Research on the bonding reliability of chemical nickel-palladium lead wires

In the "last mile" connection of semiconductor packaging, gold wire lead bonding technology is the core solution for the electrical interconnection between IC chips and PCB substrates due to its 10⁻⁸Ω level contact resistance stability and 10⁴ temperature cycling reliability. The chemical nickel-palladium (ENEPIG) surface treatment process achieves a dual performance balance of bonding and welding under the condition of 0.1μm ultra-thin gold layer by introducing a palladium layer as a gold-nickel diffusion barrier, reducing material costs by more than 40% compared to traditional electroplating soft gold processes, and driving its penetration rate in high-end fields such as 5G RF modules and automotive electronics by 15% year-on-year.

The reliability chain of electronic packaging presents a typical "short-board effect" - the failure of the bonded solder joint can lead to the paralysis of the entire electronic system. Research data shows that 32% of failures in consumer electronics rework cases are due to wire bonding defects; In automotive electronics, this percentage drops to 8%, but a single point of failure can have fatal consequences. In this paper, the relationship model of "material-process-reliability" is established by deconstructing the formation mechanism of ENEPIG coating and the metallurgical bonding process of gold wire bonding, and the key control points and optimization schemes are output systematically.

1. ENEPIG plating system and bonding process basics

A revolutionary breakthrough in ENEPIG technology lies in the construction of a copper-nickel-palladium-gold gradient functional coating, where the crystal structure and interfacial properties of each layer determine bonding properties.

1.1 Functional synergy mechanism of multi-layer coating

The ENEPIG process uses a five-step chemical deposition to form a functionally complementary coating structure:

Degreasing activation stage: Acidic cleaning agent (pH 2.5-3.5) is used to remove the copper oxide layer (CuO/Cu₂O), and a micro-etching effect (etching amount of 0.3-0.5μm) is introduced to form a microscopic roughness of Rz=1.5-2.0μm, providing a mechanical anchoring basis for the nickel layer. XPS analysis showed that the carbon content of the treated copper surface should be controlled below 5%, otherwise it would lead to a 30% decrease in the binding strength of the nickel layer. Sodium hypophosphite (NaH₂PO₂) was used as a reducing agent to form a Ni-P alloy layer (P content of 8-10wt%) at 85-90°C, and its columnar crystal structure (average grain size 0.5-1μm) should be continuously porosity-free. This layer is both a mechanical support (3-5μm thick) and a major stress zone during bonding, and TEM analysis confirms that grain boundary integrity directly affects bond strength. A 0.1-0.3μm palladium layer was deposited by autocatalytic reduction process to form a dense film with a face-centered cubic structure (density > 98%). Electrochemical impedance spectroscopy (EIS) tests show that the palladium layer can reduce the gold-nickel interdiffusion coefficient from 10⁻¹²cm²/s to 10⁻¹⁶cm²/s, effectively inhibiting the phenomenon of "black disk" (interface blackening caused by nickel corrosion). A 0.05-0.1μm pure gold layer is formed through a semi-displacement-semi-reduction mechanism that maintains good wettability during bonding (contact angle < 30°) while avoiding excessive consumption of the palladium layer. AFM characterization showed that the surface roughness of the gold layer Ra should be controlled at 0.05-0.1μm, and too high would increase the interfacial voidity during bonding. The synergistic effect of this multi-layer structure is reflected in the fact that the nickel layer provides strength support, the palladium layer blocks diffusion, and the gold layer ensures bonding, forming a "strength-barrier-wetting" functional triangle, which is the core reason why ENEPIG technology is superior to traditional chemical nickel-gold (ENIG).

1.2 Metallurgical bonding process of goldwire bonding

Thermosonic bonding achieves atomic level bonding between the gold wire and the pad through the synergistic action of temperature (150-180°C), pressure (50-100mN) and ultrasonic vibration (60-120kHz), and the process can be divided into four stages: the gold wire with a diameter of 25-50μm forms a gold ball (about 2.5 times the diameter of the wire) under the pressure of the splitting knife, and the plastic deformation occurs at the contact point with the pad, and the actual contact area is reduced from 10% of the initial Expand to over 60%. The high-speed camera shows that the deformation rate (<5 μm/ms) needs to be controlled at this stage to avoid micro-cracks. The tangential stress (100-200MPa) generated by ultrasonic vibration breaks the oxide film (Au₂O₃) on the surface of the gold layer, exposing the fresh metal surface. Energy dispersive spectroscopy (EDS) analysis confirmed that the interfacial oxygen content needed to drop below 3% to ensure subsequent diffusion. Under thermal activation, gold atoms diffuse towards the palladium layer (diffusion depth of 0.05-0.1μm), while palladium and nickel atoms migrate towards the gold sphere, forming Au-Ni-Pd ternary solid solution. Diffusion kinetics studies have shown that the diffusion coefficient at 180°C is about 10⁻¹⁴cm²/s, and it takes 15-20ms to form a continuous metallurgical bond. When the diffusion layer thickness reaches 0.1-0.15μm, a metal bonding force (about 10⁻⁸N/atom) is formed at the interface, and the solder joint shear strength is stable in the range of 7-10g. High-resolution TEM shows that the interface transition zone of a qualified solder joint should have no significant holes (porosity < 5%).

