Sensors and Actuators A 92 (2001) 242±248 Electroplating and characterization of cobalt±nickel±iron and nickel±iron for magnetic microsystems applications F.E. Rasmussena,*, J.T. Ravnkildea, P.T. Tangb, O. Hansena, S. Bouwstraa a Mikroelektronik Centret, Technical University of Denmark, Oersteds Plads, Bldg. 345 East, DK-2800 Kgs. Lyngby, Denmark b Department of Manufacturing Engineering, Technical University of Denmark, Bldg. 204, DK-2800 Kgs. Lyngby, Denmark Accepted 3 January 2001 Abstract The magnetic properties of pulse reverse (PR) electroplated CoNiFe and DC electroplated NiFe are presented. CoNiFe is a very promising material for magnetic microsystems due to the possibility of achieving a high saturation ¯ux density (Bs ) and a low coercivity (Hc ). A new bath formulation has been developed, which by means of PR electroplating makes it possible to deposit high Bs CoNiFe with a low residual stress level. The magnetic properties have been determined using a new simple measurement setup that allows for wafer level characterization. The results have been validated by comparison to measurements performed with a vibrating sample magnetometer (VSM). # 2001 Elsevier Science B.V. All rights reserved. Keywords: PR electroplating; CoNiFe; NiFe; Magnetic properties 1. Introduction 2. Fabrication Soft magnetic materials, such as PermAlloy (79 at.% Ni and 21 at.% Fe) are extensively used in electromagnetic MEMS devices, such as m-relays, m-switches, m-pumps and m-motors [1±3]. In electromagnetic devices, the attainable energy density is limited by the saturation ¯ux density (Bs ) of the soft magnetic material used. In order to further miniaturize and/or improve performance of electromagnetic devices, development of new soft magnetic high Bs materials are needed (see also [4]). A strong candidate for replacement of NiFe in many MEMS applications is the ternary alloy CoNiFe capable of providing an extremely high saturation ¯ux density combined with reasonably low coercive ®eld strength [5±7]. However, controlling the electrochemical deposition of a ternary alloy is a complex matter, since composition, magnetic and mechanical properties depend on the various plating parameters. Utilizing the pulse reverse (PR) electroplating technique, where the current is periodically reversed, the residual stress and composition can be optimized towards optimum magnetic and mechanical properties. The soft magnetic materials NiFe and CoNiFe are electroplated on 4 inch Si-wafers containing a thin layer of insulating thermal SiO2 , and an evaporated Ti/Au plating Ê ). In order to characterize the magnetic base (200/2000 A properties of the materials with the presented characterization setup, special test structures are fabricated using standard UV lithography and wet etching. The resulting test structures are illustrated in Fig. 1. The electroplated test structures have a width of 9.75 mm, a length of 32 mm and an average thickness ranging from 3 to 4 mm (varying from experiment to experiment). * Corresponding author. Tel.: 4526823493/45-25-57-00; fax: 45-45-88-77-62. E-mail address: fra@mic.dtu.dk (F.E. Rasmussen). 2.1. PR electroplating of CoNiFe A new formulation of a CoNiFe bath has been developed especially for PR electroplating (based on [5]). The bath is a chloride Ni bath, with the addition of Co and Fe as well as complexing agents. Several variations of the electrolyte have been investigated in the effort to optimize the CoNiFe plating process, and to obtain optimum magnetic properties. The composition of an electrolyte yielding high saturation ¯ux density is shown in Table 1. The CoNiFe electrolyte is designed to reduce residual stress without using stress reducing sulphur containing additives (SCAs), such as saccharin, which are known to deteriorate the magnetic properties. The 0924-4247/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 4 - 4 2 4 7 ( 0 1 ) 0 0 5 5 6 - 8 F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 243 25 liter setup utilizing a ®xed horizontal anode/cathode con®guration, and a moving paddle for agitation. 