Failure Mechanisms

[As of April, 2011]

Metallization

There are two kinds of failures related to metallization: failures in the substrate and metal electrode contact area and failures of metal interconnections.

Contact Failure

It is well known that scaling rules are applied to achieve miniaturization. For this reason, the junction depth is designed to be shallow in LSI and VLSI chips, which are highly miniaturized. However, Al metal and Si react during high-temperature processing to form alloy spikes in shallow junction contact areas due to the static electricity and surges applied while a device is being used, resulting in Al spikes reaching the junction surface as well as junction breakdown. To avoid this problem measures such as adding Si as an impurity into the Al electrode to suppress Al diffusion by the Al-Si reaction, and establishing a barrier metal under the Al metal are used.

Through-Hole Contact (Al-to-Al Contact in a Multilayered Al Structure)

Similar to contacts, scaling rules are also applied to through-hole contacts.
Therefore, open failures caused by Al step coverage at the step can occur. Also, when the through-hole contact diameter reaches 1 µm or less, contact formation using Al metal alone becomes more difficult, necessitating the use of an embedding process.

Al Metal Corrosion

Al metal corrosion is a critical problem for reliability in resin-encapsulated devices. Some cases have also been reported for hermetically sealed devices as well.^{Note 1} The following provides a brief description of the Al corrosion mechanism in resin-encapsulated devices.

1. General Model of Al Corrosion
   Figure 1 shows a diagram of a resin-encapsulated device. In general, plastic materials have moisture permeability and absorption properties by their very nature. Here, “permeability” refers to the ease of water passage, and “absorption properties” refers to the ability to absorb moisture. Resin has various kinds of ionic impurities that are introduced during the manufacturing processes. When the resin absorbs moisture, the ionic impurities are eluted and reach the internal chip surface. Depending on the size of the bias applied to fulfill the operational function of the device, positive or negative ions and moisture reach the Al metal surface by passing through micro-defects in the passivation film, thereby resulting in Al electrochemical reaction. The result is a fatal failure such as an increase in Al metal resistance or an open. This is the general mechanism of Al metal corrosion failure.
2. Moisture Penetration Route
   The primary cause of Al metal corrosion is the intrusion of external moisture. Moisture is defined as water vapor in the atmosphere. There are two penetration routes as shown in Figure 1. One is through the gaps in the interface of the lead frame and resin, and the other is through the bulk and is dependent on the moisture permeability and absorption properties of the resin. It is difficult to theorize which of these routes is more dominant since they depend on factors such as operating environment conditions and package type. Based on experimental data obtained through investigations using a moisture-sensitive chip encapsulated in a resin package, moisture penetration through the bulk can be approximated by a diffusion model.^{Note 2}

   

   Figure 1 Schematic Diagram of Resin-Encapsulated Device

3. Dependence on Applied Bias
   As a result of temperature and humidity tests conducted under acceleration conditions of 80°C, 90% RH for a PSG passivation product with bias voltage varied at 5 V, 10 V, 15 V, 20 V and 25 V, dependence of Mean Time To Failure (MTTF) on applied voltage was obtained as shown in Figure 2. From the figure, it is apparent that the MTTF decreases with increased bias voltage.
   Al corrosion that results from electrochemical reaction has different corrosion modes depending on bias polarity.
   This means that the failure mechanism varies with the polarity. The biased wiring with the relatively higher potential is called the anode, and that with the lower potential is called the cathode. Al corrosion occurring on the anode and cathode sides is referred to as anodic corrosion and cathodic corrosion, respectively. Cathodic corrosion is predominant with the Al and Al-Si metals generally used, but cracks or pinholes in the passivation can cause anodic corrosion due to impurity ions (such as Cl).
   Cathodic corrosion normally occurs in the crystal grain boundary of the Al film and appears dark when observed through an optical microscope. On the other hand, anodic corrosion is accompanied by a significant expansion of Al, sometimes causing cracks in the passivation which can propagate. In some instances, it appears as if Al is missing when the device is observed under an optical microscope. However, it sometimes remains as transparent Al₂ O₃ based on analyses using electron probe micro analyzer (EPMA) or Auger electron spectroscopy (AES).
   

   

   Figure 2 Dependency on Voltage in Humidity Resistance Acceleration Test

4. Dependence of PSG Passivation on Phosphorous Concentration
   It was previously described how the use of PSG film containing phosphorous is used in top passivation in order to subject the external ions to the gettering effect. However, an excessively high phosphorous concentration will significantly increase the potential for fatal Al metal corrosion. Phosphorous related corrosion is cathodic for Al or Al-Si metal and occurs as follows:^{Note 3, 4}
   First, when PSG absorbs moisture, P₂ O₅ in the PSG is eluted to form phosphoric acid, increasing the H⁺ ion concentration. As a result, H⁺ ions are attracted to the surface of the Al metal on the cathode side, allowing corrosion to progress according to the following reactions:
   Al + 3H⁺ → Al³⁺ + 3/2H₂ ↑
   Al³ + 3OH⁻ → Al (OH) ₃
   and
   Al (OH) ₃ + OH⁻ → AlO₂⁻ + 2H₂O
   Figure 3 shows the relative life values versus phosphorous density in PSG which causes cathodic corrosion for Test Element Groups (TEGs) and LSIs. The figure shows that the life shift is sensitive to changes in phosphorous concentration. Recently, however, the use of moisture resistant film (such as SiN) in top passivation to improve moisture resistance is eliminating this type of failure.
   

