Copper wire bonded IMC growth analysis

In situ TEM study of the growth mechanism of intermetallic compounds at copper wire bonding interfaces

Introduction: Reliability challenges in copper wire bonding technology

In the iterative process of semiconductor packaging interconnect technology, copper wire bonding is gradually shaking the traditional dominance of gold wire bonding with its significant performance and cost advantages. The electrical conductivity (58 MS/m) of copper material is 29% higher than that of gold (45 MS/m), and the thermal conductivity (401 W/m・K) is higher than that of gold (317 W/m・).K) at 26%, while the market price is only 1/5 to 1/10 of gold, making it ideal for high-density packaging applications. However, since 1992, when National Semiconductor Corporation first applied copper wire bonding technology to low-end electronic products, its large-scale industrialization process has always been subject to interface reliability issues - the excessive growth of intermetallic compounds (IMCs) formed at the copper-aluminum (Cu/Al) bonding interface in a high-temperature service environment will cause the contact resistance to rise from the initial 10⁻⁴Ω level to 10⁻²Ω level, and reduce the bonding strength by more than 40%, seriously threatening the long-term stability of the device.

Traditional research methods such as optical microscopy, Micro-XRD, and SEM-EDX have obvious limitations: non-situ characterization methods require observation of multiple groups of samples under different conditions, making it difficult to capture the dynamic process of IMC growth, and individual differences between samples can introduce large errors (usually > 15%). In contrast, in-situ high-resolution transmission electron microscopy (In-situ TEM) technology can track the evolution of interface microstructure in real time within a temperature range of 50-220°C, with a temporal resolution of up to seconds and a spatial resolution of 0.1nm, providing unprecedented observation accuracy for revealing the growth mechanism of IMC. Based on this technology, the formation law and growth kinetics of Cu/Al bonding interface IMC are systematically explored, which provides a quantitative basis for optimizing the bonding process parameters and improving the reliability of the device.

1. Experimental system and technical methods

1.1 Sample preparation process

A 22μm diameter 99.99% pure copper wire was used to connect to a 1.5μm thick Al metal pad (99.5% purity) via a thermal ultrasonic bonding process. The bonding parameters were determined after multiple rounds of optimization:

Bonding pressure: 25-35 gf (approx. 0.25-0.34N) to ensure that the copper wire forms close contact with the aluminum disc without causing damage to the aluminum layer

Ultrasonic power: 120-150 mW, operating frequency 60kHz, energy input controlled in the range of 18-22 mJ

Bonding temperature: 180°C, which not only ensures the diffusion activity of interfacial atoms, but also avoids excessive softening of the aluminum layer

After bonding, epoxy resin molding (model EPON 828) is performed to simulate the actual packaging environment, and the molding process is maintained at 175°C for 4 hours before naturally cooling to room temperature. To meet the requirements of TEM observation, the "grinding-polishing-FIB thinning" three-stage sample preparation process is adopted: the sample thickness is first reduced to 5μm by traditional mechanical grinding, then treated to less than 1μm by chemical mechanical polishing (CMP), and finally finely thinned using the FEI Helios 600i focused ion dual-beam system to obtain a transmission area with a thickness of <100nm, ensuring that the electron beam transmittance > 80%.

1.2 Configuration of in-situ observation system

The experiment uses FEI Titan 80-300 spherical aberration correction transmission electron microscope, accelerating voltage of 300kV, point resolution of 0.19nm, equipped with Gatan 628 single-tilt heat stage sample rod, temperature control accuracy of ±1°C, and adjustable heating rate range of 0.1-10°C/min. To eliminate the effects of temperature drift, each set temperature point is maintained for more than 1 hour to ensure that the hot stage and the sample are in thermal equilibrium. The in-situ heating scheme uses a stepped heating pattern (Table 1) from 50°C to 220°C, with IMC topography changes recorded at each temperature node: low temperature (50-130°C): a step every 20°C for 60-120 minutes, medium temperature (150-175°C): 150°C for 150 minutes, and 175°C extended to 500 minutes to capture the slow growth processHigh temperature section (220°C): Continuous 240 minutes, observe the characteristics of the rapid growth stage, all experiments are performed in a high vacuum environment (<1×10⁻⁵Pa) to avoid scattering of the electron beam by air molecules and oxidation of the sample surface.

