Corrosion behavior of the nickel/nickel interface during the copper‐sacrificial layer release process in micro‐electroforming

This study investigates the corrosion behavior of the nickel/nickel (Ni/Ni) interface during the release of the copper‐sacrificial layer in multilayered micro‐electroforming. By altering the pre‐ and posttreatment of multilayered micro‐electroforming (degreasing, in situ anodic treatment, pickling, repassivation, and heat treatment), six different types of Ni/Ni interfaces were obtained. The activities and surface characteristics of the pretreated Ni substrates were investigated via open circuit potential and X‐ray photoelectron spectroscopy. The microstructures of the as‐prepared Ni/Ni interfaces were determined using scanning electron microscopy; after releasing the sacrificial layer in the ferric chloride solution, their corrosion morphologies and corrosion depths were observed under a three‐dimensional microscope with an ultra‐depth of field. The Ni/Ni interface showed a pre‐existing passive layer and interfacial defects. The interface was more prone to attack than the Ni base. Pitting corrosion along the interface boundary occurs via three main processes—initiation, growth, and pit coalescence—with a corresponding reduction in the interfacial adhesion strength and ultimately the structural integrity. The combination of substrate modification (degreasing and pickling) and heat posttreatment effectively avoids localized corrosion. We believe that the surface activation and thermally induced diffusion have worked.

Many methods are now available for improving the adhesion strength, such as surface activation before plating, [8] regulation of the electrodeposition current density, [9] and ultrasound-assisted electrodeposition. [10] However, the postprocesses, photoresist removal, and sacrificial-layer release, which significantly affect the interfacial bonding strength, are rarely investigated.
The sacrificial-layer technique is essential in the layered mask molding of structures with high aspect ratios, especially the movable suspension parts in MEMS devices. Copper (Cu), zinc (Zn), titanium (Ti), and SU-8 photoresists are usually utilized as sacrificial-layer materials. Common etchants for removing the sacrificial layer are inorganic acids such as hydrochloric acid and oxidizing agents such as ferric chloride, chlorite, peroxide, and persulfate. [11][12][13][14][15] These etchants can also corrode the structural metals (e.g., Ni, NiFe), albeit with a slow etching rate. Du et al. [16] dissolved thick SU-8 sacrificial layers on Ni structural layers in a boiling inorganic acid solution and found errors between the designed and actual dimensions of the microstructure. Wu et al. [17] prepared an effective etching solution by dissolving an additive in ammonia liquor. Aided by a violent oxidant, this solution selectively etched thick sacrificial Cu laminations (>100 μm) at an etching rate of 156 μm/h while dissolving the Ni structure layers at only approximately 0.1 μm/h. An effective two-component stripper designed by Zhang et al. [18] comprising ammonium hydroxide and chlorite removes Cu at a very high rate (460 μm/h) while corroding Ni at only 0.01 μm/h; however, pits were created on the Ni structure layers. In contrast to uniform corrosion, which has a predictable rate of metal dissolution, pitting corrosion has an unexpectedly high rate of localized metal dissolution and is unacceptable in microstructures. As shown in numerous studies, [19][20][21] aggressive species, such as haloid ions and sulfate, and metallurgical heterogeneities, such as inclusions, dislocations, interfaces, and grain boundaries, facilitate the localized corrosion of metals. For a multilayered Ni microstructure, Ni/Ni interfaces are the main metallurgical heterogeneities and should be attacked in the etching process of the sacrificial layer compared with the Ni base. Especially, after the long release time of the ultrathick (>700 μm) sacrificial layer, the localized corrosion of the interfaces is severe and drastically decreases the interfacial bonding strength. Although the investigation of the corrosion behavior of the Ni/Ni interface is essential, it has not been attempted in previous literatures.
In this study, multilayered Ni structures were fabricated by electroforming. The interfacial properties of the Ni/Ni interfaces were adjusted by varying the pre-and posttreatments. The corrosion behaviors of the resulting six Ni/Ni interfaces were observed during the wetetching process of an ultrathick (>700 μm) Cu-sacrificial layer in ferric chloride solution. Based on the evolution of corrosion damage, the likely corrosion mechanism was deduced as a guide for improving the corrosion resistance performance and the interface bonding energy of the multilayered structures.

