Home

Search

Collections

Journals

About

Contact us

My IOPscience

Electrodeposited Co-Ni alloys for MEMS

This content has been downloaded from IOPscience. Please scroll down to see the full text. 2002 J. Micromech. Microeng. 12 400 (http://iopscience.iop.org/0960-1317/12/4/309) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 141.28.6.190 This content was downloaded on 13/04/2014 at 20:44

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING J. Micromech. Microeng. 12 (2002) 400405

JOURNAL OF MICROMECHANICS AND MICROENGINEERING PII: S0960-1317(02)32918-8

Electrodeposited Co-Ni alloys for MEMS

M Duch1, J Esteve1, E G omez2, R P erez-Castillejos1 2 and E Vall es
Department of Microsystems, Centre Nacional de Microelectr` onica (CNM-IMB), Campus Universitat Aut` onoma de Barcelona, Bellaterra E-08193, Barcelona, Spain 2 Laboratori de Ci` encia i Tecnologia Electroqu mica de Materials (LCTEM), Departament de Qu mica f sica, Universitat de Barcelona, Mart i Franqu` es 1, E-08028 Barcelona, Spain E-mail: Marta.Duch@cnm.es
1

Received 22 January 2002, in nal form 2 May 2002 Published 19 June 2002 Online at stacks.iop.org/JMM/12/400 Abstract The electrodeposition process has been studied and optimized in order to obtain homogeneous Co-Ni deposits on Si/SiO2/Ti/Ni substrates. The electrochemical and magnetic characterization of the deposited layers shows the features of a soft magnetic material, i.e. high saturation magnetization (1.2 T) and low coercivity, which may be varied as a function of the electrodeposition parameters. Good adherence, brilliant aspect, smooth surfaces and a constant deposition rate have been obtained in all the processed samples. The compatibility of the Co-Ni plating process with the main microelectromechanical systems (MEMS) fabrication techniques, surface and bulk, has been evaluated and these results widely demonstrate the versatility of this magnetic layer in MEMS production. Due to the high selectivity and homogeneity of the deposition process, it has been possible to pattern the deposits with a denition down to 3 m and an aspect ratio of 4.6 has been achieved. A novel method to liberate the patterned structures by sacricial etching of the seed layer has been developed. Uniform and homogeneous deposits of Co-Ni alloy have also been obtained in devices fabricated by bulk technologies. For these devices an anisotropic wet etching procedure in tetramethyl ammonium hydroxide (TMAH) has been optimized to release the magnetic layer. Low stressed free-standing structures have been obtained by both surface and bulk methods and are presented in this paper.

1. Introduction
Within the eld of microelectromechanical systems (MEMS) fabrication, the thermal and the electrostatic effects are the most widely used principles of actuation. However, other alternatives have been proposed [1], and among them, the actuators driven by magnetic forces are characterized by generating large output forces and large displacements [2, 3] in an energy-efcient way [4, 5]. For our applications, three main attributes dene the magnetic lm required. First, it must be a soft magnetic material which is characterized by a maximum saturation magnetization and a minimum remanence or coercivity. Second, in order to reduce the stress the magnetic layer brings to the actuator structure, thin lm technology is preferred. And third, the magnetic lm has to be compatible with standard
0960-1317/02/040400+06$30.00 2002 IOP Publishing Ltd

MEMS fabrication technology, granting its inclusion in batch processes. It must be observed that in spite of the various proposals found in the literature to make MEMS magnetically responsive, most of these techniques are not truly compatible with a batch fabrication process: for instance, microactuators with glued miniaturized magnets [6] and with thick or thin [7, 8] magnetic lms have been presented. To t this requirement, electrodeposition was found to be the best choice as it allows tailoring the layer characteristics by controlling the plating bath parameters. Complex magnetic alloys that consist of three or even four elements are under study and results show high moment response [9, 10]. However, due to considerations of bath and plating simplicity, the most employed magnetic material in MEMS fabrication is the Ni80Fe20 Permalloy [11]. On the other hand, Co-Ni alloys offer the possibility of shifting the magnetic 400