   

   Figure 3 Dependence of Relative Life on Phosphorous Density Due to Al Metal Corrosion
   (Experimentally Controlled Example)

5. Other Unexpected Events
   In addition to the above items, factors contributing to Al corrosion include some that are attributable to the manufacturing process, such as seal leaks in hermetically sealed devices, contamination, and flaws in passivation due to improper handling; and those contributable to customers, such as contamination during handling, penetration of flux (including Cl) during soldering, and penetration of contaminants through the resin and lead interface, causing corrosion of internal metal when moisture condenses.

Electromigration

It is well-known that introducing a high current to the metal wiring in an IC device can cause an open fault in the metal wiring and subsequent device failure. This phenomenon is called electromigration and is becoming an important failure mechanism as the scaling of ICs gets larger and miniaturization advances, such as the case with VLSI.
The following describes the mechanism of electromigration in thin film. If a large current flows in a thin film, a force is applied to the metal atoms due to the electron wind force. As a result, Al atoms diffuse in the direction of the electron flow (from cathode to anode), forming a void on the cathode side, and a hillock or whisker on the anode side.
An open failure on thin film can occur when the mass transfer in the metal becomes variable. This variable mass transfer is caused by variations in temperature or current density, or by a variable shift in metal ions such as a triple point of Al grain boundaries.
Causes include, for example:

1. Variable grain size^{Note 5}
2. Temperature gradient due to heat generation inside the device^{Note 6}
3. Metal in contact with other material^{Note 7}

The electromigration life of the thin film is generally expressed as the Median Time to Failure (MTF), establishing the following relationship:^{Note 8}

where MTF is Median Time to Failure, J is current density, n is a constant related to the current density, Ea is the activation energy, T is the absolute temperature, k is Boltzmann’s constant and A is a constant related to the material, structure and size of the metal. From this expression it can be seen that MTF increases with a decrease in current density or temperature. Also, the life distribution is in accordance with a logarithmic normal distribution with a narrow variation. Figure 3 shows an example of how to find the activation energy related to Al metal. Activation energy is metal width dependent and as the width thickens the value approaches 0.6. This is believed to be due to the shift from bulk diffusion to grain boundary diffusion by the Al metal.

Cu Metal Electromigration

With the advances in miniaturization in the silicon process, performance degradation due to increased metal resistance and metal-to-metal capacity has become problematic. To resolve this problem, the Cu metal process has been applied for the first time starting from the 0.13 generation. The Cu metal formation process employs a process called "damascene." The flow of this process is shown in Figure 5.

In this manner, electromigration failure is a potential failure mechanism for both conventional Al metal and the intrinsically different Cu metal from the standpoints of material and metal formation process. The MTF formula can also be expressed in the same manner as that for the Al metal.
However, because the Cu melting point (1083°C) is higher than the Al melting point (660°C), the Cu metal is believed to exhibit better resistance to electromigration in comparison to the Al metal. Figure 6 shows an example of the results obtained when comparing the differences in electromigration resistance that result from the differences in metal material, using the same design rules for each product. It is evident that the Cu metal exhibits a life that is approximately one digit greater than that of the Al metal at the point of MTF for electromigration failure.

The electromigration in Cu metal is characterized by the fact that there are reports indicating that the dominant diffusion route is grain boundary diffusion, like the Al metal, as well as those indicating that it is interface diffusion. A clear explanation of the failure mechanism, therefore, has not yet been concluded. Because Cu metal readily oxidizes, case examples exist in which failure to develop appropriate procedures and optimize conditions in the manufacturing process significantly degraded reliability.

Stress Migration ^{Note 9, 10, 11}

Stress migration is a failure mechanism where open failures occur simply due to extended exposure to a high-temperature environment.^{Note 12, 13} In general, LSI metal is subjected to high-temperature heat treatment during formation of the interlayer insulator after metal formation. Although stress does not occur on the metal during this high-temperature period, stress is generated in the metal after cooling due to the mismatch in thermal expansion coefficients between the metal and interlayer insulator or passivation film. The residual stress and subsequently applied heat cause void generation and diffusion in the metal and, in consequence, metal opens and open faults in vias (through-holes for connections between overlapping geometries on two adjacent routing layers).
This event is called stress migration since it is induced by internal stress.
Stress migration countermeasures, such as the addition of Cu to the Al metal, the use of barrier metal under the Al metal, and reducing passivation film stress, are taken to minimize this effect.
Acceleration of stress migration failure due to temperature does not uniformly occur due to a combined mechanism of void diffusion and stress relief. However, the apparent activation energy at 125°C or below is 0.7 eV for Al-Si and Al-Cu, and 0.9 eV for Al-Si-Cu.
In addition, stress migration tends to occur more readily in processes that use Cu metal in comparison to those that use Al metal since the Cu grain size tends to be smaller than that of Al. This type of stress migration, similar to the Al metal process, can be suppressed by the existence of a barrier metal. However, since structural placement of the barrier metal in the via is not possible, the stress migration that occurs in the via is viewed as problematic.
In general, the via readily becomes the singular point of concentrated mechanical stress, causing voids in the vicinity of the via to readily grow. There are many reported cases of this type of failure.^{Note 14, 15, 16} Figure 7 shows a case of void observation.