| Preparation of the Ni/Ni interface
The multilayered Cu-Ni structure was built on a 2-mmthick T2 copper substrate (purity ≥99.9%) with an exposed area of 30 mm × 30 mm. The Cu-sacrificial layer was prepared in a standard copper-plating electrolyte composed of 200 g/L CuSO 4 ·5H 2 O, 50 g/L concentrated H 2 SO 4 solution, and 0.15 g/L 36% HCl at a current density of 3 A/dm 2 for 1 h. The Ni structure layer was created by applying a pulse current at 50°C in a nickel sulfamate bath (pH 4.0) containing 350 g/L Ni(NH 2 SO 3 ) 2 , 10 g/L NiCl 2 , 35 g/L H 3 BO 3 , and 0.2 g/L wetting agent. The average current density, frequency, and duty ratio were 3.75 A/dm 2 , 1000 Hz, and 20%, respectively. After 5 h of plating, an approximately 200-μm-thick Ni layer was obtained. Subsequently, the layer was ground and polished to a relatively smooth surface that can serve as the substrate for the next Ni layer. The fabrication of a multilayer nickel structure simulating the MEMS electroforming process is shown in Figure 1.
Before electrodepositing the latter Ni layer, residual impurities and/or the remaining oxides were removed from the former Ni substrate through five types of pretreatments. The resulting substrates were labeled I-V as described below. Substrate-I (the control specimen) was not pretreated and was used as a standard in a comparative analysis of the treated samples. Substrate-II was electrolytically degreased at 70°C for 5 min in an alkaline bath containing 5 g/L Na 2 SiO 3 ·9H 2 O, 30 g/L Na 2 CO 3 , and 4 g/L NaOH under a constant-current condition (4 A/dm 2 ). Substrate-III was degreased and then electrolyzed for 5 min in a nickel sulfamate bath using an anode at a current density of 2 A/dm 2 . Substrate-IV was also degreased and then pickled in hydrochloric acid solution (50% vol/vol) for 10 min. Substrate-V was oxidized in concentrated nitric acid to create an artificially thickened oxidized layer. This specimen was designed to study the effect of surface nickel oxide on the interfacial properties of the structures.
The preconditioned Ni substrates were rinsed with deionized water and immediately transferred to the nickel sulfamate bath, where electrodepositing was started several times under the same plating conditions as the first Ni layer. The resulting five interfaces were designated as Interface-I, Interface-II, Interface-III, Interface-IV, and Interface-V. The multilayered Ni/Ni sample containing Interface-IV was heat-treated at 450°C for 6 h in an argon atmosphere to change its interfacial property. The changed Interface-IV was named Interface-VI.

| Specimen characterization
To test the efficiencies of the pretreatments, the open circuit potentials (OCPs) of the pretreated nickel substrates in a 5% Na 2 SO 3 solution were measured in a standard three-electrode cell consisting of a working electrode (exposed area = 1 cm 2 ), a saturated Ag/AgCl reference electrode, and a Pt foil counter electrode. All measurements were conducted on a PGSTAT302N electrochemical workstation (Metrohm Corporation) at 25°C until the OCP changed by less than 10 −6 V/s. To avoid repassivation of the pretreated nickel substrates, the oxygen content in the solution was minimized by adding Na 2 SO 3 as a reducing agent.
The surface characteristics of the pretreated Ni substrates were measured using X-ray photoelectron spectroscopy (XPS; AXIS-Ultra) with monochromatic Al Kα radiation (1486.71 eV) operating at 16 kV. The binding energies were calibrated using the C 1s hydrocarbon peak at 284.80 eV to compensate for surface charge effects.
The cross-sections of the multilayered structure were ground with 600#, 1000#, and 2000# SiC papers and polished with 3-μm diamond suspensions on a polishing pad. The polished samples were gently wiped for 3 s with a mixed solution of concentrated HNO 3 (38 ml), concentrated CH 3 COOH (100 ml), and H 2 O (10 ml). This mixture is the standard metallographic etchant for electrodeposited Ni. The cross-sectional metallographic micrographs and interfacial microstructures were observed using optical microscopy (Zeiss Axio Observer A1 m) and scanning electron microscopy (SEM; Zeiss Ultra 55).