3. Characterization Fig. 1. Illustration of magnetic measurement setup. The setup consist of an U-shaped ferrite core, two coils and electronics for signal processing. Table 1 Characteristics of CoNiFe electrolyte Compound Concentration Cobalt chloride (g/l) Iron chloride (g/l) Nickel chloride (g/l) Nickel sulfate (g/l) Boric acid (g/l) Naphtalene trisulfonic acid (g/l) 5-Sulfosalicylic acid (g/l) Deposition temperature ( C) pH 57.08 31.81 73.69 26.28 40.17 4.34 10.16 40 3.9±4.2 stress reducing ability is provided through the combination of a small amount of naphtalene trisulfonic acid (NTS) and optimized pulse plating parameters [8]. The CoNiFe ®lms are electroplated at current densities ranging from 1.5 to 2.08 A/dm2 at a temperature of 40 C. The PR electroplating process is performed in a 25 liter tank with a computer controlled pulse generator and a moving vertical anode and cathode con®guration. 2.2. DC electroplating of NiFe The DC electroplated NiFe has been fabricated for the purpose of comparison with CoNiFe. The formulation of the NiFe bath (presented in Table 2) is based on a Ni sulfate bath (Watts type) [1]. The NiFe is deposited at elevated room temperature (28 C) at current densities ranging from 0.25 to 1 A/dm2 . The electroplating process is performed in a The magnetic properties of the electroplated NiFe and CoNiFe alloys have been investigated using a new setup allowing for wafer level characterization of electroplated ®lms. The setup exploits basic electromagnetic principles, and can easily be fabricated with simple and inexpensive electronics. The setup consists of an U-shaped ferrite core, two coils and electronics for ampli®cation and integration of the signals. A sinusoidal signal is applied to the primary coil in series with the resistor R1. The voltage across R1 is ampli®ed and displayed on an oscilloscope in order to obtain the current through the primary coil, I1 , and thereby the magnetic ®eld strength in the alloy sample, Ha . The relation between current and magnetic ®eld strength is given by Amperes law I X N1 I1 H dl Hi li Hy ly Ha la (1) i where N1 is the number of turns on the primary coil and l the length of the magnetic path in different parts of the closed magnetic circuit (see Fig 20). The indices y and a denote the yoke and the alloy sample, respectively. Eq. (1) is valid under the assumption that no current flows in the secondary coil, i.e. I2 0, and that no air gap is present between the ferrite core (the yoke) and the alloy sample, i.e. lair 0. If it, additionally, is assumed that the flux in the yoke equals the flux in the alloy sample, i.e. fy fa Ay By Aa Ba , where A is the cross-sectional area and B is the magnetic flux density, and assuming that there is a linear relationship between B and H (B m0 mr H), then Hy can be expressed as a function of Ha . Accordingly, Ha can be obtained by slightly rearranging Eq. (1) Ha N 1 I1 Aa =Ay mr;a =mr;y ly la (2) Since Aa Ay (and mr;a mr;y ) Eq. (2) can be reduced to Ha N 1 I1 la (3) Table 2 Characteristics of NiFe electrolyte Compound Concentration Nickel sulfate (g/l) Iron sulfate (g/l) Nickel chloride (g/l) Boric acid (g/l) Saccharin (g/l) Sodium lauryl sulfate 1 cm (g/l) Deposition temperature ( C) pH 200 8 5 25 2 0.1 28 3 Fig. 2. Illustration of magnetic path through yoke and alloy sample. It is assumed, that I2 0 and that no airgap is present between ferrite core and alloy sample. 244 F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 Thus, yielding a simple relation between primary current and magnetic field strength in the alloy sample. The magnetic ¯ux density in the alloy sample Ba can be obtained by measuring the induced voltage in the secondary coil. According to the law of induction, the induced voltage E is related to the magnetic ¯ux f2 through the secondary coil as E t df2 dt (4) Using f2 N2 Ay By together with the assumption fy fa Ay By Aa Ba and an inverting amplifier, the magnetic flux density in the alloy sample can be written as Z 1 E t dt (5) Ba N2 Aa The full hysteresis curve is obtained by plotting Ba as a function of Ha . The electroplated NiFe and CoNiFe ®lms were characterized with respect to composition by quantitative energy dispersive X-ray (EDX) analysis, with an estimated accuracy of 0.5 at.% (error bars not visible on given curves). The residual stress levels of the electroplated ®lms were determined at room temperature (22 C) using the wafer curvature method. Cooling the wafers from the deposition temperature to room temperature results in a thermal stress contribution. As a general rule of thumb, the thermal induced stress equals 3 MPa (tensile) for each degree the deposition temperature is elevated above room temperature. Thus, the reported stress levels in this work include a thermal contribution of approximately 18 MPa (tensile) for NiFe alloys (deposition temperature 28 C) and 54 MPa (tensile) for CoNiFe alloys (deposition temperature 40 C). 