Data analysis methods

The TEM images were processed using Gatan DigitalMicrograph software to achieve accurate measurement of IMC thickness through the following steps: Image preprocessing: Background noise reduction (Gaussian filtering, σ=1.0) and contrast enhancement, Boundary recognition: Automatically identifying the interface between the IMC and the substrate using the Canny edge detection algorithmStatistical analysis: 10 measurement positions were randomly selected at each observation point, and the average value was taken as the IMC thickness at that time, with the standard deviation controlled within 5%. Kinetic analysis was based on the parabolic growth model (X²=Kt), data fitting was performed using Origin software, and the activation energy of the reaction was calculated by the Arrhenius equation, and the goodness of fit (R²) needed to be > 0.95 to ensure the reliability of the results.

2. Experimental results and micro mechanism analysis

2.1 Dynamic evolution of IMC phase composition

Initial state observations prior to annealing (Figure 3) showed the presence of discretely distributed granular IMCs at the Cu/Al bonding interface, 20-40 nm in diameter, and island-like. High-resolution TEM (HRTEM) images combined with fast Fourier transform (FFT) analysis showed that these initial phases were mainly Cu₉Al₄(γ₁ phases) with lattice constants a=0.874nm and cubic crystal structures, while detecting a small amount of CuAl₂ (θ phase), accounting for about 15% of the total IMC. This is different from the earlier conclusion that "no IMC generation after bonding" was mainly due to the fact that the experimental sample underwent a 175°C heat treatment during the molding process, which facilitated the initial reaction at the interface.

After 24 hours of stepped annealing, the IMC formed a significant double junction close to the Cu side: a continuously distributed Cu₉Al₄ phase, about 210 nm thick, and the electron diffraction pattern showed that its [110] crystal orientation had a certain orientation relationship with the Cu matrix, close to the Al side: the CuAl₂ phase with a thickness of about 130 nm showed a typical sheet morphology, and a slight stress concentration phenomenon was observed at the interface with the Al matrix. It is worth noting that the other three stable phases (CuAl, Cu₄Al₃, Cu₃Al₂) predicted in the Cu-Al binary phase diagram were not detected throughout the in-situ observation, possibly due to the higher temperatures required for their formation (>300°C) or were metastable under experimental conditions, making it difficult to form observable continuous phase regions.

2.2 Temperature-dependent growth characteristics

In-situ real-time observations capture the evolution of IMC thickness with temperature (Figure 2), showing a significant temperature dependence:

Low temperature zone (<175°C): IMC has almost no obvious growth at 50°C, and the thickness increases slowly in the range of 70-130°C, reaching about 80nm at 130°C, with an average growth rate of 0.012nm/min, and in the medium temperature zone (175°C): the growth rate jumps to 0.083nm/min, and the thickness increases from 110nm to 195nm within 150 minutes, showing an acceleration trend. High temperature zone (220°C): The thickness increases rapidly from 195nm to 340nm in 240 minutes, and the growth rate reaches 0.604nm/min, at which point the Al layer is close to complete depletion

By fitting the relationship between IMC thickness and time at different temperatures (Fig. 5a), it was confirmed that its growth was in line with the parabolic law (X²∝t), indicating that the process is controlled by diffusion, and the activation energy of Cu atoms diffusing through the IMC layer to the Al side is a key factor in determining the growth rate. The thickness is further linearly fitted to the square root of time (Fig. 5b) to obtain the reaction rate constant K at each temperature, which provides basic data for subsequent kinetic analysis.

2.3 Calculation of growth kinetics parameters

The reaction rate constant of IMC calculated based on the experimental data is shown in Table 2, and the K value is 2.1×10⁻¹⁸ cm²/s at 150°C, increases to 5.8×10⁻¹⁸ cm²/s at 175°C, and reaches 2.3×10⁻¹⁷ cm²/s at 220°C, showing an exponential growth trend. lnK to 1/T was plotted (Fig. 6) to obtain a linear fitting equation: lnK = -11054.3/T + 18.8, correlation coefficient R²=0.992.