Printed in the UK

Electrodeposited Co-Ni alloys for MEMS

Table 1. List of the optimum range of operation for all the Co-Ni bath parameters. NiCl2 concentration 0.70.9 mol dm
3

CoCl2 concentration 0.20.3 mol dm

3

H3BO3 concentration 2040 g dm

3

Saccharin concentration 0.51.2 g dm

3

Temperature 5060 C

pH 33.5

Current Density (mA /cm2)

characteristics from para- to ferromagnetic by just increasing the Co content of the layer [12], which can be of main interest in the fabrication of magnetic microactuators. As a consequence of the previous considerations, an electrodeposited Co-Ni thin lm was developed and optimized for MEMS fabrication processes. Demonstrating prototypes for using the Co-Ni alloy in surface and bulk technologies were fabricated. The results that demonstrate the versatility of the presented electrodeposition process for silicon-based microactuators are provided in this paper.

10 5 0 -5 -10 -15

2. Electrodeposition process and characterization of the deposits

2.1. Electrodeposition process A simple chloride bath was developed and optimized for p-type 440 cm 100 silicon wafers [13] in order to obtain homogeneous Co-Ni deposits. Surfaces were prepared for better adhesion and minimum mechanical stress by seed layer on a previously sputtering a Ti/Ni (1000/ 500 A) grown phosphosilicate glass (PSG) layer 1 m thick. The electrodeposition process was performed in a three-electrode cell where no external magnetic eld was applied, and a potenciostat/galvanostat (EG&G 273) was used. All potentials were referred to a Ag/AgCl reference electrode containing 1 mol dm3 NaCl, which was mounted in a Luggin capillary. As the counter electrode, a nickel sheet was used. The bath was prepared from NiCl2 6H2O and CoCl2 6H2 O, boric acid and saccharin. An exhaustive study of the parameters affecting the electrodeposition procedure was carried out [14] and as a consequence, optimum operation ranges were determined for all parameters, which are shown in table 1. Co-Ni deposition is anomalous-type, i.e. preferential deposition of the less noble alloy component, Co in our case, occurs. Boric acid was added to the bath in order to control the pH of the solution and improve the quality of the deposit. Saccharin is a well-known plating additive that lowers the stress of the electrodeposited lm. As it was found [14], saccharin content in the alloy superior to 0.9 g dm3 led to no relevant effect, which agrees with other recent studies [15]. To avoid hydrogen evolution, pH has to be carefully controlled in the electrodeposition process. The pH of each solution was measured before and after deposition and no deviation was observed from the initial value. The temperature of deposition was chosen to be in the range of 5060 C in order to attain a sufcient deposition rate as well as a good appearance for the deposited lms. Once the bath parameters were established, the electrodeposition took place under potentiostatic conditions at 750 to 850 mV. Gentle stirring of the solution was maintained during deposition by using a moderate ow of Ar gas.

-800

-600

-400

-200

200

Reference Potential (mV)

Figure 1. Voltammetric curves of a 0.9 mol dm3 NiCl2 + 0.2 mol dm3 CoCl2 + 30 g dm3 H3BO3 + 0.7 g dm3 saccharin solution on a Si/SiO2/Ti/Ni substrate: pH = 3, T = 55 C and V = 50 mV s1.