4. Results and discussion 4.1. The DC electroplating of NiFe The electroplated NiFe alloys exhibit the well known anomalous codeposition phenomenon, where preferential deposition of Fe results in a larger Fe content in the deposit, than directly given by the content of Ni and Fe in the electrolyte (see Table 2). The Ni and Fe content as a function of current density is presented in Fig 3. The residual stress in the electroplated NiFe ®lms was found to decrease with increasing current density. Typically, a stress level of 160±170 MPa was obtained for ®lms deposited at 1 A/dm2 . The current ef®ciency of the NiFe electroplating process varied from 62 to 80%, which is rather low, but in agreement with existing results published on electroplating of NiFe [9,10]. The current ef®ciency was found to increase with increasing current density in the investigated region 0.25±1 A/dm2 . The decreasing stress as a function of current density is believed to be related to the increasing current ef®ciency. At lower current Fig. 3. Composition of DC electroplated NiFe vs. plating current density. The composition is measured using quantitative energy dispersive X-ray analysis. ef®ciencies secondary cathode reactions, such as evolution of hydrogen may contribute signi®cantly to the formation of stress due to diffusion of hydrogen into the electroplated alloy [11]. 4.2. PR electroplating of CoNiFe Like the electroplating process of NiFe, electroplating of CoNiFe also exhibits anomalous codeposition behavior. However, the mutual relation between deposition of Co, Ni and Fe in electroplating of the ternary CoNiFe alloy is more complicated than in electroplating of binary alloys, such as NiFe. The composition of the CoNiFe alloy exhibits a strong dependence on current density as shown in Fig 4. The average current density can be regarded as an effective cathodic plating current density. The illustrated dependence of the current density is general for the observations done in this work. The Co content is almost constant as a function of current density, whereas the Fe content increases Ð with a corresponding decrease in Ni content Ð for increasing current densities. These observations indicate an intensi®ed Fig. 4. Composition of PR electroplated CoNiFe vs. average plating current density. The composition is measured using quantitative energy dispersive X-ray analysis. F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 Fig. 5. Composition of CoNiFe as a function of current density for different PR electroplating baths (I±IV). The arrows indicate the direction of increased plating current density. Regions A and B corresponds to CoNiFe compositions potentially having a coercivity of less than 160 A/m [5]. The dashed line defines a region of CoNiFe compositions potentially having a saturation flux density above 1.8 T [5]. anomalous codeposition behavior of Ni and Fe at higher current densities. A three element plot showing the composition of CoNiFe ®lms deposited from different PR electroplating baths (I±IV) is shown in Fig. 5. Bath II has an increased content of Fe compared to bath I. Bath III corresponds to bath II with increased pH. Bath IV has an increased Fe and Co content, and a decreased Ni content compared to bath III (shown in Table 1). The arrows indicate the directions of increased current density. Regions A and B in Fig. 5 correspond to CoNiFe compositions potentially having a coercivity of less than 160 A/ m [5]. The dashed line de®nes the compositional region reported to have a saturation ¯ux density (Bs ) above 1.8 T [5]. Generally it can be stated, that increased Fe content and decreased Ni content enhance Bs , when the Co content retains a constant value lying between 40 and 65 at.