel film thickness distribution is also demonstrated in Table 1. There is a noteworthy thickness variation for samples fabricated with a 20 mA/cm2 current density. This variation is reduced significantly in the case of 1 mA/cm2. As an example, the thickness variation for sample A3 of Table 1 is depicted via a contour plot in Fig. 6, where the distribution is observed to be symmetric with respect to the two contact points designated with C. Film thickness decreases toward the two contact points, while it increases toward the remaining two corners, thereby resulting in a saddle-type distribution throughout the die. This thickness distribution indicates the necessity to track individual cantilevers before and after release rather than assigning an average cantilever thickness to the whole die. Table 2 lists resulting tip deflections after release. Since DC and pulse plating lead to similar cantilever tip deflections for most of the cases, DC plated samples are chosen for subsequent stress gradient measurements. Upon release, cantilever tip deflections are found to be upwards or downwards depending on employed process parameters. Combinations of different current densities and electrolyte temperatures demonstrate a transition between upward and downward deflections as shown in Fig. 7. Adopting the sign convention in Fig. 2, deflections of upward bent cantilevers will be taken as positive. Moduli of elasticity of cantilevers are reported in Fig. 8. The mean modulus of elasticity of nickel is found to be in the range of 128.2–184.4 GPa, which is smaller than the modulus of elasticity of bulk nickel of 219 GPa [23]. Having the moduli of elasticity and cantilever tip deflections characterized, it becomes now possible to link stress gradients to electroplating parameters via Eq. (1). Thus as the prime outcome of this study, Fig. 9 summarizes the residual stress gradient variation within the current density and electrolyte temperature ranges of 1–20 mA/cm2 and 30–60 °C, respectively. The stress gradient is found to vary in the range of 126.5 to +65.6 MPa/lm. What is interesting about the results is the fact that obtaining straight
Fig. 7. A plot of cantilever tip deflections. The trend shows that obtaining a straight cantilever is possible for a certain combination of deposition parameters.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 8. The modulus of elasticity measurements with respect to electroplating temperature and current density.
Fig. 9. Stress gradients with respect to the electroplating temperature and current density. Numbers I–IV refer to blanked-coated Ni films for microstructure characterization experiments.
cantilevers after release is possible with a suitable choice of deposition parameters. For example, current densities of approximately 14 mA/cm2 and 5 mA/cm2 result in stress-gradient-free, straight beams for plating temperatures of 45 °C and 30 °C, respectively, without the use of any stress-reducing agents. Additional Ni film samples are electroplated with fabrication parameters designated with I–IV in Fig. 9 to evaluate the effects of the current density and electrolyte temperature on film internal structure by performing texture and grain size measurements. Diffractions from Si substrate, Au seed layer and Ni film are seen on the XRD patterns in Fig. 10. Crystallite size is estimated for the nickel films from I to IV as 30.9 nm, 46.1 nm, 53.2 nm and 46 nm, respectively.
5. Discussion As shown in Table 2, DC and pulse plating follow the same trend in terms of cantilever tip deflection. Similarity between DC- and pulse-plated cantilevers deposited with the same current density in terms of internal structure and residual stress is also
Fig. 10. XRD patterns of approximately 1-lm-thick films electroplated at conditions shown in Fig. 9. For better visualization of differences, an offset is introduced between plots in vertical direction.