Several methods were used for the study of the electrodeposition process. In the voltammetric curves (gure 1), typical features for a nucleation and growth process are shown, and only one oxidation peak is detected during the positive scan, corresponding to the alloy oxidation [16]. The j-t transients obtained from potentiostatic deposition, in the voltage range from 750 to 850 mV, show a gradual increase in current with time, which evolves into a constant current after 4 min. 2.2. Characterization of the deposit The general characterization of the deposits was performed with samples containing 6070% Co, grown from a bath composition of 0.2 mol dm3 CoCl2 + 0.9 mol dm3 NiCl2. The Co-Ni lms electrodeposited under these conditions (table 1) were homogeneous, compact and ne-grained, thanks to the fact that the seed layer favours a high nucleus density, leading to a smoother plating base. The rms-average surface roughness of the samples was quantied to be 0.04 m (data from interferometric surface analysis microscope, Zygos New View 100). No relevant variation was observed from samples with different layer thicknesses. The deposition rate was calculated from the average thickness of the deposit at different deposition times (gure 2). In this case, a Mechanical Perlometer measured the thickness of the layers (Tencor, Alpha-step 200, 5 nm resolution). For the selected plating conditions, the obtained deposition rate was 0.8 m min1. X-ray diffraction phase analysis (Siemens D-500 diffractometer) showed that the deposited Co-Ni alloy corresponded to a face-centred cubic solid solution [14]. On the other hand, data obtained from x-ray photoelectron spectroscopy (XPS, PHI 5600 multitechnique system) 401

M Duch et al

Co-Ni Thickness (m)

1 .5
20 15 10 5 0 0 200 400 600 800 1000

M, Specific Magnetization (T)

1 .0 0 .5 0 .0 -0 .5 -1 .0 -1 .5

Time (s)
Figure 2. Dependence of the thickness of the Co-Ni layer on deposition time, obtained at 800 mV during 20 min under the same conditions as in gure 1.

t = 3600s t = 1000s

conrmed that, under stirring conditions, a constant alloy composition, throughout the thickness of the deposit, was obtained. Magnetic characterization of the Co-Ni lms was performed using a pendulum-type magnetometer (Manics DSM 8) at room temperature. The samples were processed to eliminate the silicon substrate and the PSG layer, thus only the Co-Ni alloy, and the Ni/Ti layers remained. Then the samples were triturated and measured under a magnetic eld coplanar to the sample tray, that is coplanar to the sample. Results demonstrate that the electrodeposited alloy behaves as a soft magnetic material, with low coercivity (HC) and high saturation magnetization (MS). For the operation ranges of the plating bath listed in table 1, the measured average coercivity was 4 kA m1 (54 Oe) and the average saturation magnetization was 1.2 T (995.6 emu cm3). The volume of the samples is required in order to determine their magnetization. The procedure chosen here was weighing each sample on a precision mass scale (10 g resolution) and then multiplying the result by the density. Since Ni and Co have almost the same density, varying the composition of the alloy does not lead to density variations. It must be observed that no magnetic eld was applied during deposition in order to minimize the remanence of the lms, as required for soft magnetic lms. In gure 3 representative examples of the magnetization curves obtained from the studied Co-Ni lms are shown. The only difference between the samples was the electrodeposition time, since both were plated from a 0.9 mol dm3 NiCl2 + 0.2 mol dm3 CoCl2 solution at 800 mV and 55 C, to obtain a 63% Co:37% Ni alloy composition. It can be observed that the slope of the magnetization curve is related to the magnetic permeability and decreases as the electrodeposition time increases. This effect may be due to a reduction of the effective magnetic eld, as the demagnetizing eld increases with thickness. The dependence of the magnetic characteristics on the plating parameters in table 1 was studied. The saturation magnetization was found to remain almost constant for all the deposition parameters within the optimum ranges of operation. Coercivity exhibited strong dependence on the Co content, but no relevant dependence on any of the other bath parameters. As shown in gure 4, coercivity diminishes when the percentage of Co in the alloy decreases, allowing a minimization of this 402

-1 5 0

-7 5

75

150

H, Magnetic Field (kA/m)

Figure 3. Magnetization curves of Co-Ni lms electrodeposited from 0.9 mol dm3 NiCl2 + 0.2 mol dm3 CoCl2 solution at pH = 3, 50 C and 800 mV for electrodeposition times of 1000 s and 3600 s.
8

H (kA /m )

6 4 2 0 60 65 70 75 80 85 90 95

%Co
Figure 4. Evolution of the coercivity parameter (HC) as a function of the Co content in the deposited alloy.

parameter. In the experiments carried out, values for the coercivity down to 1.07 kA m1 (13.5 Oe) were obtained, that is a decrease of 84% from its higher value.

