High-current-density electrodeposition using pulsed and constant currents to produce thick CoPt magnetic films on silicon substrates

In recent years, thick-film hard magnetic materials have been increasingly investigated for their potential applications in MEMS, sensors and other microsystems.^1,2 L1₀-ordered equiatomic alloys of CoPt and FePt have shown tremendous promise because of the combination of (i) strong hard magnetic properties arising from large magnetocrystalline anisotropy, (ii) excellent corrosion resistance and chemical/thermal stability, and (iii) ability to be deposited in thick films using relatively simple, low-cost electroplating methods.^3,4

Previous research in CoPt electrodeposition on silicon has focused on films <10 μm at low current densities (<200 mA/cm²) and room temperature (<0.2 μm/min),⁵ but many applications demand films that are much thicker, perhaps in excess of 100 μm. Recently, electroplated films up to 60 μm with high coercivity (800 kA/m) have been reported by plating at on metal substrates using 1 A/cm² current density performed at 70 °C (plating rates up to 3 μm/min).⁶ Although this is a promising result, widespread application demands the ability to plate the films into photolithographically defined patterns on semiconductor substrates.

Plating thick films on silicon (or other semi-conductors) presents challenges including (i) substrate adhesion (particularly on highly polished semiconductor substrates), (ii) stress build up which can result in cracking,⁷ delamination, and film peeling, (iii) limited current efficiency of Pt and Pt alloys,⁸ and (iv) delamination/swelling of common polymeric electroplating molds/masks (typically photoresists).

This article explores very high current densities (up to 1 A/cm²) and heating the bath (up to 70 °C) for achieving very thick (over 20 μm) electroplated CoPt films on silicon substrates. Additionally pulsed electrodeposition is explored for reducing film stress, decreasing or eliminating cracks in the CoPt films, and improving substrate adhesion. Pulse plating has been known to effect internal stress in other types of electroplated films,^9,10 but has not been widely studied for CoPt.

This section details the experimental conditions and procedures used to plate CoPt films at high current densities, the subsequent annealing process, and the material and magnetic characterization techniques.

Samples were electroplated on (100) silicon substrates with a sputtered seed layer or 25 nm Cr/500 nm Cu. Because of the aggressive nature of the plating process (high-temperatures and high-current-densities), traditional thick, photoresist molds deteriorated in the bath. To combat this problem, 500-nm-thick alumina masks were used to localize the plating to a 2-mm-diameter circular window, and were not intended to produce well-defined vertical sidewalls around the deposit.

Samples were plated using a 25 mm x 25 mm high-purity Pt foil for the anode in 100 mL of electroplating bath. The distance between the anode and the cathode (substrate) was ∼25 mm. The electrolyte used 0.0515 M cobalt sulfamate to supply Co ions and 0.05 M diamine-dinitrito platinum (p-salt) to supply Pt ions. The bath also contained 0.1 M of ammonium citrate, which forms coordination complexes with the metal ions to bring the reduction potential of Co and Pt closer together and also acts as a pH buffer. The pH was precisely adjusted to 7 using NaOH. The metal ion concentrations in the bath were precisely determined by inductively coupled plasma mass spectrometry (ICP-MS).

Plating current densities were varied from 300 mA/cm² to 1 A/cm². The bath temperature was held constant at 70 °C, and the duration was 1 hr for all samples, except the sample plated at 1 A/cm², which was only 20 min. A Dynatronix DuPR10 Pulse Power Supply was used for the pulse-plated sample, and a Keithley 2400 power supply was used for the DC samples. For the pulsed sample, a 0.5 duty cycle was used with a frequency of 1000 Hz, with these parameters informed by a set of screening experiments (not reported here). The 0.5 duty cycle meant delivering peak current densities of 600 mA/cm² to produce an average current density of 300 mA/cm².

To promote a phase transformation to the L1₀ phase, which provides the desired magnetic properties in the films,¹¹ the samples were annealed in a tube furnace at 675 °C with a ramp time of 33 min, a dwell time of 30 minutes, and were then allowed to gradually cool to room temperature.

Average sample thickness was determined using a Tencor AS500 mechanical profilometer prior to annealing. After annealing, energy-dispersive X-ray spectroscopy (EDS) analysis (18 kV) was used to determine the CoPt ratio in the films and cross-sectional imaging was performed using a FEI Nova 430 Scanning Electron Microscope (SEM). The magnetic properties of each sample were tested using a vibrating sample magnetometer (VSM), ADE Technologies EV9. Both in-plane and out-of-plane features were recorded using a maximum applied field of ± 1800 kA/m. The in-plane results were used for sample-to-sample comparison.

Figure 1 shows the plating rates for films plated at high-current densities and 70 °C referenced against a previously reported plating rate at 50 mA/cm² and room temperature. Figure 2 shows corresponding cross-section SEM images, showing distinct morphologies differences for the samples. A current density of 1 A/cm² yielded plating rates of 1.5 μm/min with some regions of the film growing even faster. The deposit showed a porous structure, and EDS analysis indicated the films to be 70% Pt, although the composition was not totally uniform throughout. More moderate current densities (300 – 500 mA/cm²) yielded slower plating rates (0.35 μm/min) but denser deposits and a more desirable CoPt ratio between 50%-55% Pt. However, these films exhibited poor adhesion to the substrate often peeling or buckling during or after annealing.

Subsequent experiments focused on pulse plating in an attempt mitigate stress and improve adhesion. Table I shows an example direct comparison between a pulsed and constant-current sample (300 mA/cm²). Using pulsed currents, the film remained attached the substrate throughout the entire duration of the plating and annealing and was 6 μm thicker than its constant-current counterpart (24 μm vs. 18 μm). Pulse plating also influenced the elemental ratio and the magnetic properties. Magnetically the pulse-plated film was inferior to DC films but still displayed hard magnetic characteristics.

Table II and Figure 3 summarize the magnetic properties—coercivity, remanence, and maximum energy product (BH_max)—for one pulse-plated sample (Sample A) and three constant current samples (B, C, & D). Sample C, which was plated at 500 mA/cm², showed the highest BH_max and highest remanence. Sample B, plated at 300 mA/cm² showed the highest coercivity and an energy product close to that of Sample C. The magnetic properties of Sample D, DC plate at 1 A/cm², were negatively affected, likely due to being 70% Pt.

The use of high-current densities (300 - 1000 mA/cm²) and heated baths (70 °C) in electroplating CoPt on silicon substrates provides significant improvement in plating rates (>0.25 μm/min) and attainable film thicknesses (20+ μm in only 1 hr) as compared to lower current densities and room temperature. Films exhibited desirable hard-magnetic properties: high coercivities (815 kA/m), and high energy product (46 kJ/m³). Substrate adhesion poses a challenge to films plated at 300-500 mA/cm², but pulse plating shows potential for improved substrate adhesion with the added benefit of increased film thickness over an equal time period. While the pulse plated film in this paper was magnetically inferior to the constant-current samples, it stands to reason that further optimization could improve the magnetic properties.

The authors thank the staff of the UF Research Service Centers (RSC’s) for assistance with microfabrication and material characterization. This research was sponsored by the Defense Advanced Research Projects Agency (DARPA) Microsystems Technology Office (MTO) and the Army Research Office, and was accomplished under Grant Number W911NF-17-1-0050. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, DARPA MTO, Army Research Office or the U.S. Government.