65
emphasized in the literature [24]. Thus, stress gradient characterization is performed only for DC-plated cantilevers. Since properties of electroplated films are highly sensitive to deposition conditions such as current density, plating temperature, pH and chemistry of the electrolyte and film thickness, it is quite difficult to find an identical study in the literature for comparison of the results of an experiment. However, trends showing relationships between two independent parameters can still be extracted as explained below. The effect of electroplating conditions on the modulus of elasticity is investigated thoroughly in several studies [25]. Variation of modulus of elasticity was attributed to changes in the film internal structure by the introduction of non-metallic species into deposited films by hydrolysis with increasing electroplating temperature and shifting from the instantaneous nucleation and growth mechanism to progressive nucleation mechanism by increasing current density. According to the stress gradient results in Fig. 9, there is a transition from negative to positive stress gradient at electroplating temperatures of 30 and 45 °C. This transition is expected to be related to the film internal structure. Effects of the electrolyte temperature and current density on the film internal structure are investigated systematically by studying cross sections of electroplated films. As shown in Fig. 11a, cross-sections of electroplated nickel films are formed as follows: Coating and development of a photoresist (PR) layer on a silicon substrate with an RF-magnetron-sputtered Au seed layer continues with growth of nickel films electroplated at operating conditions shown in Fig. 9 with numbers I–IV. Then, the PR layer is removed to allow SEM examination of the cross section of deposited nickel films. Fig. 11b demonstrates SEM micrographs of the cross sections of 1-lm-thick films fabricated using designated conditions. In I, low current density results in a needle-like internal structure. However, the current density of 20 mA/cm2 and electroplating temperature of 30 °C lead to somewhat strongly textured structure as shown in III. The cross section shown in IV seems to have smoother surface compared with the surface shown in III. Blanket-coated nickel film samples employed for the SEM crosssection evaluation are also used for texture and grain size measurements. Several XRD studies [24,26] state that the texture of nickel films deposited on <1 1 1> oriented gold changes from <1 1 0> orientation at low current densities to <1 0 0> texture by increasing current density. Amblard et al. [26] explained this shift with a specific inhibition mechanism correlating to adsorbed hydrogen for a nickel sulfate based electrolyte. In this study, the crystallographic orientation of the polycrystalline Au seed layer evaluated before the texture of Ni films indicates that the preferred orientation is indeed in <1 1 1> direction with fiber texture and {1 1 1} planes parallel to substrate surface. As seen in Fig. 10, (1 1 1) peak of Ni overlaps with (2 0 0) peak of Au. Although Au layer shows strong <1 1 1> orientation and the contribution of its (2 0 0) peak at 2h = 44.5° might be low, (2 0 0), (2 2 0) and (3 1 1) peaks of Ni are employed for the estimation of the preferred orientation. All of the analyzed nickel films possess fiber texture in normal direction to the sample surface and corresponding planes are found to be parallel to the sample surface. Contrary to theoretical predictions for electroplated Ni films, all films (I–IV in Fig. 9) show a preferred orientation mainly in <5 1 1> crystal direction. A very weak <1 1 1> texture is found to develop for all films electroplated at high current density (II–IV in Fig. 9). In all cases, grain sizes of microstructures are on the order of a few tens of nanometers. Deposition at high current density (II–IV in Fig. 9) results in larger crystallite sizes of approximately 50 nm. However, the effect of deposition temperature on the crystallite size is not as significant as the current density. Dependence of the crystallite size on the current density can be explained by
66
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
Fig. 11. (a) Preparation of cross-sections for scanning electron microscopy evaluation. (b) SEM micrographs of film cross-sections. Each letter corresponds to a specific fabrication condition given in Fig. 9.
inhibited nickel film growth as a result of hydrogen evolution on the cathode pronounced at low current densities [27,28]. Fritz et al. [24] analyzed cross-sections of electroplated thick (15 lm) films and found that columnar internal structure formed at low current densities turns into granular structure as employed current density increases. However, in this study, where we address much thinner films, there is no such transition and internal structure seems to be granular for all cases. Temporal evolution of stress gradients under room conditions is also studied by tracking tip displacements of various cantilevers 6 and 15 months after the release. Cantilevers were kept in closed containers under ambient conditions throughout this time. Creep effect due to residual stresses, thermal loadings due to mean ambient temperature shifts, hydrogen desorption from a deposited film with time are held responsible for temporal changes [29]. In this study, shifts in tip deflection are found to be negligible over a period of 15 months. This observation underlines employment of single-layer instead of multilayer cantilevers to obtain time-independent tip deflections which improve reliability of sensors. 6. Conclusions Nickel electroplating is a popular technique for the fabrication of micromechanical structures. With miniaturized structures finding more and more applications, a systematic stress gradient
characterization of micrometer-thick nickel films with electroplating parameters is needed. This study is intended to provide a guideline for recipe formation aiming at straight nickel microstructures. For this purpose microcantilevers are used as transducers for residual stress gradient measurement. Residual stress gradient is found to vary in the range of about 130 to 70 MPa/lm, where a transition from downward to upward deflection is observed. Hence, a window of electroplating parameters is established yielding zero residual stress gradients, i.e. straight cantilevers. For the deposition temperature of 30 °C, a current density of about 5 mA/ cm2 is needed, whereas a current density of approximately 14 mA/cm2 leads to straight cantilevers at 45 °C. No such condition was obtained at 60 °C. Furthermore, stress gradients are observed to exhibit negligible change 15 months after first measurements highlighting improved reliability of sensors made of monolayer cantilevers. Acknowledgements Authors thank Prof. Hakan Urey for access to resonance frequency measurement setup. Contributions by Erhan Ermek and Zuhal Tasdemir during characterization experiments are acknowledged. Discussions on electroplating and pulse plating by Prof. Levent Demirel, Dr. Annamaria Miko and Prof. Ozgur Birer are also gratefully acknowledged. Funding was provided by Tubitak under Grant no. 109E222 and 111E184.