3. Applications of electrodeposited Co-Ni alloys in MEMS technology

To this point, the Co-Ni electrodeposition process had been studied and the lms plated on at silicon-based substrates already characterized. The next goal was to evaluate the versatility of Co-Ni lms when used for the fabrication of MEMS. First of all, a bath composition and its plating conditions were chosen as the standard for further studies of the Co-Ni alloy. In terms of its adherence to the substrate, and its magnetic response, the 60% Co: 40% Ni composition alloy was chosen as the optimum Co-Ni lm for our applications. The standard electrodeposition process for this layer was carried out from a 0.2 mol dm3 CoCl2 + 0.9 mol dm3 NiCl2 + 30 g dm3 H3BO3 + 0.7 g dm3 saccharin bath at 800 mV, pH = 3 and 55 C. As most of the Si-based MEMS devices fabricated today are based on planar technology, bulk technology or a combination of both, the compatibility of the electrodeposited

Electrodeposited Co-Ni alloys for MEMS

Table 2. Details of the photolithographic process for the three photoresists used. Photoresist HPRI 6512 (OHC) MAP 1275 (Microresist Technologies) AZ 4562 (Hoestch) Soft bake 100 C 50 s 100 C 30 min 100 C 30 min

Exposure 7 mW cm 8s 75 mW cm2 45 s 305 W 28 s

2

Post expbake 110 C 50 s

Developer HPRD 428 (OHC) HPRD 428 (OHC) 351 B

Hard bake 115 C 30 min 100 C 45 min 80 C 2h

Thickness 2 m 6 m 17 m

Co-Ni alloy with MEMS fabrication technologies was investigated for each of the two mentioned standard methods. Novel etching techniques for Co-Ni lms were developed that make compatible the magnetic alloy with the two main MEMS fabrication technologies. 3.1. Planar technology This method consists of an UV photoresist patterning procedure followed by an electrodeposition process. The photolithographic process was carried out on a silicon substrate where a Ti/Ni seed layer had been sputtered. Prior to the coating process, a soft bake was carried out in order to stabilize the substrate. Three different photoresists (HPRI 6512 by OHC, MAP 1275 by Microresist Technologies and AZ 4562 by Hoestch) were used to get photoresist microstructures of several heights, as the thickness of the deposit depends on the height of the photoresist walls. The details of the photolithographic process for the three photoresists are given in table 2. Once the photoresist structures were dened, the electrodeposition process was performed at standard conditions, on small samples (50 mm 200 mm) as well as on 4 inch wafers. Co-Ni layers up to 14 m thick were achieved when the AZ 4562 photoresist was used, as the thickness of the plated layer is limited by the thickness of the photoresist. After the electrodeposition step, photoresist was removed by acetone. In the fabrication of MEMS devices, the capability of freeing, partially or totally, the magnetic layer is as important as the ability to pattern the Co-Ni layer. For this aim, a new procedure was developed: a Ni seed layer was used as a sacricial layer and was etched by a diluted nitric acid solution (HNO3/H2O 1:4). The studies carried out show that this solution has its optimum working point 30 min after being min1. For this working mixed, where the etch rate is 500 A point the etching times are very short, allowing the removal of the Ni seed layer without visibly damaging the Co-Ni deposit. 3.2. Bulk technology According to the promising results achieved in the planar technology, the next aim was to investigate the possibility of maintaining these features in bulk technologies. In this case, test devices were fabricated electrodepositing a Co-Ni lm on a silicon surface previously micromachined by anisotropic wet etching. The rst step consisted of an anisotropic wet etching in tetramethyl ammonium hydroxide (TMAH) solution from the back side of the wafer so as to get silicon membranes. The TMAH etching was carried out at 80 C, and a nitrogen stirring system was applied to avoid variations in local