| Chemical etching of the sacrificial layer
Ferric chloride solution is extensively used as a Cu etchant in the printed circuit board and surface finishing industries. Here, it was chosen for releasing the ultrathick copper-sacrificial layer (>700 μm) because of its high etching rate and good selectivity. The solution was prepared by mixing 30 g of FeCl 3 ·6H 2 O, 2 ml of HCl, and 200 ml of H 2 O. The as-prepared Cu/Ni multilayered structure was incised into 5 mm × 5 mm × 2 mm blocks using a cutting machine. After polishing, the block was placed in a string bag and suspended in the etchant solution for 120 min. Etching images and depth profiles were analyzed using an LH-WN-YH 3D microscope with F I G U R E 1 Electroforming process of a multilayer nickel structure ultra-depth field manufactured by Chengdu Liyang Precision Machinery Co., Ltd.

| Activity and surface characteristics of the pretreated substrates
After mechanical grinding and polishing, dust, organic pollutants, and a passive oxide layer remained on the surfaces of the Ni substrates. Without treatment, these impurities would inhibit bonding between the Ni substrate and the electrodepositing layer, resulting in poor adhesion strength and low corrosion resistance. [22,23] To improve the interfacial bonding strength, the surface was activated before plating. By measuring the OCP before and after pretreatment, we can determine the degree of electrode activation-the more negative the potential, the more active the substrate. The OCP results of the corresponding substrates are shown in Table 1. The stable potential of the untreated Substrate-I was −0.428 V (vs. Ag/AgCl). The stable potentials of Substrate-II with the deoiling treatment and Substrate-III with the in situ anodic treatment were slightly negatively shifted from that of Substrate-I. Substrate-IV pretreated by two-step degreasing and pickling attained the most negative potential among the samples (−0.528 V vs. Ag/AgCl), whereas the OCP of Substrate-V ( − 0.331 V vs. Ag/AgCl) was more positive than that of the untreated one because it was repassivated in concentrated nitric acid.
The surface characteristics of the substrates were determined from the XPS measurements. The pretreated substrates were rinsed with deionized water and tested as soon as possible. As shown in the Ni2p 3/2 spectra, the oxide layer on the substrate was changed by the pretreatments (Figure 2a-e). The two main peaks at 852.5 and 855.7 eV originated from metallic Ni (Ni 0 ) and Ni (OH) 2 , respectively. [24,25] The broad shake-up peak at approximately 860.9 eV was the satellite peak of the main peaks and was possibly attributable to multielectron excitation. The percentages of metallic Ni (Ni 0 %) on the substrates were calculated from the area ratios of the Ni 0and Ni(OH) 2 -related peaks. The Ni 0 % of Substrate-IV (48%) was more than double that of Substrate-III (18%) and approximately five times those of Substrate-I (10%) and Substrate-II (9%). The Ni 0 % of Substrate-V was nearly 0. Overall, the OCP and XPS results showed that Substrate-IV was more activated than the other samples, whereas Substrate-V was passivated (least activated). The surface-activation mechanism of pickling can effectively improve the proportion of metallic atoms on the substrate by forming tight Ni electrodeposits. After electroforming, the Ni layer, the impurities on the pretreated substrate became inclusions and imperfections at the Ni-Ni interface.

| Microstructural characterization of the Ni/Ni interface
The cross-sectional characteristics of the as-prepared multilayered Ni structures were observed by SEM. The results are shown in Figure 3. The boundary lines (or interfaces between two Ni layers) are obvious and are sequentially indicated with Roman numerals. SEM micrographs of the Ni/Ni interfaces were obtained for a more detailed interpretation of the interfacial microstructures ( Figure 4). Unlike the Ni base, the interface shows some imperfections, including differently oriented dislocations, where atoms are mismatched between adjacent lattices, and cavities, where organic/oxide residuals on the substrate hindered the formation of the electrodeposited layer. Interface-I, Interface-II, and Interface-III were of the same width (∼100 nm), but the sizes and numbers of their cavities decreased in the order Interface-I > Interface-II > Interface-III. Interface-IV was approximately 50 nm wide and contained only a few tiny cavities. In contrast, Interface-V was as wide as 250 nm. The number of interfacial defects was proportional to the width of the interface, cavity size, and a number of cavities on the interface. It appears that the least and most severe imperfections appeared on Interface-IV and Interface-V, respectively.