To suppress Cu process stress migration failure, in particular void occurrence and growth, development of a reduced stress process as described for the Al metal process above, application of an interlayer insulator with an expansion coefficient approximate to that of the Cu metal, and use of a barrier metal with a high melting point, such as Ti or Ta, have been confirmed as effective methods.
Countermeasures incorporated at the metal design phase are also effective. It is also possible to suppress void generation by diffusing the stress at locations where stress collects in the metal.^{Note 17} For instance, the stress that collects in a via can be alleviated and subsequent stress migration can be suppressed by making the metal volume (metal width and film thickness), which is the void supply source, no larger than necessary or by creating multiple vias in the area of connection with a large surface area metal.
The speed of progression of stress migration can be expressed by the product of the stress component and diffusion component as follows:

Where, R indicates the speed of stress migration progression, C indicates a coefficient, T₀ indicates the metal formation temperature or interlayer film formation temperature, T indicates the test temperature, N indicates the acceleration coefficient, Ea indicates the activation energy, and k indicates Boltzmann’s constant. Both the stress component and diffusion component depend on test temperature. The stress component increases as the test temperature lowers in comparison with the metal formation temperature, and the diffusion component increases as the test temperature increases. The speed of progression is expressed as the product of these two components, resulting in a peak value at a certain temperature.

Note 1: Bibliography. David B. Willmott; "Investigation of Metalization Failures of Glassed Sealed Ceramic Dual in Line Integrated Circuits," (1900), p. 158

Note 2: Bibliography. Umezu, Kunihiro; "Degradation of Plastic Encapsulated Semiconductor Parts due to Moisture," Reliability study at electronic and communication society, R79-56, (1979), p. 75

Note 3: Bibliography. W. M. Paulson and R. W. Kirk; "The Effects of Phosphorus Doped Passivation Glass on Corrosion of Aluminum," 12th Annual Proc. Rel. Phys., (1974), p. 172

Note 4: Bibliography. S. P. Sim and R. W. Lawson; "The Influence of Plastic Encapsulants and Passivation Layers on the Corrosion of Thin Aluminum Films Subjected to Humidity Stress," 17th Annual Proc. Rel. Phys., (1979), p. 103

Note 5: Bibliography. E. Nagasawa; "Electromigration of Sputtered Al-Si Alloy Films," Proc. of Annual Rel. Phys. Symp., (1978), p. 64

Note 6: Bibliography. Francois M. D'Hearle; "Electromigration and Failure Electronics: An Introduction," Proc. of the IEEE, Vol. 59, No. 10, (1971)

Note 7: Bibliography. J. R. Black; "Electromigration of Al-Si Alloy Films," Proc. Annual Rel. Phys. Symp., (1978), p. 233

Note 8: Bibliography. M. C. Shine and F. M. D'Heurle; "Activation Energy for Electromigration in Aluminum Films Alloyed Copper," IBM J. Res. Dev., Vol. 15, No. 5, (1971), p. 378

Note 9: Bibliography. N. Owada, K. Hinoda, M. Horiuchi, T. Nishida, K. Nakata and K. Mukai; “Stress Induced Slit-like Void Formation in a Fine-Pattern Al-Si Interconnect during Aging Test,” IEEE 2nd International VLSI Multilevel Interconection Conference Proc., (1985), p. 173

Note 10: Bibliography. Tsuda; “New Mechanism Proprosal for Al Open Failures,” Nikkei Microdevice, September (1985), p. 50

Note 11: Bibliography. A.Tezaki, et al.; “Measurement of Three Dimentional Stress and Modeling of Stress Induced Migration Failure in Aluminium Interconnects,” IRPS, (1990), p. 221

Note 12: Bibliography. J. Klema, et al.; Proc. of IRPS (1984), pp. 1-5

Note 13: Bibliography. J. Curry, et al.; Proc. of IRPS, (1984), pp. 6-8

Note 14: Bibliography. T. Ohshima, et al.; Proc. of IEDM, (2002), pp. 61-68

Note 15: Bibliography. M. Kawano, et al.; Proc. of IITC, (2003), pp. 210-212

Note 16: Bibliography. K. Yoshida., et al.; Proc. of IEDM, (2002), pp. 247-250

Note 17: Bibliography. H. Yamamoto, et al.; IEDM Tech Dig., (1987), p. 205

• Products
  • Reliability Information
    • Semiconductor Device Reliability
      • Failure Mechanisms
        • Metallization
• Applications
• Company Profile
• News and Event
• Support
• Environmental Activities of previous Semiconductor company
• Environmental Activities of previous Storage Products Company

• Terms of Use
• Site Map