Nickel Characterization for Interconnect Analysis Yuriy Shlepnev #1 , Scott McMorrow *2 # Simberian Inc. 3030 S Torrey Pines Dr., Las Vegas, NV, 89146, USA 1 shlepnev@simberian.com * Teraspeed Consulting Group LLC 121 North River Drive, Narragansett, RI, 02882, USA 2 scott@teraspeed.com Abstract—Landau-Lifshits model of ferromagnetic metal permeability is proposed in the paper for broad-band characterization of nickel in PCB/packaging interconnects made of copper plated with nickel and gold (ENIG finish). Unknown parameters of the plated nickel are identified with the measured generalized modal S-parameters of nickel-plated microstrip line segment and electromagnetic analysis of the same segment with multi-layered conductor interior model. The model predicts dispersive frequency dependency of nickel permeability with a resonance between 2 and 3 GHz. The resonance produces an anomaly in the insertion loss and group delay consistent with the experimental data. I. INTRODUCTION Electroless nickel/immersion gold (ENIG) interconnect finish is widely used over the last few years as a lead-free alternative to protect copper PCB and packaging interconnects from oxidation. ENIG finish produces a clean surface suitable for soldering for a long period of time. A layer of gold preserves interconnects from oxidation and may reduce signal attenuation at high frequencies. Nickel serves as a barrier preventing inter-diffusion of gold into copper. Appropriate modelling of the nickel layer on such boards is important for accurate prediction of signal degradation effects. Electrical characterization of nickel for analysis of microwave and digital signal propagation in interconnects with ENIG finish is the subject of this paper. Nickel is a ferromagnetic metal with relatively large resistivity and non-unit frequency-dependent relative permeability [1]-[2]. A good review of nickel characterization is provided in [3]. Nickel is probably the most mysterious metal widely used in electronics. DC resistivity of pure nickel, reported by different authors, ranges from 4 to 5 of resistivity of copper [3]. It was also observed that the resistivity of the nickel layer can be 3-10 time of pure nickel [4]. Reports on the permeability of nickel are even more confusing – from 1 to 600 at DC and from 1 to 20 at RF/microwave frequencies [3]. The differences in outcome can be explained by the differences in the samples, process impurities and differences in investigation techniques. Thus nickel must be characterized only in the context of a particular manufacturing technology and for a particular application. Here, we will focus on ENIG finish technology and interconnects for multi-gigabit or microwave signals with the electromagnetic fields transverse to the wave propagation direction. A relatively thick layer of gold (1-5 µm) on top of nickel electrically shields the nickel layer at microwave frequencies and reduces the signal degradation due to the large resistance and ferromagnetic properties of the nickel [5]. Effect of nickel can be safely neglected in these cases. In low-cost PCBs/packages the thickness of gold layer is usually below 0.1-0.2 µm. This is smaller than the current skin depth in gold at microwave frequencies (skin depth for gold is about 1.4 µm at 3 GHz and about 0.78 µm at 10 GHz). Anomalies in the insertion loss or attenuation at frequencies from 2 to 4 GHz due to ENIG finish with thin gold layer were observed earlier [5]-[7]. Though, the authors were not able to accurately reproduce such behaviour in simulations with simplified conductive or non-dispersive permeability models for nickel. ENIG plating model was constructed in [4] as an increase in total resistivity of the finished conductor. Ferromagnetic properties of nickel were neglected which led to large discrepancies between the model and measurements. Electromagnetic models with non-dispersive permeability of nickel were used in [5], [6] and did not reproduce the observed anomaly in insertion loss. In [3] it was suggested to use second order Debye permeability model for nickel, but parameters of the model were not identified and it was not verified with measurements. In this paper we start from experimental observation of the anomaly in insertion loss and group delay in interconnects plated with nickel and gold. The resonance can be observed on both original S-parameters and reflection-less generalized modal S-parameters. Next, we build an electromagnetic model of plated interconnects with the physics-based dispersive Landau-Lifshits (L-L) permeability model [8], to characterize the nickel layer. The model is based on the theory of moving boundaries between uniformly magnetized layers and predicts a resonance at the microwave frequency band. Assuming frequency-dependent properties of dielectrics, and other conductors and interconnect geometry are known, we will adjust parameters of L-L model of nickel in the computed model to match the measured generalized modal S-parameters. The derived permeability model can be used for reliable prediction of interconnect behaviour for that particular manufacturing technology over the frequency range where the measured and computed S-parameters were matched. Note that thin layers of electroplated nickel are also used in micro-electro mechanical structures (MEMS) and in micro- fabricated filters operating at RF/microwave frequencies [3]. 978-1-4577-0811-4/11/$26.00 ©2011 IEEE 524