As is clear from the XPS and SEM results, the superficial state of the substrate critically affects the interfacial properties of the substrate. As the substrates are progressively cleaned and activated, the interface becomes thinner and the cavities on the interface become smaller. Maintaining an activated nickel substrate during the transfer from the activated process to the plating solution seems to be impractical, as evidenced by the XPS results. For instance, Substrate-IV, which was activated in the hydrochloric acid solution for a long time, showed notable imperfections. However, an SEM characterization of the Interface-VI microstructure (Figure 4f) revealed a small width (20 nm) and no significant cavities. According to previous studies, [26,27] the thermally induced diffusion of Ni and O atoms at the interface decreases the number of density dislocations and defects, thereby improving the interfacial properties. Heat treatment could further reduce the interfacial defects.

| Behavior and mechanism of corrosion at the Ni/Ni interface
The ferric chloride solution selectively released the asprepared laminated Ni structures with an ultrathick Cusacrificial layer (see Section 2). Optical micrographs of the etching faces and the corresponding corrosion-damage profiles of the laminated structures were taken under a 3D microscope with an ultra-depth field. Figure 5 shows the etching images of the multilayered Cu/Ni/Ni structure containing Interface-Ι, Interface-II, and Interface-III, and Figure 6 shows the corrosion profiles of the multilayered Cu/Ni/Ni structure, including Interface-IV and Interface-V. Both structures were etched for 120 min. Figure 7 shows the etching morphologies of Interface-VI after releasing the Cu sacrificial layer in ferric chloride solution for 140 min. The passivation of nickel by surface growth of Ni 2+ anodic oxides in acidic and alkaline environments has been well-documented. [28,29] The passive film was dense and stable, preventing access of environmental species to the F I G U R E 2 X-ray photoelectron spectroscopy Ni2p 3/2 peaks of the nickel substrates: (a) Substrate-I, (b) Substrate-II, (c) Substrate-III, (d) Substrate-IV, and (e) Substrate-V F I G U R E 3 Cross-sectional metallography of the multilayered Ni structure observed via optical microscopy. Interface-I, Interface-II, Interface-III, Interface-IV, and Interface-V are sequentially labeled by their corresponding Roman numerals metal. Therefore, when exposed to ferric chloride solution, the Cu layer dissolved more easily than the passivated Ni layer. The different etching rates of the Cu-sacrificial layer and Ni-electroformed layer are shown in Figures 5b,c  and 6b,c. After etching the Ni-electroformed layer, the Cu/ Ni interface height decreased by approximately 700 μm. The etching rate of Cu was approximately 350 μm/h. The Ni/Ni interfaces were more susceptible to corrosion than the nickel base. Deep etching trenches formed along Interface-Ι, Interface-II, and Interface-III (Figure 5d-f) with approximate penetration depths of 30, 28, and 20 μm, respectively. Such preferential corrosion actually debonded Interface-V (Figure 6a). No obvious etching trench appeared on Interface-IV; however, shallow corrosion pits with diameters of 7-10 μm (indicated by red arrows in Figure 6d) were formed. After exposure to ferric chloride solution for 140 min, the altitude of the intercept between the Cu-sacrificial layer and the Ni structure layer was approximately 850 μm (Figure 7b,c); however, Interface-VI remained almost intact and uniform corrosion continued  (Figure 7a,d). These results confirm that heat treatment can significantly improve the resistance to localized corrosion. In terms of etching type and depth, the localized interfacial-corrosion resistance decreased in the order of Interface-VI > Interface-IV > Interface-III > Interface-Ι-Interface-II > Interface-V. Overall, the resistivity of the interface to chemical etching was directly related to the interfacial bonding performance, which itself depended on the quality of the substrate pretreatment. During the selective release of the ultrathick sacrificial layer, the interfacial-corrosion extent of the laminated Ni structure proportionally reduced the interfacial bonding strength and ultimately damaged the structural integrity.