% [5]. The development process of the CoNiFe electrolytes in this work has, therefore, been focused on impeding Ni deposition and increasing Fe deposition, while maintaining a high Co content. The Fe content in the deposited alloy is increased at the expense of Ni by increasing the Fe content in the electrolyte. Correspondingly, the Co content can be increased in the alloy by increasing the Co content in the electrolyte. The pH value of the electrolyte also in¯uences the composition of the CoNiFe alloy. Thus, an increase in Fe content in the alloy was observed at increased pH (II ! III). The resulting stress levels in the PR electroplated CoNiFe ®lms were low with values ranging from 69 to 101 MPa including a thermal contribution of 54 MPa. The variation of the stress level with current density is shown in Fig. 6. As evident the residual stress increases as a function of current density. Furthermore, an increase in the residual stress level was observed for increasing Fe content (see Fig. 7). The relation between Fe content in the CoNiFe alloy and residual stress level indicates, that a compositional 245 Fig. 6. Residual stress level of PR electroplated CoNiFe films as a function of average plating current density. The stress level has been determined from the wafer curvature method. Fig. 7. Residual stress level of PR electroplated CoNiFe as a function of Fe content. The stress level has been determined from the wafer curvature method. optimum yielding optimum magnetic properties and minimum residual stress can be obtained. The current ef®ciency of the PR electroplating process has been optimized, since this parameter is of great importance for industrial application of the process. The current ef®ciency exhibited a strong dependence on the pH value of the electrolyte (see Fig. 8). Fig. 8. Current efficiency of CoNiFe PR electroplating process at two different plating temperatures as a function of pH value. 246 F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 Increasing the pH value from 1.7 to 2.3 improved the current ef®ciency from 66 to 86%. The current ef®ciency was further improved to 96% by decreasing the temperature of the electrolyte (from 43 to 40 C). The current ef®ciency remained stable (96±97%) for current densities in the investigated region 1.5±2.08 A/dm2 . This level of ef®ciency is high compared to the current ef®ciency of CoNiFe electroplating processes reported in literature (e.g. [7]). The pH value was observed to increase as a function of plating time. This is attributed to evolution of hydrogen occurring despite the high current ef®ciency. A constant increase in pH of 0.04 was observed in plating experiments converting a charge of 900 C. As a consequence the pH of the electrolyte has to be continuously monitored and adjusted. In this work, the pH was allowed to vary from 3.9 to 4.2 before adjustment was performed. 4.3. Magnetic properties The magnetic properties of main interest are the saturation ¯ux density Bs and the coercivity Hc. The magnetic properties of CoNiFe and NiFe ®lms with the most promising compositions were characterized using a vibrating sample magnetometer (VSM). The results of the VSM measurements are shown in Fig. 9. The measurements on the CoNiFe alloy reveal a high saturation ¯ux density, ca. 1.7 T, and a reasonably low coercivity, ca. 510 A/m. The coercivity is low, but not below 160 A/m as reported in [5]. The low coercivity obtained in [5] is attributed to a small grain size in the material of ca. 10±20 nm, which is obtainable in the presence of mixed crystalline phases [6]. The present results of Hc indicate, that the optimum relation between crystalline phases (fcc and bcc) is not present in the developed material. As evident from Fig. 9, the DC electroplated NiFe exhibits an extremely low coercivity, and a fairly high saturation ¯ux density. The value of Hc is on the edge of the detection limit Fig. 9. Hysteresis loops of selected CoNiFe and NiFe samples obtained with a VSM. The characterized CoNiFe alloy with a composition of 58.5 at.% Co, 18.1 at.% Ni and 23.4 at.% Fe is located in region A in the three element plot shown in Fig 5. Fig. 10. Hysteresis loops of selected CoNiFe sample obtained with the new characterization setup (compensated) and with a VSM. The characterized CoNiFe alloy with a composition of 65.8 at.% Co, 13.1 at.% Ni and 21.2 at.