According to the Arrhenius equation K=K₀exp (-Q/RT), the activation energy Q=23.8 kcal/mol (about 99.6 kJ/mol) is calculated

, refers to the prefactor K₀=1.645×10⁻³ cm²/s Compared with the results of the non-in situ study (Table 3), the activation energy obtained in this experiment is between the SEM in situ measurement (26 kcal/mol) and the TEM in situ measurement (Cu₉Al₄ 18.06 kcal/mol, CuAl₂ 14.49 kcal/mol). This difference is mainly due to the fact that the in situ technique is able to capture the continuous growth process of the IMC, avoiding the errors introduced by sample preparation in the non-in-situ method (typically ±15%), so the kinetic parameters obtained in this study have greater confidence.

The final formula for in-situ growth is:

X² = 1645×10⁻⁸ exp (-11054.3/T)・t

where X is the IMC thickness (cm), T is the absolute temperature (K), and t is the annealing time (s)

3. Discussion: technical value and process enlightenment

3.1 Methodological advantages of in-situ technology

Compared with traditional in situ studies, the in situ TEM technology used in this experiment demonstrates three significant advantages:

Dynamic tracking capability: The transition process from Cu₉Al₄ to CuAl₂ phase was observed in real time for the first time, and it was found that 175°C was the turning point in the growth rate of the two phases

Improved data accuracy: Systematic errors caused by sample-to-sample differences are eliminated by continuous measurement of the same batch of samples, reducing the activation energy calculation error from ±3 kcal/mol to ±1 kcal/mol

Mechanism reveal depth: The grain boundary migration phenomenon during IMC growth was directly observed, confirming that the diffusion path mainly proceeded along the [100] crystal direction of the Cu₉Al₄ phase

These findings revise the previous perception that "Cu/Al IMC is mainly CuAl₂", clarify the dominance of the Cu₉Al₄ phase in the low-temperature stage, and provide a microscopic basis for accurate prediction of interface evolution.

Guiding significance for the optimization of the bonding process

Based on the experimentally obtained growth kinetics formulas, the parameters of the copper wire bonding process can be quantitatively optimized: it is recommended to control the annealing temperature of the mold seal below 150°C, at which time the annual growth thickness of IMC is < 50nm, which can ensure a service life of more than 5 years, appropriately increase the ultrasonic power (140-150mW) in the bonding stage, promote the uniform distribution of the initial Cu₉Al₄ phase, and reduce the anisotropy of later growthAfter applying these suggestions, the high-temperature storage (150°C) life of its copper bonding products was extended from 1000 hours to more than 3000 hours, and the change rate of contact resistance was controlled within 8%.

3.3 Establishment of reliability prediction model

Combining the IMC growth data from this study with the device failure criterion (typically an IMC thickness of > 500nm or a > 20% increase in contact resistance) can be used to establish a reliability prediction model:

At 85°C operating temperature, the estimated failure time > 10 years

At 125°C high temperature, the life is reduced to 3-5 years

If there is a local hot spot (> 175°C), early failure may occur within 1 year

This model has been adopted by an automotive electronics manufacturer for reliability evaluation of its automotive MCU, using thermal simulation to locate potential hot spots and take thermal optimization measures, resulting in AEC-Q100 Grade 2 qualification.

conclusion

In this study, the growth mechanism and kinetic characteristics of IMC at the Cu/Al copper bonding interface were systematically revealed by in-situ high-resolution transmission electron microscopy, and the main conclusions are as follows:

1. There are granular IMCs at the initial interface after bonding, with Cu₉Al₄ (about 85%) and a small amount of CuAl₂ (about 15%), with a diameter of 20-40nm.

During the annealing process, the IMC formed a double-layer structure, with the Cu₉Al₄ phase near the Cu side and the CuAl₂ phase near the Al side, and no other Cu-Al intermediate phases were detected.

The growth of IMC conformed to the parabolic law (X²=Kt), and the reaction activation energy was calculated to be 23.8 kcal/mol, and the growth formula based on in situ data was established.

175°C is the critical temperature of the growth rate, below which the growth rate is slow, and above this temperature the growth rate is significantly accelerated.

The research results provide a quantitative basis for the optimization of copper wire bonding process parameters, reliability evaluation and chip heat dissipation design, and help promote the large-scale application of copper wire bonding technology in the field of high-end electronic packaging. Future work will focus on the inhibitory effect of multi-alloying on IMC growth to further improve interface stability.