This process is extremely sensitive to the quality of the coating - the grain boundary defects of the nickel layer will become the stress concentration point, the discontinuity of the palladium layer will lead to abnormal diffusion of gold and nickel, and the thin gold layer will accelerate the consumption of the palladium layer during the ultrasonic process.

2. Failure mode analysis and key control factors

The root cause of wire bond failure is that the interfacial bond strength is not strong enough to withstand external stresses (bonding process forces or service environment forces), and three typical failure modes and their control paths can be identified by tensile testing (test standard IPC-TM-650 2.4.19).

2.1 Mechanism and control of gold wire pull-off failure (mode 1).

This mode shows that when the gold wire is separated from the pad, there is no obvious damage to the gold layer on the surface of the pad (residual gold layer > 90%), and the tensile value is < 3g, which is mainly due to insufficient wettability of the gold layer or poor diffusion and bonding.

Experimental data show that the bond tension fluctuates significantly (standard deviation > 0.8 g at a palladium thickness of < 0.1 μm), and 30% of the solder joints have a single point failure. By comparing the SEM cross-sectional view of the 0.03μm and 0.3μm palladium layers, it was found that there were obvious grain boundary channels (width >5nm) in the thin palladium layer, resulting in a 5-fold increase in the diffusion rate of gold to the nickel layer after reflow, forming a brittle compound (AuNi₃). ）。 It is recommended to control the palladium thickness at 0.1-0.3μm, where the diffusion blocking efficiency can reach more than 95%. Surface profiles measured with white light interferometer show a 40% reduction in the initial bond contact area and a 25% decrease in tensile force values when the Ra value increases from 0.05μm to 0.2μm. An excessively rough surface (Ra>0.15 μm) leads to a longer time for gold wire to fill gaps and increased ultrasonic energy loss. The optimization scheme included a 10% reduction in micro-etching time (from 60s to 54s) and a 2°C increase in nickel layer deposition temperature to control the overall roughness within the range of Ra=0.08±0.02μm. EDS analysis found a 30% decrease in bond tension when the surface carbon content of the gold layer >8% or the presence of sulfur (>0.5%). The main source of contamination comes from finger contact (including grease) in the post-process process or sulfides in the storage environment. Controls include: "Glove-tweezers" dual protection operation, controlled relative humidity of the storage environment at 40-50%, and nitrogen protection packaging (oxygen content < 5%).

2.2 Root causes and countermeasures of gold-palladium layer pull-off failure (mode 2).

This pattern manifests as carrying part of the gold palladium layer (residual gold layer < 50%) when the gold wire is pulled off, exposing the nickel layer, with a tensile value of 3-5g, and the core problem is corrosion or lattice failure on the surface of the nickel layer. The chemical nickel layer is highly active (surface energy > 800mJ/m²) and forms an oxide layer (NiO) when exposed to air for more than 30 seconds. Experimental data show that the washing time is extended from 60s to 300s, the density of corrosion pits on the surface of the nickel layer is increased from 5 to 30 /μm², and the bonding yield is reduced from 99% to 82%. The optimized washing scheme was a three-stage countercurrent flushing (conductivity < 10μS/cm) with a total time of 90±10s and an air transfer time of < 15s. When the conductivity of the washed water increased from 10μS/cm to 50μS/cm, the corrosion rate of the nickel layer increased by a factor of 3 due to high ion concentrations (especially Cl⁻ >10ppm) can destroy the passivation film of nickel. It is recommended to install an online conductivity monitor (accuracy ±1μS/cm), set 15μS/cm as the warning value, and conduct weekly ICP-MS testing to ensure that the total amount of heavy metal ions < 5ppm. : Palladium tank pH deviations from the norm (4.5±0.2) can lead to uneven deposition rates, resulting in localized thin palladium zones. When the pH dropped to 4.0, the porosity of the palladium layer increased from 1% to 5%, and the risk of corrosion in the nickel layer increased significantly. Controls include pH stability with an automated titration system and deposition rate detection every hour (target 10-15 nm/min).