% Fe is located in region A in the three element plot shown in Fig. 5. of 10 A/m of the VSM. The saturation ¯ux density is approximately 1.1 T, which is high compared to values of Bs for PermAlloy given in literature (ca. 0.8±1.1 T) [3,10]. The developed CoNiFe and NiFe have a reasonably high permeability in the range of 2000±3000; roughly corresponding to the permeability of other electroplated high m materials, such as NiFeMo, given in the literature [4]. In order to validate the functionality of the simple characterization setup illustrated in Fig. 1, CoNiFe and NiFe samples have been characterized using the new setup and a VSM. The resulting hysteresis loops for a CoNiFe ®lm with a composition of Co 65.8 at.%, Ni 13.1 at.% and Fe 21.2 at.% are shown in Fig. 10. The hysteresis loop obtained with the new characterization setup deviated slightly from the hysteresis loop obtained with the VSM. The magnitude of the measured magnetic ®eld strength H was in agreement with the values obtained with the VSM, whereas the measured ¯ux density B was 20± 30% larger than the corresponding VSM values of B. The deviation of B is primarily attributed to insuf®cient sample thickness (3±4 mm), which increases the loss of ¯ux to the ambient environment and direct coupling between primary and secondary coil. The hysteresis loop shown in Fig. 10 has been compensated for this deviation, subsequently exhibiting an overall shape in agreement with the VSM result. In order to verify that the alloy sample thickness is important for obtaining reliable values of B, a commercially available amorphous soft magnetic alloy, VitroVac 6025, has been characterized with the new characterization setup. The 20 mm thick VitroVac 6025 sample is known to have a saturation ¯ux density Bs of 0.55 T, a permeability mr approaching 100,000 and an extremely low coercivity. The hysteresis loop of VitroVac 6025 obtained with the new characterization setup is shown in Fig. 11. As evident the saturation ¯ux density is measured to approximately 0.55 T, and the obtained hysteresis loop reveals an extremely high permeability as well as an extremely low F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 Fig. 11. Hysteresis loop of VitroVac 6025 obtained with the new characterization setup. VitroVac 6025 is known to have a saturation flux density of 0.55 T, and an extremely low coercivity. coercivity. Thus, the obtained magnetic properties of VitroVac 6025 is in close agreement with the data stated by the manufacturer of VitroVac (Vacuumschmeltze). The increased accuracy of the characterization result is attributed to the increased sample thickness. Since the magnetic reluctance of the alloy sample decreases when the sample thickness increases an increased fraction of the total ¯ux ¯ows through the alloy sample. Thus, the ¯ux density B inferred from the measured induced voltage E t in the secondary coil, becomes an increasingly better estimate of the actual ¯ux density in the alloy sample. 247 [2] E.J. O'Sullivan, et al., Integrated, variable-reluctance magnetic minimotor, IBM J. Res. Dev. 42 (1998) 681±691. [3] C. Liu, T. Tsao, G.-B. Lee, J.T.S. Leu, Y.W. Yi, Y.-C. Tai, C.-M. Ho, Out-of-plane magnetic actuators with electroplated permalloy for fluid dynamics control, Sens. Actuators A 78 (1999) 190±197. [4] W.P. Taylor, M. Schneider, H. Baltes, M.G. Allen, A NiFeMo electroplating bath for micromachined structures, Electrochem. Solid-State Lett. 2 (1999) 624±626. [5] T. Osaka, M. Takai, K. Hayashi, K. Ohashi, M. Saito, K. Yamada, A soft magnetic CoNiFe film with high saturation magnetic flux density and low coercivity, Nature 392 (1998) 796±798. [6] T. Osaka, M. Takai, Y. Sogawa, T. Momma, K. Ohashi, M. Saito, K. Yamada, Influence of crystalline structure and sulfur inclusion on corrosion properties of electrodeposited CoNiFe soft magnetic films, J. Electrochem. Soc. 146 (1999) 2092±2096. [7] I. Tabakovic, S. Riemer, V. Inturi, P. Jallen, A. Thayer, Organic additives in the electrochemical preparation of soft magnetic CoNiFe films, J. Electrochem. Soc. 147 (2000) 219±226. [8] P.T. Tang, P. Mùller, H. Dylmer, An electroplating method of forming platings of nickel, cobalt, nickel alloys or cobalt alloys, PCT Application, WO 97/00980, 9 January 1997. [9] J.T. Ravnkilde, V. Ziebart, O. Hansen, H. Baltes, Mechanical characterization of electroplated nickel-iron, Proc. Eurosens. XIII (1999) 383±386. [10] J.