concentrations that can produce very rough surfaces or even the appearance of hillocks [17]. Furthermore, a water-cooling system was included to reduce changes in concentration due to evaporation. Following this procedure, 15 m thick silicon membranes were fabricated, although more complex structures can be obtained by processing the front side of the wafer. As an example, some new test devices were developed in order to study the behaviour of the Co-Ni magnetic layer in front of an external magnetic eld. These devices were fabricated by patterning the front side of the membranes through dry etching (RIE) techniques until a cantilever was obtained. Once the silicon structures had been fabricated, a Ti/Ni /500 A) lm was sputtered into the cavities to act (1000 A as the seed layer during the plating process. Afterwards, standard Co-Ni coatings were potentiostatically deposited for different deposition times producing membranes of 2, 4 and 6 m thickness. For some applications it may be important to obtain standalone Co-Ni structures, that is, to be able to remove silicon from the area of interest. For instance, in the fabrication of the test devices, the next step was to remove the 15 m silicon layer over the Co-Ni lm (front side of the wafer) to obtain free Co-Ni membranes. This process was carried out using a 25% TMAH solution at the same conditions as the anisotropic etching reported before. Under these conditions, the silicon was completely removed at an etch rate of 25 m h1. No damage was produced on the Co-Ni surface during the silicon etching. Low roughness and minimum stress were obtained in the free Co-Ni cantilevers. Although it would be possible, in this case the seed layer was not removed since its small thickness did not change signicantly the magnetic or the mechanical properties of the Co-Ni lm.

4. Results
Good adherence, brilliant aspect, smooth nal surfaces and rather constant deposition rates have been obtained for all the samples, independently of how the substrates were processed. 4.1. Planar technology In the devices fabricated by the surface method detailed in section 3.1, a high selectivity was obtained in all the samples, independent of the photoresist that was used. As can be seen in the SEM photograph (gure 5), microstructures with dimensions down to 3 m have been achieved. The best aspect ratio that has been obtained is 4.6 and belongs to the samples processed with the thickest photoresist (AZ 4562). The electrodeposition of Co-Ni alloys was performed on small samples as well as on 4 inch wafers. In both cases an 403

M Duch et al

Figure 5. SEM image of 3 m wide structures of a 60% Co: 40% Ni alloy. Figure 7. SEM image of a detail of the edge effect at the cavity corner.

(a)

Figure 6. Free-standing Co-Ni cantilever with the same composition as in gure 5.

(b) (c)

almost constant deposition rate was observed which coincided with the velocity recorded for simple at substrates (gure 2), where photoresist was not used to shape the deposited alloy. When entire wafers were plated, a thickness inhomogeneity of around 12% between the centre (higher) of the sample and the rim (lower) was observed. In gure 6, an example of a device processed to partially free the Co-Ni structure is shown. Due to the geometry of the device only the thinnest parts were released, producing two stand-alone cantilevers xed to the substrate by one end through the mass of the device. The etching process is stopped by time control, since the etching velocity is known (see section 3.1). It is also noticeable from the picture that no relevant stress is observed in the released cantilevers. 4.2. Bulk technology In spite of the distance between the top and the bottom of the cavity, where the electrodeposition took place, more than 300 m, an almost constant deposition rate of 0.8 m min1 404

Figure 8. Photographs of Co-Ni deposit obtained from the same solution and temperature as in gure 5. (a) View of the sample showing the front side and the back side of the free-standing Co-Ni structure. (b) and (c) Samples of a video showing the bending effect induced by a permanent magnet on a 2 m thick Co-Ni lm/1500 A Ti-Ni/8 m thick silicon.