Y. Kilinc et al. / Microelectronic Engineering 134 (2015) 60–67
References [1] T. Itoh, S. Kawamura, K. Kataoka, T. Suga, Sens. Actuators, A 123–124 (2005) 490. [2] A. Ozturk, H.I. Ocakli, N. Ozber, H. Urey, I.H. Kavakli, B.E. Alaca, Photonics Technology Letters, IEEE 20/23 (2008) 1905. [3] E. Timurdogan, N. Ozber, S. Nargul, S. Yavuz, M.S. Kilic, I.H. Kavakli, H. Urey, B.E. Alaca, Biosens. Bioelectron. 26 (1) (2010) 195. [4] O. Cakmak, C. Elbuken, E. Ermek, A. Mostafazadeh, I. Baris, B. Erdem Alaca, I.H. Kavakli, H. Urey, Methods 63 (3) (2013) 225. [5] E. Uzunlar, B. Beykal, K. Ehrlich, D. Sanli, A. Jonáš, B.E. Alaca, A. Kiraz, H. Urey, C. Erkey, J. Supercrit. Fluids 81 (2013) 254. [6] H. Majjad, S. Basrour, P. Delobelle, M. Schmidt, Sens. Actuators, A 74 (1–3) (1999) 148. [7] J. Zhang, J. Mater. Process. Technol. 123 (3) (2002) 329. [8] Y.-H. Zhang, G.-F. Ding, Y.-L. Cai, H. Wang, B. Cai, Mater. Charact. 57 (2) (2006) 121. [9] Y.B. Wu, G.F. Ding, J. Wang, C.C. Zhang, H. Wang, Microelectron. Eng. 87 (11) (2010) 2035. [10] C.W. Baek, Y.-K. Kim, Y. Ahn, Y.-H. Kim, Sens. Actuators A 117 (2005) 17. [11] Z.M. Zhou, Y. Zhou, H. Mao, J. Mater. Sci. Technol. 25 (1) (2009) 77. [12] C. Genzel, Int. Centre Diffr. Data 44 (2001). [13] A. Atkinson, S.C. Jain, J. Raman Spectrosc. 30 (10) (1999) 885.
67
[14] H.-H. Yang, M.-H. Seo, C.-H. Han, J.-B. Yoon, J. Micromech. Microeng. 23 (5) (2013) 055016. [15] S. He, J.S. Chang, L. Li, H. Ho, Sens. Actuators, A 154 (1) (2009) 149. [16] A.K. Chinthakindi, D. Bhusari, B.P. Dusch, J. Musolf, B.A. Willemsen, E. Prophet, M. Roberson, P.A. Kohl, J. Electrochem. Soc. 149 (8) (2002) H139. [17] W. Fang, J.A. Wickert, J. Micromech. Microeng. 6 (3) (1996) 301. [18] V. Mulloni, F. Giacomozzi, B. Margesin, Sens. Actuators, A 162 (1) (2010) 93. [19] T. Fritz, M. Griepentrog, W. Mokwa, U. Schnakenberg, Electrochim. Acta 48 (20) (2003) 3029. [20] J. Luo, M. Pritschow, A. Flewitt, S. Spearing, N. Fleck, W. Milne, J. Electrochem. Soc. 153 (10) (2006) D155. [21] W.A. Brantley, J. Appl. Phys. 44 (1) (1973) 534. [22] F. Bachmann, R. Hielscher, H. Schaeben, Solid State Phenom. 160 (2010) 63. [23] W.N. Sharpe Jr., N. William, Mechanical Properties of MEMS Material, CRC Press, 2002. [24] T. Fritz, W. Mokwa, U. Schnakenberg, Electrochim. Acta 47 (1) (2001) 55. [25] J.K. Luo, A.J. Flewitt, S.M. Spearing, N.A. Fleck, W.I. Milne, Mater. Lett. 58 (17– 18) (2004) 2306. [26] J. Amblard, I. Epelboin, M. Froment, G. Maurin, J. Appl. Electrochem. 9 (2) (1979) 233. [27] M. Vogelaere, V. Sommer, H. Sprigborn, U. Mishelsen-Mohammadein, Electrochim. Acta 47 (2001) 55. [28] K.C. Chan, N.S. Qu, D. Zhu, Text. Microstruct. 30 (1-2) (1997) 43. [29] W. Fang, H.-C. Tsai, C.-Y. Lo, Sens. Actuators, A 77 (1) (1999) 21.