To diagnose the corrosion type and elucidate the corrosion mechanism, the evolution of corrosion damage was investigated at Interface-I under the same treatment as the sacrificial-layer release process. The samples were placed in a string bag and suspended in ferric chloride solution for 30, 50, 70, 90, and 120 min. Submicrometerscale pits were initiated along with the interface at 30 min and grew to microscale pits over time (see Figure 8). Pits continually appeared, grew, and fused with neighboring ones until a boundary trench formed over the entire interface. This type of pitting corrosion was limited to microscopically thin zones along with the interface of the boundary. Figure 9 shows the expected front and cutaway views of the Ni/Ni interface during the modified pitting corrosion process. Many authors [30,31] have proposed that pit initiation occurs via local breakthrough of the passive oxide film in the presence of halogen ions, especially Cl − , a small-volume halogen with high penetrating ability. This study proposes that at Interface-I, atoms are less thermodynamically stable at the dislocation and cavity sites in the crystal structure of the base metal than at perfect lattice sites; accordingly, the passivating oxide film is thinner and more vulnerable to corrosion than at perfect sites (Figure 9a,e). At dislocation and cavity sites, Cl − ions preferentially diffuse or penetrate the interface, where they react with the metal to form a soluble compound (Figure 9b). The propagation of pits aligns with the surface-defect distribution ( Figure 9f). As the active pits are surrounded by large passive areas with different potentials, they stimulate the formation of many passive-active microcells. [32,33] In the present study, the anodic area was inside the pit (Ni − 2e − → Ni 2+ ) and the cathodic zone was the ambient passive area (see Figure 9c,g). The main cathodic processes are oxygen dissolution in solution and ferric reduction (O 2 + 2H 2 O + e − → 4OH − ; Fe 3+ + e − → Fe 2+ ). Based on the pit-depth change from 30 to 70 min (Figure 8), the initiatory pit growth rate of the pits was determined as only 0.12 μm/min. As the exposure time increases, the circulation of microcells causes a series of reactions and chemical modifications. Inside the pit, the pH is decreased by hydrolysis of dissolved nickel cations (Ni 2+ + 2H 2 O → Ni(OH) 2 + 2H + ), further stimulating the anodic attack. Therefore, the etching rate (calculated from the pit-depth change between 70 and 90 min in Figure 8) reached approximately 0.55 μm/min. Furthermore, pitting corrosion occurs within the interface zone and spreads across the surface. As multiple adjacent pits are simultaneously fused, the pit mouth widens and the ion exchange between the interior and exterior of the pits accelerates, thereby decreasing the local acidity. As shown in Figure 8, the pit-etching rate slowly decreased to 0.47 μm/min between 90 and 120 min, preventing deeper penetration of the pits. Consequently, a wide and shallow boundary trench, rather than a deep and narrow cavity, formed along with the interface (Figure 9d,h).
In conclusion, pitting corrosion of the Ni/Ni interface during the release of the ultrathick Cu-sacrificial layer occurred by three main processes-initiation, growth, and pit coalescence. Pit initiation is the kinetic limiting step. The pit initiation times were approximately 30 min at Interface-I ( Figure 8) and approximately 40 and 60 min at Interface-II and Interface-III, respectively. At Interface-IV, the pit initiation time was extended by 100-120 min (Figure 6d). Even after 140 min of selective etching, no visible pits appeared on Interface-VI (Figure 7d). Pit initiation is closely related to the interfacial property of the interface. A larger number of interface defects are associated with a longer pit F I G U R E 8 Evolution of corrosion damage along Interface-I during immersion in ferrite chloride solution. The red curves at later immersion times show the cross-sectional contour lines of the pits labeled by pit depth and the diameter of the pit mouth initiation time. Furthermore, pit coalescence largely influences the pit growth rate and corrosion shape.
The release process of an ultrathick (>700 μm) sacrificial layer is time consuming and may require several days. It also requires a strong oxidizer that either contains haloid ions (such as hydrochloric acid and cupric chloride) or decomposes to haloid ions (such as chlorite, hypochlorite, and periodate). The localized corrosion behavior of multilayered Ni or NiFe microstructures observed in the ferric chloride solution is also expected in these etchants. Such corrosion severely decreases the mechanical performance, reliability, and longevity of microdevices. Proper substrate pretreatments and heating treatments that eliminate the connecting defects in the interface zone through surface activation and thermally induced diffusion mechanisms will significantly improve the pitting-resistance properties of the interface. F I G U R E 9 Schematics of the proposed pitting corrosion mechanism of the Ni/Ni interface: front view (a-d) and cutaway view (e-h)