was observed for all the samples on the membrane. However, a thickness inhomogeneity can be observed at the corners of the membrane (gure 7) due to the edge effect [18]. The test devices developed to study the response of the Co-Ni alloy when an external magnetic eld is applied are shown in gure 8(a). The body of the cantilever (7.1 mm Ti/Ni 4.1 mm) was formed on 8 m of silicon, 1500 A and nally 2 m of the Co-Ni lm with a 60% Co: 40% Ni composition. During the experiments, the external magnetic eld was generated by a Nd-Fe-B magnet (MS = 1.2 T, HC = 890 kA m1) and a high response of the cantilever was observed (gures 8(b)(c)). High adherence, robustness

Electrodeposited Co-Ni alloys for MEMS

and exibility of the deposited layer must also be noticed. In spite of the large movement produced by the magnet, no damage was observed in the Co-Ni deposit even after more than 50 cycles of displacement upwards and downwards.

References
[1] Muralt P 1999 MME99 Proc. (Gyf sur Ivette, France) [2] Zou J, Chen J and Liu Ch 2001 Transducers01 Proc. (Munich, Germany 1582) [3] Judy J W, Muller R S and Zappe H H 1995 J. Microelectromech. Syst. 4 162 [4] Busch-Vishniac I J 1992 Sensors Actuators A 33 207 [5] Nami Z, Anh C H and Allen M G 1996 J. Micromech. Microeng. 6 337 [6] Wagner B, Benecke W, Engelmann G and Simon J 1992 Sensors Actuators A 32 598 [7] Lagorce L K and Allen M G 1996 MEMS96 Proc. (San Diego, CA) [8] Liakopoulos T M, Zhang W and Ahn C H 1996 IEEE Trans. Magn. 32 5154 [9] Osaka T, Takai M, Hayashi K, Ohashi K, Saito M and Yamada K 1998 Nature 392 796 [10] Rasmussen F E, Ravnkilde J T, Tang P T, Hansen O and Bouwstra S 2000 Eurosensors 14th Proc. (Copenhagen, Denmark, 915) [11] Ruan M, Shen J and Wheeler C B 2001 MEMS01 Proc. (Interlaken, Switzerland) [12] L ochel B and Maciossek A 1996 J. Electrochem. Soc. 143 3343 [13] G omez E and Vall es E 1999 J. Appl. Electrochem. 29 805 [14] Duch M, Esteve J, G omez E, P erez-Castillejos R and Vall es E 2002 Development and characterization of Co-Ni alloys for microsystems applications J. Electrochem. Soc. 149 C201 [15] Yang H H, Myung N V, Yee J, Park D-Y, Yoo B-Y, Schwartz M, Nobe K and Judy J W 2001 Ferromagnetic micromechanical magnetometers Transducers01 (Munich, 2001) [16] G omez E, Ram rez J and Vall es E 1998 J. Appl. Electrochem. 28 71 [17] Merlos A, Acero M, Bao M H, Bausells J and Esteve J 1993 Sensors Actuators A 37-38 73743 [18] Shenoy R V and Datta M 1996 J. Electrochem. Soc. 143 544

5. Conclusions
A simple bath from chloride salts, boric acid and saccharin was developed by optimizing the plating conditions required to obtain homogeneous Co-Ni alloys over Si/SiO2/Ti/Ni substrates. The obtained layers show the features of a soft magnetic material with a high saturation magnetization (1.2 T) and small values of coercivity, which can be further decreased by reducing the content of Co. Two methods have been successfully developed in order to demonstrate the versatility of the Co-Ni electrodeposition in MEMS technologies: one based on planar technology and the other, on bulk technology. In both cases good adherence, smooth nal surfaces and uniform deposition rates were obtained. Free-standing structures have been fabricated by sacricial etching of the seed layer as well as by micromachining of the silicon bulk structures. As the main conclusion it can be said that a soft magnetic Co-Ni lm fully compatible with MEMS technologies has been obtained.

Acknowledgments
The authors wish to thank Zen on Navarro, Ana S anchez and Josep Montserrat from the CNMs clean room staff for their help in this work.

405