-M. Quemper, S. Nicolas, J.P. Gilles, J.P. Grandchamp, A. Bosseboeuf, T. Bourouina, E. Dufour-Gergam, Permalloy electroplating through photoresist molds, Sens. Actuators A 74 (1999) 1±4. [11] P.T. Tang, Fabrication of micro components by electrochemical deposition, Ph.D. Thesis, The Institute of Manufacturing Engineering (IPT), Building 424, Technical University of Denmark, DK-2800 Lyngby, Denmark, 1998. Biographies 5. Conclusion A new bath formulation has been developed for PR electroplating of CoNiFe. A current ef®ciency of 96% and a residual stress of 70 MPa have been obtained. The magnetic properties of CoNiFe exhibit a high Bs of 1.7 T and reasonably low Hc of 510 A/m. In comparison, the electroplated NiFe shows an extremely low Hc of less than 10 A/m and a Bs of 1.1 T. The results show that PR electroplated CoNiFe could become a serious alternative to NiFe. Magnetic measurements using a new characterization setup were validated with a VSM with a reasonable result. Acknowledgements The authors would like to acknowledge Dr. J.C. Lodder and T. Bolhuis, Information Storage Technology Group (ISTG), MESA Research Institute, Twente University, The Netherlands, for performing the VSM measurements. References [1] W.P. Taylor, O. Brand, M.G. Allen, Fully integrated magnetically actuated micromachined relays, J. MEMS 7 (1998) 181±191. Frank Engel Rasmussen was born in 1975 in Copenhagen, Denmark. He received his MSc in Electrical Engineering (Master's thesis title: Development of Soft Magnetic Materials for Microsystems Applications) from Mikroelektronik Centret (MIC), the Technical University of Denmark, in 2000. He is currently working towards an industrial PhD degree within the field of advanced packaging at the Danish company Oticon A/S. Jan Tue Ravnkilde received his MSc in Electrical Engineering from the Technical University of Denmark in 1997 and is currently working towards his PhD degree. He is developing electroplated microelectromechanical systems for fabrication on preprocessed integrated circuits. Peter T. Tang received his MSc in Chemical Engineering from the Technical University of Denmark in 1991, and his PhD in 1998 based on a thesis entitled ``Fabrication of Micro Components by Electrochemical Deposition''. He is presently employed as assistant research professor by the Department of Manufacturing Engineering at the Technical University of Denmark. He has worked on various projects within the field of electrochemistry and electrochemical deposition and holds two patents. Current research topics are fabrication of tools for microinjection moulding, as well as electrodeposition of alloys and pure metals for microcomponents. Ole Hansen received his MSc in Electrical Engineering from the Semiconductor Laboratory, the Technical University of Denmark in 1977. Since 1977, he has done research first at the Semiconductor Laboratory, the Technical University of Denmark, now at Mikroelektronik Centret (MIC), the Technical University of Denmark, in the fields of Bipolar Technology, CMOS Technology, Micromachining of Silicon, Microsystems Technology, Microdevices by Metal Plating, Scanning Probe Technology, and Biochemistry in Microsystems. Since 1994, he has 248 F.E. Rasmussen et al. / Sensors and Actuators A 92 (2001) 242±248 been an associate professor at MIC. He is presently teaching two lecture courses: Semiconductor Technology, and Semiconductor Devices. Siebe Bouwstra received his MSc in Mechanical Engineering in March 1984 from the University of Twente, The Netherlands. In April 1984, he joined the Micromechanics research group at the University of Twente as an Associate Researcher, where he was active in public±private collaboration projects with Oce Copiers B.V. and ASM International B.V., respectively. He received his Doctorate degree in March 1990 based on his thesis entitled ``Resonating Microbridge Mass Flow Sensor''. After this he was awarded a Research Fellowship of the Royal Dutch Academy of Sciences, with which he collaborated with the Center for Integrated Sensors and Circuits of the University of Michigan, USA. In November 1992, he was appointed Associate Professor at Mikroelektronik Centret at the Technical University of Denmark, where he has been responsible for the Micro-Electro-Mechanical Systems research and education programme. Since January 2000, he has joined the private company Microcosm Technologies B.V. in The Netherlands, the European technical office of Microcosm Technologies Inc., where he is responsible for the company's Technical Services in Europe.