2.3 Causes and prevention of copper-nickel separation failure (mode 3).

This pattern shows that the solder joint is intact but the nickel layer is separated from the copper base, and the tensile force value is < 2g, which is fundamentally due to the presence of contaminants at the copper-nickel interface, resulting in insufficient bonding force (<3N/cm). The tape residue used in the pre-PCB process (mainly acrylate) is difficult to remove during the micro-etching process, forming a 1-5 μm barrier. EDS analysis showed that the carbon content in the residual glue area was > 30% and the oxygen content was > 20%, completely blocking the copper-nickel binding. Preventive measures include: replacing the tape with a magnetic fixture, adding a plasma cleaning process (power 300W, time 60s) when necessary to ensure a residual detection rate of < 0.1 /m². The Al₂O₃ abrasive (particle size 1-3μm) used in the pretreatment will form a hard isolation point if it is adsorbed by the residual adhesive. By improving the cleaning process (adding 2 bar high-pressure spraying), the particle residue can be reduced from 10 /cm² to less than 1 /cm². Validation was performed under a 60x microscope and no particles of > 5 μm were allowed per plate. Bare copper plates stored for more than 24 hours will form an oxide layer of 5-10 nm, resulting in a 40% decrease in the adhesion of the nickel layer. Solutions include antioxidation treatment (organic protective film) or continuous production with the "copper surface treatment - nickel chemical" process (intervals of < 4 hours).

3. Optimization and interaction of bonding process parameters

The final quality of the wire bond is the result of the combination of material properties and process parameters, and the influence weights of key parameters can be quantified by DOE experiments.

3.1 Synergy between bonding pressure and threading method

A 2-factor 3-level all-factor experiment (pressure: 50/75/100mN; Method: Forward / Reverse), analyze the test data of 1000 solder joints: the average tensile force (7.2g) of the forward punch method (the first solder joint is on the chip side) is 24% higher than that of the reverse punch method (5.8g), because the PCB pad is first stressed during the reverse punch, which can easily lead to microcracks in the nickel layer. The increase in pressure from 50mN to 100mN resulted in a 15% increase in the pull force of the forward stroke (from 6.8g to 7.8g), while the counter-hitting method increased more significantly (from 5.0g to 6.5g), indicating that the increased pressure partially compensated for the shortcomings of the counter-strike style. When the pressure is ≥ 75mN, the standard deviation of the tensile force of the forward stroke method is < 0.3g, and the process capability index Cpk=1.67. The Cpk of the counter-hitting method is only 1.2 even at 100mN pressure, indicating that the process stability of the forward stroke method is better.

Recommended parameter combination: preferential use of forward punching method, bonding pressure 75-100mN (adjusted according to component size); When a counterattack is required, the pressure is set to 100mN and the ultrasonic energy is increased by 10% (from 100mW to 110mW).

Matching relationship between temperature and ultrasound parameters

Further response surface experiments showed a significant interaction between temperature and ultrasonic frequency: at 150°C, the ultrasonic frequency increased from 60kHz to 120kHz, the tensile value increased by 20% (from 6.5g to 7.8g), at 180°C the optimal frequency range narrowed to 80-100kHz, and too high a frequency (120kHz) caused excessive softening of the gold layer and a 10% decrease in tensile force

The optimized thermal ultrasonic parameters are as follows: temperature 165±5°C, ultrasonic frequency 90±5kHz, at this time, the thickness of the intermetallic compound layer at the bonding interface is moderate (0.12μm), which has both strength and toughness.

4. Reliability guarantee system and prospects

The reliability control of ENEPIG wire bonding requires the establishment of a full-process prevention mechanism, forming a closed-loop system of "parameter-detection-verification".

4.1 Key parameter control matrix

4.2 Technological development trends

Developed nanocomposite nickel layers (with W or Mo elements) to replace palladium layers, preliminary results show that their diffusion blocking capacity reaches 80% of that of palladium layers and reduces material costs by 50%. Developed ultrasound-assisted room temperature bonding technology to achieve a tensile force of more than 7g at 80°C, suitable for heat-sensitive applications such as flexible substrates. Machine learning algorithms are used to analyze the force-displacement curves of the bonding process to achieve 100% online prediction of solder joint quality (> 95% accuracy). The development of ENEPIG wire bonding technology has proven that high-reliability connections can be achieved at low cost through micro-control of material design and precise matching of process parameters. In the future, with the development of advanced packaging technologies such as chiplets, this technology will evolve in the direction of finer pitch (<25μm) and higher strength (>10g), providing continuous technical support for semiconductor packaging.